Love the nano-texture on the Studio Display, but my MacBooks have always suffered from finger oil rubbing the screen from the keys. Fingerprint oil on nano-texture sounds like a recipe for disaster.
For my current laptop, I finally broke down and bought a tempered glass screen protector. It adds a bit of glare, but wipes clean — and for the first time I have a one-year-old MacBook that still looks as good as new.
I put a thin screen cleaner/glasses cleaner cloth on the keyboard whenever I close the lid. That keeps the oils off the screen as well as prevents any pressure or rubbing from damaging the glass.
My one concern is that nano-texture apple displays are a little more sensitive to damage, and even being super careful with my MBPs I get the little marks from the keyboard when you carry the laptop with your hand squeezing the lid and bottom (a natural carry motion).
A piece of paper also kind of works, but it's a bit ridiculous to constantly keep that around with you and try to reach 95% consistency. Eventually I stopped trying because it was just more infrastructure to protect the thing than I'd prefer.
I'm not sure we can leverage the neural cores for now, but they're already rather good for LLMs, depending on what metrics you value most.
A specced out Mac Studio (M2 being the latest model as of today) isn't cheap, but it can run 180B models, run them fast for the price, and use <300W of power doing it. It idles below 10W as well.
It seems they also update the base memory on MacBook Air:
> MacBook Air: The World’s Most Popular Laptop Now Starts at 16GB
> MacBook Air is the world’s most popular laptop, and with Apple Intelligence, it’s even better. Now, models with M2 and M3 double the starting memory to 16GB, while keeping the starting price at just $999 — a terrific value for the world’s best-selling laptop.
Ohh, good catch. Sneaking that into the MBP announcement. I skimmed the page and missed that. So a fourth announcement couched within the biggest of the three days.
The only older configs that Apple sells are the M2 and M3 Airs, which were bumped. Everything else is now on M4, or didn't have an 8gb base config (Mac Studio, Mac Pro)
Wow, I didn't expect them to update the older models to start at 16GB and no price increase. I guess that is why Amazon was blowing the 8GB models out at crazy low prices over the past few days.
Costco was selling MB Air M2 8 GB for $699! Incredible deal.
I’ve been using the exact model for about a year and I rarely find limitations for my typical office type work. The only time I’ve managed to thermally throttle it has been with some super suboptimal Excel Macros.
It would make more sense to discontinue the smaller model along with some other updates to the line. Or in other words, Air won't receive any other updates this week unfortunately.
They did? The tweet that announced stuff from the head of marketing did not mention 3 days.
That said, I believe you. Some press gets a hands-on on Wednesday (today) so unless they plan to pre-announce something (unlikely) or announce software only stuff, I think today is it.
That's disappointing. I was expecting a new Apple TV because mine needs replacement and I don't really feel inclined to get one that's due for an upgrade very soon.
The current-gen Apple TV is already overpowered for what it does, and extremely nice to use. I can think of very few changes I would like to see, and most of them are purely software.
Mine has 128GB of onboard storage... but Apple still bans apps from downloading video, which annoys me.
The streaming apps virtually all support downloading for offline viewing on iPhone, but the Apple TV just becomes a paperweight when the internet goes out, because I'm not allowed to use the 128GB of storage for anything.
If they're not going to let you use the onboard storage, then it seems unlikely for them to let you use USB storage. So, first, I would like them to change their app policies regarding internal storage, which is one of the purely software improvements I would like to see.
I use a dedicated NAS as a Plex server + Plex app on Apple TV itself for local streaming, which generally works fine. Infuse app can also index and stream from local sources.
But there are some cases like e.g. watching high-res high-FPS fractal zoom videos (e.g. https://www.youtube.com/watch?v=8cgp2WNNKmQ) where even brief random skipped frames from other things trying to use WiFi at the same time can be really noticeable and annoying.
"This is a huge week for the Mac, and this morning, we begin a series of three exciting new product announcements that will take place over the coming days," said Apple's hardware engineering chief John Ternus, in a video announcing the new iMac.
See that’s the thing. Given that somehow you need 1TB to get the matte screen, I feel like Apple is using it as a way to upsell. It would indicate that perhaps Apple won’t offer a matte MacBook Air.
I think spec-wise the Air is good enough for almost everyone who isn't doing video production or running local LLMs, I just wish it had the much nicer screen that the Pro has. But I suppose they have to segregate the product lines somehow.
Well, the issue for me with memory on these new models is that for the Max, it ships with 36GB and NO expandable memory option. To get more memory that's gated behind a $300 CPU upgrade (plus the memory cost).
It'll be interesting to see the reaction of tech commentators about this. So many people have been screaming at Apple to increase the base RAM and stop price gouging their customers on memory upgrades. If Apple Intelligence is the excuse the hardware team needed to get the bean counters on board, I'm not going to look a gift horse in the mouth!
It wouldn't surprise me if people typically use more storage on their phone than their computer. The phone should probably have a higher base storage than the base storage of their laptops.
Extremely interesting point. My initial reaction to your comment is that it is a crazy thing to say, but the more I think about it the more I agree with you. On my phone is where I have tons of 4k 30/60FPS videos, high resolution photos (with live), and videos downloaded on Netflix and YouTube.
On my Mac I don't have any of these things, it's mostly for programming and some packages. I'm almost always connected to Wi-Fi (except on planes) so I don't really need any photos or videos.
The only people that I see have high storage requirements on Macs are probably video/media creators? As a programmer I'm totally fine with 512GB, but could probably live with 256GB if I wanted to be super lean.
I've seen a lot of people complaining about 8GB but honestly my min spec M1 Air has continued to be great. I wouldn't hesitate to recommend a refurb M1 8GB Air for anyone price conscious.
Yeah, this one caught me off guard. We just purchased a MacBook Air in the last month and if we bought the same one now, we would save $200. Apple support would not price match/correct that, so we will be returning it and purchasing anew.
Thanks to your comment. I persuaded my friend who purchased an M3 Air 24GB recently and we got 200$ back (Remuneration for price drop valid for 14 days after the date of DELIVERY) where we live
No support for M3 or M4 powered machines currently.
> All Apple Silicon Macs are in scope, as well as future generations as development time permits. We currently have support for most machines of the M1 and M2 generations.[^1][^2]
Have anyone tried it recently, specifically the trackpad? I tried the Fedora variant a few months ago on my M1 Macbook and it was horrible to use the trackpad, it felt totally foreign and wrong.
I feel you, but Apple's trackpad prowess is not an easy thing to copy. It's one of those things I never expect anyone else to be able to replicate the level of deep integration between the hardware and software.
It's 2024, and I still see most Windows users carrying a mouse to use with their laptop.
btw, there is a recent interview with an Asani dev focusing on GPUs, worth a listen for those interested in linux on apple silicon. The reverse engineering effort required to pin down the GPU hardware was one of the main topics.
For many years I treated Windows or macOS as a hypervisor - if you love Linux but want the Mac hardware, instant sleep & wake, etc, putting a full screen VM in Parallels or similar is imo better than running Linux in terms of productivity, although it falls short on “freedom”.
I do the same thing, but there are two big caveats:
1. Nested virtualization doesn't work in most virtualization software, so if your workflow involves running stuff in VMs it is not going to work from within another VM. The exception is apparently now the beta version of UTM with the Apple Virtualization backend, but that's highly experimental.
2. Trackpad scrolling is emulated as discrete mouse wheel clicks, which is really annoying for anyone used to the smooth scrolling on macOS. So what I do is use macOS for most browsing and other non-technical stuff but do all my coding in the Linux VM.
$ swift repl
Welcome to Apple Swift version 6.0.2 (swiftlang-6.0.2.1.2 clang-1600.0.26.4).
Type :help for assistance.
1> import Virtualization
2> VZGenericPlatformConfiguration.isNestedVirtualizationSupported
$R0: Bool = false
This is still half the speed of a consumer NVidia card, but the large amounts of memory is great, if you don't mind running things more slowly and with fewer libraries.
A 24gb model is fast and ranks 3.
A 70b model is slow and 8.
A top tier hosted model is fast and 100.
Past what specialized models can do, it's about a mixture/agentic approach and next level, nuclear power scale. Having a computer with lots of relatively fast RAM is not magic.
Thanks, but just to put things into perspective, this calculation has counted 8 channels which is 4 DIMMs and that's mostly desktops (not dismissing desktops, just highlighting that it's a different beast).
Desktops are two channels of 64 bits, or with DDR5 now four (sub)channels of 32 bits; either way, mainstream desktop platforms have had a total bus width of 128 bits for decades. 8x64 bit channels is only available from server platforms. (Some high-end GPUs have used 512-bit bus widths, and Apple's Max level of processors, but those are with memory types where the individual channels are typically 16 bits.)
The vast majority of any x86 laptop or desktops are 128 bits wide. Often 2x64 bit channels up till last year or so, now 4x32 bit DDR5 in the last year or so. There are some benefits to 4 channels over 2, but generally you are still limited by 128 bits unless you buy a Xeon, Epyc, or Threadripper (or Intel equiv) that are expensive, hot, and don't fit in SFFs or laptops.
So basically the PC world is crazy behind the 256, 512, and 1024 bit wide memory busses apple has offered since the M1 arrived.
Was this example intended to describe any particular device? Because I'm not aware of anything that operates at 8800 MT/s, especially not with 64-bit channels.
That seems unlikely given the mismatched memory speed (see the parent comment) and the fact that Apple uses LPDDR which is typically 16 bits per channel. 8800MT/s seems to be a number pulled out of thin air or bad arithmetic.
Heh, ok, maybe slightly different. But apple spec claims 546GB/sec which works out to 512 bits (64 bytes) * 8533. I didn't think the point was 8533 vs 8800.
I believe I saw somewhere that the actual chips used are LPDDR5X-8533.
Effectively the parents formula describes the M4 max, give or take 5%.
Fewer libraries? Any that a normal LLM user would care about? Pytorch, ollama, and others seem to have the normal use cases covered. Whenever I hear about a new LLM seems like the next post is some mac user reporting the token/sec. Often about 5 tokens/sec for 70B models which seems reasonable for a single user.
Is there a normal LLM user yet? Most people would want their options to be as wide as possible. The big ones usually get covered (eventually), and there are distinct good libraries emerging for Mac only (sigh), but last I checked the experience of running every kit (stable diffusion, server-class, etc) involved overhead for the Mac world.
> This is still half the speed of a consumer NVidia card, but the large amounts of memory is great, if you don't mind running things more slowly and with fewer libraries.
But it has more than 2x longer battery life and a better keyboard than a GPU card ;)
It's not the memory being unified that makes it fast, it's the combination of the memory bus being extremely wide and the memory being extremely close to the processor. It's the same principle that discrete GPUs or server CPUs with onboard HBM memory use to make their non-unified memory go ultra fast.
I thought it meant that both the GPU and the CPU can access it. In most systems, GPU memory cannot be accessed by the CPU (without going through the GPU); and vice versa.
CPUs access GPU memory via MMIO (though usually only a small portion), and GPUs can in principle access main memory via DMA. Meaning, both can share an address space and access each other’s memory. However, that wouldn’t be called Unified Memory, because it’s still mediated by an external bus (PCIe) and thus relatively slower.
No, unified memory usually means the CPU and GPU (and miscellaneous things like the NPU) all use the same physical pool of RAM and moving data between them is essentially zero-cost. That's in contrast to the usual PC setup where the CPU has its own pool of RAM, which is unified with the iGPU if it has one, but the discrete GPU has its own independent pool of VRAM and moving data between the two pools is a relatively slow operation.
An RTX4090 or H100 has memory extremely close to the processor but I don't think you would call it unified memory.
I don't quite understand one of the finer points of this, under caffeinated :) - if GPU memory is extremely close to the CPU memory, what sort of memory would not be extremely close to the CPU?
I think you misunderstood what I meant by "processor", the memory on a discrete GPU is very close to the GPUs processor die, but very far away from the CPU. The GPU may be able to read and write its own memory at 1TB/sec but the CPU trying to read or write that same memory will be limited by the PCIe bus, which is glacially slow by comparison, usually somewhere around 16-32GB/sec.
A huge part of optimizing code for discrete GPUs is making sure that data is streamed into GPU memory before the GPU actually needs it, because pushing or pulling data over PCIe on-demand decimates performance.
> CPU trying to read or write that same memory will be limited by the PCIe bus, which is glacially slow by comparison, usually somewhere around 16-32GB/sec.
If you’re forking out for H100’s you’ll usually be putting them on a bus with much higher throughput, 200GB/s or more.
> So for example if you have a server with 16 DDR5 DIMMs (sticks) it equates to 1,024 GB/s of total bandwidth.
Not quite as it depends on number of channels and not on the number of DIMMs. An extreme example: put all 16 DIMMs on single channel, you will get performance of a single channel.
If you're referring to the line you quoted, then no, it's not wrong. Each DIMM is perfectly capable of 64GiB/s, just as the article says. Where it might be confusing is that this article seems to only be concerning itself with the DIMM itself and not with the memory controller on the other end. As the other reply said, the actual bandwidth available also depends on the number of memory channels provided by the CPU, where each channel provides one DIMM worth of bandwidth.
This means that in practice, consumer x86 CPUs have only 128GiB/s of DDR5 memory bandwidth available (regardless of the number of DIMM slots in the system), because the vast majority of them only offer two memory channels. Server CPUs can offer 4, 8, 12, or even more channels, but you can't just install 16 DIMMs and expect to get 1024GiB/s of bandwidth, unless you've verified that your CPU has 16 memory channels.
Apple is using LPDDR5 for M3. The bandwidth doesn't come from unified memory - it comes from using many channels. You could get the same bandwidth or more with normal DDR5 modules if you could use 8 or more channels, but in the PC space you don't usually see more than 2 or 4 channels (only common for servers).
Unrelated but unified memory is a strange buzzword being used by Apple. Their memory is no different than other computers. In fact, every computer without a discrete GPU uses a unified memory model these days!
Eh… not quite. Maybe on an Instinct. Unified memory means the CPU and CPU means they can do zero copy to use the same memory buffer.
Many integrated graphics segregate the memory into CPU owned and GPU owned, so that even if data is on the same DIMM, a copy still needs to be performed for one side to use what the other side already has.
This means that the drivers, etc, all have to understand the unified memory model, etc. it’s not just hardware sharing DIMMs.
The new idea is having 512 bit wide memory instead of PC limitation of 128 bit wide. Normal CPU cores running normal codes are not particularly bandwidth limited. However APUs/iGPUs are severely bandwidth limited, thus the huge number of slow iGPUs that are fine for browsing but terrible for anything more intensive.
So apple manages decent GPU performance, a tiny package, and great battery life. It's much harder on the PC side because every laptop/desktop chip from Intel and AMD use a 128 bit memory bus. You have to take a huge step up in price, power, and size with something like a thread ripper, xeon, or epyc to get more than 128 bit wide memory, none of which are available in a laptop or mac mini size SFF.
Memory interface width of modern CPUs is 64-bit (DDR4) and 32+32 (DDR5).
No CPU uses 128b memory bus as it results in overfetch of data, i.e., 128B per access, or two cache lines.
AFAIK Apple uses 128B cache lines, so they can do much better design and customization of memory subsystem as they do not have to use DIMMs -- they simply solder DRAM to the motherboard, hence memory interface is whatever they want.
> Memory interface width of modern CPUs is 64-bit (DDR4) and 32+32 (DDR5).
Sure, per channel. PCs have 2x64 bit or 4x32 bit memory channels.
Not sure I get your point, yes PCs have 64 bit cache lines and apple uses 128. I wouldn't expect any noticeable difference because of this. Generally cache miss is sent to a single memory channel and result in a wait of 50-100ns, then you get 4 or 8 bytes per cycle at whatever memory clock speed you have. So apple gets twice the bytes per cache line miss, but the value of those extra bytes is low in most cases.
Other bigger differences is that apple has a larger page size (16KB vs 4KB) and arm supports a looser memory model, which makes it easier to reach a large fraction of peak memory bandwidth.
However, I don't see any relationship between Apple and PCs as far as DIMMS. Both Apple and PCs can (and do) solder dram chips directly to the motherboard, normally on thin/light laptops. The big difference between Apple and PC is that apple supports 128, 256, and 512 bit wide memory on laptops and 1024 bit on the studio (a bit bigger than most SFFs). To get more than 128 bits with a PC that means no laptops, no SFFs, generally large workstations with Xeon, Threadrippers, or Epyc with substantial airflow and power requirements
FYI cache lines are 64 bytes, not bits. So Apple is using 128 bytes.
Also important to consider that the RTX 4090 has a relatively tiny 384-bit memory bus. Smaller than the M1 Max's 512-bit bus. But the RTX 4090 has 1 TB/s bandwidth and significantly more compute power available to make use of that bandwidth.
The M4 max is definitely not a 4090 killer, does not match it in any way. It can however work on larger models than the 4090 and have a battery that can last all day.
My memory is a bit fuzzy, but I believe the m3 max did decent on some games compared to the laptop Nvidia 4070 (which is not the same as the desktop 4070). But highly depended on if the game was x86-64 (requiring emulation) and if it was DX11 or apple native. I believe apple claims improvements in metal (the Apple's GPU lib) and that the m4 GPUs have better FP for ray tracing, but no significant changes in rasterized performance.
I look forward to the 3rd party benchmarks for LLM and gaming on the m4 max.
> The new idea is having 512 bit wide memory instead of PC limitation of 128 bit wide.
It's not really a new idea, just unusual in computers. The custom SOCs that AMD makes for Playstation and Xbox have wide (up to 384-bit) unified memory buses, very similar to what Apple is doing, with the main distinction being Apples use of low-power LPDDR instead of the faster but power hungrier GDDR used in the consoles.
Yeah, a lot of it is just market forces. I guess going to four channels is costly for the desktop PC space and that's why that didn't happen, and laptops just kind of followed suite. But now that Apple is putting pressure on the market, perhaps we'll finally see quad channel becoming the norm in desktop PCs? Would be nice...
"Unified memory" doesn't really imply anything about the memory being located on-package, just that it's a shared pool that the CPU, GPU, etc. all have fast access to.
Also, DRAM is never on-die. On-package, yes, for Apple's SoCs and various other products throughout the industry, but DRAM manufacturing happens in entirely different fabs than those used for logic chips.
It's mostly an IBM thing. In the consumer space, it's been in game consoles with IBM-fabbed chips. Intel's use of eDRAM was on a separate die (there was a lot that was odd about those parts).
For comparison, a Threadripper Pro 5000 workstation with 8x DDR4 3200 has 204.8GB/s of memory bandwidth.
The Threadripper Pro 7000 with DDR5-5200 can achieve 325GB/s.
And no, manaskarekar, the M4 Max does 546 GB/s not GBps (which would be 8x less!).
Thanks for the numbers. Someone here on hackernews got me convinced that a Threadripper would be a better investment for inference than a MacBook Pro with a M3 Max.
On PC desktops I always recommend getting a mid-range tower server precisely for that reason. My oldest one is about 8 years old and only now it's showing signs of age (as in not being faster than the average laptop).
Yes, you could buy a brand new (announced weeks ago) AMD Turin. 12 channels of DDR5-6000, $11,048 and 320 watts (for the CPU) and get 576GB/sec peak.
Or you could buy a M3 max laptop for $4k, get 10+ hour battery life, have it fit in a thin/light laptop, and still get 546GB/sec. However those are peak numbers. Apple uses longer cache lines (double), large page sizes (quadruple), and a looser memory model. Generally I'd expect nearly every memory bandwidth measure to win on Apple over AMD's turin.
AnandTech did bandwidth benchmarks for the M1 Max and was only able to utilize about half of it from the CPU, and the GPU used even less in 3D workloads because it wasn't bandwidth limited. It's not all about bandwidth. https://www.anandtech.com/show/17024/apple-m1-max-performanc...
Indeed. RIP Anandtech. I've seen bandwidth tests since then that showed similar for newer generations, but not the m4. Not sure if the common LLM tools on mac can use CPU (vector instructions), AMX, and Neural engine in parallel to make use of the full bandwidth.
FWIW I ran a quick test of gemma.cpp on M3 Pro with 8 threads. Similar PaliGemma inference speed to an older AMD (Rome or Milan) with 8 threads. But the AMD has more cores than that, and more headroom :)
I doubt you'll get 10+ hours on battery if you utilize it at max. I don't even know if it can really sustain the maximum load for more than a couple of minutes because of thermal or some other limits.
The 14" MBP has a 72 watt-hour battery and the 16" has a 100 watt-hour battery.
At full tilt an M3 Max will consume 50 to 75 watts, meaning you get 1 to 2 hours of runtime at best, if you use the thing full tilt.
That's the thing I find funny about the Apple Silicon MBP craze, sure they are efficient but if you use the thing as a workstation, battery life is still not good enough to really use unplugged.
Most claiming insane battery life are using the thing effectively as an information appliance or a media machine. At this game the PC laptops might not be as efficient but the runtime is not THAT different provided the same battery capacity.
You lose out on things like expandability (more storage, more PCIe lanes) and repairability though. You are also (on M4 for probably a few years) compelled to use macOS, for better or worse.
There are, in my experience, professionals who want to use the best tools someone else builds for them, and professionals who want to keep iterating on their tools to make them the best they can be. It's the difference between, say, a violin and a Eurorack. Neither's better or worse, they're just different kinds of tools.
I was sorely tempted by the Mac studio, but ended up with a 96GB ram Ryzen 7900 (12 core) + Radeon 7800 XT (16GB vram). It was a fraction of the price and easy to add storage. The Mac M2 studio was tempting, but wasn't refreshed for the M3 generation. It really bothered me that the storage was A) expensive, B) proprietary, C) tightly controlled, and D) you can't boot without internal storage.
Even moving storage between Apple studios can be iffy. Would I be able to replace the storage if it died in 5 years? Or expand it?
As tempting as the size, efficiency, and bandwidth were I just couldn't justify top $ without knowing how long it would be useful. Sad they just didn't add two NVMe ports or make some kind of raw storage (NVMe flash, but without the smarts).
> Even moving storage between Apple studios can be iffy.
This was really driven home to me by my recent purchase of an Optane 905p, a drive that is both very fast and has an MTBF measured in the hundreds of years. Short of a power surge or (in California) an earthquake, it's not going to die in my lifetime -- why should I not keep using it for a long time?
Many kinds of professionals are completely fine with having their Optanes and what not only be plugged in externally, though, even though it may mean their boot drive will likely die at some point. That's completely okay I think.
Yeah memory bandwidth is one of the really unfortunate things about the consumer stuff. Even the 9950x/7950x, which are comfortably workstation-level in terms of compute, are bound by their 2 channel limits. The other day I was pricing out a basic Threadripper setup with a 7960x (not just for this reason but also for more PCIe lanes), and it would cost around $3000 -- somewhat out of my budget.
This is one of the reasons the "3D vcache" stuff with the giant L3 cache is so effective.
I don't think so? That PDF I linked is from 2015, way before Apple put focus on it through their M-series chips... And the Wikipedia article on "Glossary of computer graphics" has had an entry for unified memory since 2016: https://en.wikipedia.org/w/index.php?title=Glossary_of_compu...
For Apple to have come up with using the term "unified memory" to describe this kind of architecture, they would've needed to come up with it at least before 2016, meaning A9 chip or earlier. I have paid some attention to Apple's SoC launches through the years and can't recall them touting it as a feature in marketing materials before the M1. Do you have something which shows them using the term before 2016?
To be clear, it wouldn't surprise me if it has been used by others before Intel did in 2015 as well, but it's a starting point: if Apple hasn't used the term before then, we know for sure that they didn't come up with it, while if Apple did use it to describe A9 or earlier, we'll have to go digging for older documents to determine whether Apple came up with it
There are actual differences but they're mostly up to the drivers. "Shared" memory typically means it's the same DRAM but part of it is carved out and can only be used by the GPU. "Unified" means the GPU/CPU can freely allocate individual pages as needed.
We run our LLM workloads on a M2 Ultra because of this. 2x the VRAM; one-time cost at $5350 was the same as, at the time, 1 month of 80GB VRAM GPU in GCP. Works well for us.
At some point there should be an upgrade to the M2 Ultra. It might be an M4 Ultra, it might be this year or next year. It might even be after the M5 comes out. Or it could be skipped in favour of the M5 Ultra. If anyone here knows they are definitely under NDA.
About 10-20% of my companies gpu usage is inference dev. Yes horribly not efficient usage of resources. We could upgrade the 100ish devs who do this dev work to M4 mbp and free up gpu resources
High availability story for AI workloads will be a problem for another decade. From what I can see the current pressing problem is to get stuff working quickly and iterate quickly.
Can you elaborate, are those workflows in queue or can they serve multiple users in parallel ?
I think it’s super interesting to know real life workflows and performance of different LLMs and hardware, in case you can direct me to other resources.
Thanks !
They aren't going to be using fp32 for inferencing, so those FP numbers are meaningless.
Memory and memory bandwidth matters most for inferencing. 819.2 GB/s for M2 Ultra is less than half that of A100, but having 192GB of RAM instead of 80gb means they can run inference on models that would require THREE of those A100s and the only real cost is that it takes longer for the AI to respond.
3 A100 at $5300/mo each for the past 2 years is over $380,000. Considering it worked for them, I'd consider it a massive success.
From another perspective though, they could have bought 72 of those Ultra machines for that much money and had most devs on their own private instance.
The simple fact is that Nvidia GPUs are massively overpriced. Nvidia should worry a LOT that Apple's private AI cloud is going to eat their lunch.
Having 128GB is really nice if you want to regularly run different full OSes as VMs simultaneously (and if those OSes might in turn have memory-intensive workloads running on them).
This is definitely tempting me to upgrade my M1 macbook pro. I think I have 400GB/s of memory bandwidth. I am wondering what the specific number "over half a terabyte" means.
At least in the recent past, a hindrance was that MacOS limited how much of that unified memory could be assigned as VRAM. Those who wanted to exceed the limits had to tinker with kernel settings.
I wonder if that has changed or is about to change as Apple pivots their devices to better serve AI workflows as well.
you'd probably save money just paying for a VPS. And you wouldn't cook your personal laptop as fast. Not that people nowadays keep their electronics for long enough for that to matter :/
I'm curious about getting one of these to run LLM models locally, but I don't understand the cost benefit very well. Even 128GB can't run, like, a state of the art Claude 3.5 or GPT 4o model right? Conversely, even 16GB can (I think?) run a smaller, quantized Llama model. What's the sweet spot for running a capable model locally (and likely future local-scale models)?
You'll be able to run 72B models w/ large context, lightly quantized with decent'ish performance, like 20-25 tok/sec. The best of the bunch are maybe 90% of a Claude 3.5.
If you need to do some work offline, or for some reason the place you work blocks access to cloud providers, it's not a bad way to go, really. Note that if you're on battery, heavy LLM use can kill your battery in an hour.
I am always wondering if one shouldn't be doing the resource intensive LLM stuff in the cloud. I don't know enough to know the advantages of doing it locally.
Have they published this ahead of other pages or is it just me?
The linked Apple Store page says "MacBook Pro blasts forward with the M3, M3 Pro, and M3 Max chips" so it seems like the old version of the page still?
Looking at how long the 8gb lasted it's a pretty sure bet that now you won't need to upgrade for a good few years.
I mean, I have a MacBook air with 16gb of ram and it's honestly working pretty well to this day. I don't do "much" on it though but not many people do.
I'd say the one incentive a MacBook Pro has over the air is the better a screens and better speakers. Not sure if it's worth the money.
My hypothesis is Apple is mostly right about their base model offerings.
> I mean, I have a MacBook air with 16gb of ram and it's honestly working pretty well to this day. I don't do "much" on it though but not many people do.
If an HN user can get along with 16gb on their MacBook Air for the last X years, most users were able to get by with 8gb.
It's just a tactic to get a higher average price while being able to advertise a lower price. What makes it infuriating is memory is dirt cheap. That extra 8GB probably costs them $10 at most, but would add to utility and longevity of their hardware quite a bit.
They are supposed to be "green" but they encourage obsolescence.
They align need with more CPU and margin. Apple wants as few SKUs as possible and as much margin as possible.
8GB is fine for most use cases. Part of my gig is managing a huge global enterprise with six figures of devices. Metrics demonstrate that the lower quartile is ok with 8GB, even now. Those devices are being retired as part of the normal lifecycle with 16GB, which is better.
Laptops are 2-6 year devices. Higher end devices always get replaced sooner - you buy a high end device because the productivity is worth spending $. Low end tend to live longer.
People looking for low prices buy PC, they don't even consider Mac. Then they can have a computer with all the "higher numbers", which is more important than getting stuff done.
I bought a framework back in 2020 or so and really wish I just waited a little longer and spent a few hundred bucks more on the M1.
It's fine, but the issue is linux sleep/hibernate - battery drain. To use the laptop after a few days, I have to plug it in and wait for it to charge a little bit because the battery dies. I have to shut it down (not just close the screen) before flying or my backpack becomes a heater and the laptop dies. To use a macbook that's been closed for months I just open it and it works. I'll pay double for that experience. If I want a computer that needs to be plugged in to work I have a desktop for that already. The battery life is not good either.
Maybe it's better now if I take the time to research what to upgrade, but I don't have the time to tinker with hardware/linux config like I did a few years ago.
I don't mind spending a thousand bucks every 7 years to upgrade my laptop. I've had this macbook air since 2020 and besides the speakers don't being the best... I have no complaints.
I don't really see a world where this machine doesn't last me a few more years. If there's anything i'd service would be the battery, but eh. It lasts more than a few hours and I don't go out much.
Even the "few thousand" for a bit of extra RAM and SSD are highway robbery.
Anyways to each their own, I also had things break and repairability isn't really a thing with Apple hardware (while it easily could be if they wanted - even if difficult and by certified technicians instead of myself)
To be fair, the link in this story is to a press release. Arguably there are probably many things in it that can be considered "misleading" in certain contexts.
What's the deal with running Linux on these anyway? Could one conceivably set up an M4 mini as headless server? I presume Metal would be impossible to get working if MacOS uses proprietary drivers for it...
This is the first compelling Mac to me. I've used Macs for a few clients and muscle memory is very deeply ingrained for linux desktops. But with local LLMs finally on the verge of usability along with sufficient memory... I might need to make the jump!
Wish I could spin up a Linux OS on the hardware though. Not a bright spot for me.
It won't have all the niceties / hardware support of MacOS, but it seamlessly coexists with MacOS, can handle the GPU/CPU/RAM with no issues, and can provide you a good GNU/Linux environment.
IIRC one of the major factors holding back M3 support was the lack of a M3 mini for use in their CI environment. Now that there's an M4 mini hopefully there aren't any obstacles to them adding M4 support
How? What cloud providers offer it? MacStadium and AWS don't.
I guess you could have a physical MBP in your house and connect it to some bring-your-own-infrastructure CI setup, but most people wouldn't want to do that.
How do you imagine that a cloud computing platform designed around running Macs with macOS would work for testing an entirely different OS running on bare metal on hardware that doesn't have a BMC, and usefully catching and logging frequent kernel panics and failed boots?
It's a pretty hard problem to partially automate for setups with an engineer in the room. It doesn't sound at all feasible for an unattended data center setup that's designed to host Xcode for compiling apps under macOS.
Getting M4 Mac Mini CI might be a while considering the amount they've changed the package. Too tall to go into a 1U, smaller so the rack frames need to be redesigned. Power button now on the bottom so switch actuation need finagling.
In general for local LLMs, the more memory the better. You will be able to fit larger models in RAM. The faster CPU will give you more tokens/second, but if you are just chatting with a human in the loop, most recent M series macs will be able to generate tokens faster than you can read them.
That also very much depends on model size. For 70B+ models, while the tok/s are still fast enough for realtime chat, it's not going to be generating faster than you can read it, even on Ultra with its insane memory bandwidth.
NextSTEP which macOS is ultimately based on is indeed older than Linux (first release was 1989). But why does that matter? The commenter presumably said "Linux" for a reason, i.e. they want to use Linux specifically, not any UNIX-like OS.
Sure. But not everybody. That’s how I ended up on a Mac. I needed to develop for Linux servers and that just sucked on my windows laptop (I hear it’s better now?). So after dual booting fedora on my laptop for several months I got a MacBook and I’ve never looked back.
BSD is fun (not counting MacOS in the set there), but no, my Unix experiences have been universally legacy hardware oversubscribed and undermaintained. Not my favorite place to spend any time.
I miss Linux, it respected me in ways that MacOS doesn't. But maintaining a sane dev environment on linux when my co-workers on MacOS are committing bash scripts that call brew... I am glad that I gave up that fight. And yeah, the hardware sure is nice.
IIRC brew supports linux, but it isn't a package manager I pay attention to outside of some very basic needs. Way too much supply chain security domain to cover for it!
It does, but I prefer to keep project dependencies bound to that project rather than installing them at wider scope. So I guess it's not that I can't use Linux for work, but that I can't use Linux for work and have it my way. And if I can't have it my way anyway, then I guess Apple's way will suffice.
Yep. That's roughly 20% per generation improvement which ain't half-bad these days, but the really huge cliff was going from Intel to the M1 generation.
M1 series machines are going to be fine for years to come.
It feels like M1 was the revolution, subsequent ones evolution - smaller fabrication process for improved energy efficiency, more cores for more power, higher memory (storage?) bandwidth, more displays (that was a major and valid criticism for the M1 even though in practice >1 external screens is a relatively rare use case for <5% of users).
Actually wasn't M1 itself an evolution / upscale of their A series CPUS that by now they've been working on since... before 2010, the iPhone 4 was the first one with their own CPU, although the design was from Samsung + Intrinsity, it was only the A6 that they claimed was custom designed by Apple.
In early 2020, I had an aging 2011 Air that was still struggling after a battery replacement. Even though I "knew" the Apple Silicon chips would be better, I figured a 2020 Intel Air would last me a long time anyway, since my computing needs from that device are light, and who knows how many years the Apple Silicon transition will take take anyway?
Bought a reasonably well-specced Intel Air for $1700ish. The M1s came out a few months later. I briefly thought about the implication of taking a hit on my "investment", figured I might as well cry once rather than suffer endlessly. Sold my $1700 Intel Air for $1200ish on craigslist (if I recall correctly), picked up an M1 Air for about that same $1200 pricepoint, and I'm typing this on that machine now.
That money was lost as soon as I made the wrong decision, I'm glad I just recognized the loss up front rather than stewing about it.
Exact same boat here. A friend and I both bought the 2020 Intel MBA thinking that the M1 version was at least a year out. It dropped a few months later. I immediately resold my Intel MBA seeing the writing on the wall and bought a launch M1 (which I still use to this day). Ended up losing $200 on that mis-step, but no way the Intel version would still get me through the day.
That said...scummy move by Apple. They tend to be a little more thoughtful in their refresh schedule, so I was caught off guard.
When I saw the M1s come out, I thought that dev tooling would take a while to work for M1, which was correct. It probably took a year for most everything to be compiled for arm64. However I had too little faith in Rosetta and just the speed upgrade M1 really brought. So what I mean to say is, I still have that deadweight MBA that I only use for web browsing :)
Oh yes, my wife bought a new Intel MBA in summer 2020... I told her at the time Apple planned its own chip, but it couldn't be much better than the Intel one and surely Apple will increase prices too... I was so wrong.
I have the OG 13" MBP M1, and it's been great; I only have two real reasons I'm considering jumping to the 14" MBP M4 Pro finally:
- More RAM, primarily for local LLM usage through Ollama (a bit more overhead for bigger models would be nice)
- A bit niche, but I often run multiple external displays. DisplayLink works fine for this, but I also use live captions heavily and Apple's live captions don't work when any form of screen sharing/recording is enabled... which is how Displaylink works. :(
Not quite sold yet, but definitely thinking about it.
I've never kept any laptop as long as I've kept the M1. I was more or less upgrading yearly in the past because the speed increases (both in the G4 and then Intel generations) were so significant. This M1 has exceeded my expectations in every category, it's faster quieter and cooler than any laptop i've ever owned.
I've had this laptop since release in 2020 and I have nearly 0 complaints with it.
I wouldn't upgrade except the increase in memory is great, I don't want to have to shut down apps to be able to load some huge LLMs, and, I ding'ed the top case a few months ago and now there's a shadow on the screen in that spot in some lighting conditions which is very annoying.
I hope (and expect) the M4 to last just as long as my M1 did.
My 2015 15" MBP is also still kickin, is/was an absolutely fabulous unit. Was my work machine for 3-4 years, and now another almost-6 years as my personal laptop. My personal use case is obviously not very demanding but it's only now starting to really show its age.
I also have a M1 from work that is absolutely wonderful, but I think it's time for me to upgrade the 2015 with one of these new M4s.
Honestly, my Thinkpad from 2015 was still used in my family until recently. The battery was pretty bad, same as on my 2015 MBP, but other than that, I put Fedora on it, and it was still really fast.
Longevity is not only a thing of MBPs. OTOH, IIRC, some 2017-2019 MBPs (before the Mx switch) were terrible for longevity, given their problematic keyboard.
That is amazing. Mine lasted for a super long time as well, and like you, I upgraded everything to its max. I think it was the last model with a 17 inch screen.
Sold mine last year for $100 to some dude who claimed to have some software that only runs on that specific laptop. I didn't question it.
I was also a mid-2012 MBP user. I eventually got the M2 MBA because I was investing in my eyesight (modern displays are significantly better). I was never impressed with the touchbar-era macs, they didn't appeal to me and their keyboards were terrible.
I think this M-series macbook airs are a worthy successor to the 2012 MBP. I fully intend to use this laptop for at least the same amount of time, ideally more. The lack of replaceable battery will probably be the eventual killer, which is a shame.
Ah my 2013 mbp died in 2019. It was the gpu. No way to repair it for cheap enough so I had to replace it with a 2019 mbp which was the computer I kept the shortest (I hated the keyboard).
I used that same model for 5 years until I finally upgraded in 2017 and totally regretted it, the upgrade was not worth it at all, I would have been just as happy with the 2012. I quickly replaced it again with the "Mea Culpa" 2019 where they added back in ports, etc, would have been just about worth the upgrade over the 2012, 7 years later, but again, not by a big margin.
The 2012 MBP 15" Retina was probably the only machine I bought where the performance actually got better over the years, as the OS got more optimized for it (the early OS revisions had very slow graphics drivers dealing with the retina display)
The M1 Pro on the other hand, that was a true upgrade. Just a completely different experience to any Apple Intel laptop.
I still have my 2015, and it lived just long enough to keep me going until the death of the touch bar and horrible keyboard, which went away when I immediately bought the M1 Pro on release day.
You'll be glad you did. I loved my 2015 MBP. I even drove 3 hours to the nearest Best Buy to snag one. That display was glorious. A fantastic machine. I eventually gave it to my sister, who continued using it until a few years ago. The battery was gone, but it still worked great.
When you upgrade, prepare to be astonished.
The performance improvement is difficult to convey. It's akin to traveling by horse and buggy. And then hopping into a modern jetliner, flying first class.
It's not just speed. Display quality, build quality, sound quality, keyboard quality, trackpad, ports, etc., have all improved considerably.
> The battery was gone, but it still worked great.
A family 2018 Macbook Air got a second life with a battery replacement. Cheap kit from Amazon, screwdrivers included, extremely easy to do. Still in use, no problems.
The performance jump between a top-of-the-line intel MBP (I don't remember the year, probably 2019) and the m1 max I got to replace it.. was rather like the perf jump between spinning disks and SSDs.
When I migrated all my laptops to SSDs (lenovos at the time, so it was drop-dead simple), I thought to myself, "this is a once-in-a-generation feeling". I didn't think I would ever be impressed by a laptop's speed ever again. It was nice to be wrong.
I've just replaced a 2012 i5 mbp, and used it for Dev work and presentations into 2018.
It has gotten significantly slower the last 2 years, but the more obvious issue is the sound, inability to virtual background, and now lack of software updates.
But if you had told me I'd need to replace it in 2022 I wouldn't believe you
I was considering upgrading to an M3 up until about a month ago when Apple replaced my battery, keyboard, top case, and trackpad completely for free. An upgrade would be nice as it no longer supports the latest MacOS, but at this point, I may just load Ubuntu on the thing and keep using it for another few years. What a machine.
How do you justify this kind of recurring purchases, even with selling your old device? I don't get the behaviour or the driving decision factor past the obvious "I need the latest shiny toy" (I can't find the exact words to describe it, so apologies for the reductive description).
I have either assembled my own desktop computers or purchased ex corporate Lenovo over the years with a mix of Windows (for gaming obviously) and Linux and only recently (4 years ago) been given a MBP by work as they (IT) cannot manage Linux machines like they do with MacOS and Windows.
I have moved from an intel i5 MBP to a M3 Pro (?) and it makes me want to throw away my dependable ThinkPad/Fedora machine I still uses for personal projects.
There are 2 things I was always spending money on, if I felt is not the almost best achievable: my bed and my laptop. Even the phone can be 4 years old iPhone, but the laptop must be best and fast. My sleep is also pretty important. Everything else is just "eco".
In my country you can buy a device and write off in 2 years, VAT reimbursed, then scrap it from the books and you sell it to people without tax payed to people who otherwise would pay a pretty hefty VAT. This decreases your loss of value to like half.
It's tax avoidance, not evasion. If it's fully legal then I don't know why wouldn't you recommend it. If you are against it, you can easily pay more in taxes than required yourself.
Apple has a pretty good trade-in program. If you have an Apple card, it's even better (e.g. the trade-in value is deducted immediately, zero interest, etc.).
Could you get more money by selling it? Sure. But it's hard to be the convenience. They ship you a box. You seal up the old device and drop it off at UPS.
I also build my desktop computers with a mix of Windows and Linux. But those are upgraded over the years, not regularly.
I spend easily 100 hours a week using it not-as-balanced-as-it-should-be between the two.
I don't buy them because I need something new, I buy them because in the G4/Intel era, the iterations were massive and even a 20 or 30% increase in speed (which could be memory, CPU, disk -- they all make things faster) results in me being more productive. It's worth it for me to upgrade immediately when apple releases something new, as long as I have issues with my current device and the upgrade is enough of a delta.
M1 -> M2 wasn't much of a delta and my M1 was fine.
M1 -> M3 was a decent delta, but, my M1 was still fine.
M1 -> M4 is a huge delta (almost double) and my screen is dented to where it's annoying to sit outside and use the laptop (bright sun makes the defect worse), so, I'm upgrading. If I hadn't dented the screen the choice would be /a lot/ harder.
I love ThinkPads too. Really can take a beating and keep on going. The post-IBM era ones are even better in some regards too. I keep one around running Debian for Linux-emergencies.
Yeah, I always have AppleCare. I view it as part of the cost of a mac (or iPhone).
And yeah, this incident reminded me of why it's important to back up as close to daily as you can, or even more often during periods when you're doing important work and want to be sure you have the intermediate steps.
Interesting. I have found occasion to use it for pretty much every Mac I've owned since the 1980s! I'm not sure how much money it's saved compared to just paying for repairs when needed, but I suspect it may come out to:
1) a slight overall savings, though I'm not sure about that.
2) a lack of stress when something breaks. Even if there isn't an overall savings, for me it's been worth it because of that.
Certainly, my recent Mac repair would have cost $1500 and I only paid $300, and I think I've had the machine for about 3 years, so there's a savings there but considerably less recent stress. That's similar to the experience I've had all along, although this recent expense would have probably been my most-expensive repair ever.
I've spilled liquid on my MacBook's once every 10 years on average. Last in 2014, then again last month. Accidents happen.
As I've noted in a sibling comment, I'll probably stop purchasing mobile Macs until the repair story on Macbooks is improved -- the risk for accidents and repairs is simply much higher on portable machines. That's only going to happen through third-party repair (which I think would simultaneously lead Apple to lower their first-party repair costs, too).
"SSD is soldered on" is a bit of glossing over of the issue with the M-series Macs.
Apple is putting raw NAND chips on the board (and yes soldering them) and the controller for the SSD is part of the M-series chip. Yes, apple could use NVMe here if you ignore the physical constraints and ignore fact that it wouldn't be quite as fast and ignore the fact that it would increase their BOM cost.
I'm not saying Apple is definitively correct here, but, it's good to have choice and Apple is the only company with this kind of deeply integrated design. If you want a fully modular laptop, go buy a framework (they are great too!) and if you want something semi-modular, go buy a ThinkPad (also great!).
I need macOS for work. Now that the writing is on the wall for Hackintosh (which I used to do regularly while purchasing a Mac every few years, most recently in 2023 and 2018, because I love that choice), I don't have a choice. I used to spend 10-20 hours per third party machine for that choice.
I don't truly mind that they solder on the SSD, embed the controller into the processor -- you're right that it's great we have choice here. I mind the exuberant repair cost _on top of_ Apple's war on third party repair. Apple is the one preventing me to have choice here, I have to do the repair through them, or wait until schematics are smuggled out of China and used/broken logic boards are available so that the repair costs what it should: $300 to replace 2 chips on my logic board (still mostly labor, but totally a fair price).
I love Apple for their privacy focus and will continue to support them because I need to do Mac and iOS development, but I will likely stop buying mobile workstations from them for this reason, the risk of repair is simply much higher and not worth this situation.
Mine fell off from the roof of a moving car at highway speeds and subsequently spent 30 mins being run over by cars until it was picked back up. Otherwise no complaints.
I had an 2019 i9 for a work laptop. It was absolutely awful, especially with the corporate anti-virus / spyware on it that brought it to a crawl. Fans would run constantly. Any sort of Node JS build would make it sound like a jet engine.
That was the worst laptop I've ever had. Not only was it turning the jet engines on when you tried to do something more demanding that moving mouse around, it throttled thermally so much that you literally could not move that mouse around.
>I've never kept any laptop as long as I've kept the M1
What different lives we live. This first M1 was in November 2020. Not even four years old. I’ve never had a [personal] computer for _less_ time than that. (Work, yes, due to changing jobs or company-dictated changes/upgrades)
That’s a feature, not a bug, for some. When I upgraded to an M series chip MacBook, I had to turn up the heat because I no longer had my mini space heater.
Exactly my thoughts. I don't understand whether I'm really spoiled, or is the crowd here weird about upgrading for some reason - if you have a laptop from 4-5 years ago, the new one would be 2-5x faster in vast majority of things - even if not critical for your workflow, it would feel SO MUCH nicer - so if it's something you use for 100h / week, shouldn't you try to make it as enjoyable as reasonably possible?
Other example - I'm by no means rich, but I have a $300 mechanical keyboard - it doesn't make me type faster and it doesn't have additional functionality to a regular $30 Logitech one - but typing on it feels so nice and I spend so much of my life doing it, that to me it's completely justified and needed to have this one then.
> I've never kept any laptop as long as I've kept the M1.
I still have a running Thinkpad R60 from 2007, a running Thinkpad T510 from 2012, and a modified running Thinkpad X61 (which I re-built as an X62 using the kit from 51nb in 2017 with a i7-5600U processor, 32 GB of RAM and a new display) in regular use. The latter required new batteries every 2 years, but was my main machine until 2 weeks ago when I replaced it with a ThinkCentre. During their time as my main machine, each of these laptops was actively used around 100 hours per week, and was often running for weeks without shutdown or reboot. The only thing that every broke was the display of the R60 which started to show several green vertical bars after 6 years, but replacement was easy.
Bought the same machine from the same store last December, been using it ever since with a big grin on my face. I'll probably consider upgrading in 2028 or later.
I have the same one, but everyone I know with an M series Mac says the same thing. These are the first machines in a long time built to not only last a decade but be used for it.
I still use my MacBook Air M1 and given my current workloads (a bit of web development, general home office use and occasional video editing and encoding) I doubt I’ll need to replace it in the coming 5 years. That’ll be an almost 10 year lifespan.
M1 Pro compared to Intel was so big step ahead that I suppose we all are still surprised and excited. Quiet, long battery life and better performance. By a lot!
I wonder if M4 really feels that much faster and better - having M1 Pro I'm not going to change quickly, but maybe Mac Mini will land some day.
Honestly it was a game changer. Before I'd never leave the house without a charger, nowadays I rarely bring it with me on office days, even with JS / front-end workloads.
(of course, everyone else has a macbook too, there's always someone that can lend me a charger. Bonus points that the newer macbooks support both magsafe and USB-C charging. Added bonus points that they brought back magsafe and HDMI ports)
It's the other way around, isn't it? MagSafe was removed in the 2016-2019 model years (not sure why; maybe to shave off another bit of thickness?), and then brought back in 2020 to MacBook Pro and 2022 to MacBook Air.
Personally, I practically never use MagSafe, because the convenience of USB C charging cables all over the house outweighs the advantages of MagSafe for me.
I have the M1 MacBook Pro with MagSafe and I still charge it via USB C simply because I can't be bothered to carry around another cable when all of my other peripherals are USB C.
That's interesting, I don't use USB A at all whatsoever anymore, and frankly I don't use the SD card reader or HDMI port either. Maybe I should just get a MacBook Air 15" but I do like the 120 hz screen of the Pro models, that's the main reason I'm holding out.
With whisky i feel like id never need anything else. That said, the benchmark jump in the m4 has me thinking i should save up and grab a refurb in a year or two
I switched from a 2014 MacBook pro to a 2020 M1 MacBook Air, yeah the CPU is much faster, but the build quality and software is a huge step backwards. The trackpad is feels fake, not nearly as responsive, keyboard also feel not as solid. But now I'm already used to it.
They also feel very bulky/innelegant while still being fragile for the most part and not really hitting workstation level territory.
I don't understand how people are enamored with those things, sure it's better in some way than what it was before but it's also very compromised for the price.
It's a very robust and capable small laptop. I'm typing this to a M1 Macbook Air.
The only thing to keep in mind, is that the M1 was the first CPU in the transition from Intel CPUs (+ AMD GPUs) to Apple Silicon. The M1 was still missing a bunch of things from earlier CPUs, which Apple over time added via the M1 Pro and other CPUs. Especially the graphics part was sufficient for a small laptop, but not for much beyond. Better GPUs and media engines were developed later. Today, the M3 in a Macbook Air or the M4 in the Macbook Pro have all of that.
For me the biggest surprise was how well the M1 Macbook Air actually worked. Apple did an outstanding job in the software & hardware transition.
Yep. Bought an M1 Max in 2021 and it still feels fast, battery lasts forever. I’m sure the M4 would be even quicker (Lightroom, for example) but there’s little reason to consider an upgrade any time soon.
I too bought a 2020 MBA M1, it was great initially, but now seems like its getting throttled, same goes with my iPhoneX, I used to love Apple, but its just pathetic that they throttle older devices just to get users to upgrade.
Disingenuous to mention the x86 based MacBooks as a basis for comparison in their benchmarks; they are trying to conflate current-gen Intel with what they shipped more than 4 years ago.
Are they going to claim that 16GB RAM is equivalent to 32GB on Intel laptops? (/sarc)
Lot's of people don't upgrade on the cadence that users on this forum do. Someone was mentioning yesterday that they are trying to sell their Intel Mac {edit: on this forum] and asking advice on getting the best price. Someone else replied that they still had a 2017 model. I spoke to someone at my job (I'm IT) who told me they'd just ordered a new iMac to replace one that is 11 years old. There's no smoke and mirrors in letting such users know what they're in for.
Given that they also compare it to an M1 in the same aside, I'd say you're wrong.
> Up to 23.8x faster basecalling for DNA sequencing in Oxford Nanopore MinKNOW when compared to the 16-inch MacBook Pro with Core i9, and up to 1.8x faster when compared to the 16-inch MacBook Pro with M1 Pro.
Yup, I'm a developer who still primarily works on a 2018 Intel Mac. Apple's messaging felt very targeted towards me. Looking forward to getting the M4 Max as soon as possible!
Ben Bejarin said that around 50% of the installed base is still using Macs with Intel chips. You’ll keep hearing that comparison until that number goes down.
If you're willing to play, here are plenty of lenders who will finance this purchase.
If it affects your earning power to that extent, you should probably pony up and save in the long run, probably just a few years until you see returns.
Caste system usually can't be bypassed by paying a monthly subscription fee.
I will note that making it a subscription will tend to increase the overall costs, not decrease. In an environment with ready access to credit, I think offering on a subscription basis is worse for consumers?
To all these comments, I'm not talking practically for myself, I'm talking about what a revolutionary "think different" company would do for its users and projection into the world. Starting with taking away the friction of changing models and the impact of all these product lines, which sometimes instantly become less valuable (2023's "8GB is more than enough"), would be a good start, and Apple if anyone could amortize this on behalf of their userbase.
Another observation; I've travelled the world and rarely see people who could use robust, secure products the most (vulnerable people) using Apple products. They're all packing second-tier Samsung or LG Androids and old Windows notebooks (there are decent Samsung, LG, Android, Windows products, but that's not what they have access to).
If you’re willing to buy from a retailer you can usually get two or three year financing terms. sell it at the end of the payment term for half (or more) of what you paid in total and get a new one on a similar plan if you want.
don’t think it’s wise though, i bought a base m1 pro mbp when it launched and don’t feel a need to upgrade at all yet. i’m holding off for a few more years to grab whenever the next major increase in local llm capability and battery life comes.
> All MacBook Pro models feature an HDMI port that supports up to 8K resolution, a SDXC card slot, a MagSafe 3 port for charging, and a headphone jack, along with support for Wi-Fi 6E and Bluetooth 5.3.
No Wifi 7. So you get access to the 6 GHz band, but not some of the other features (preamble punching, OFDMA):
The iPhone 16s do have Wifi 7. Curious to know why they skipped it (and I wonder if the chipsets perhaps do support it, but it's a firmware/software-not-yet-ready thing).
Yeah, this threw me as well. When the iMac didn’t support WiFi 7, I got a bit worried. I have an M2, so not going to get this, but the spouse needs a new Air and I figure that everything would have WiFi 7 by then, and now I don’t think so.
Faster is always nice, makes sense. But do you really need WiFi 7 features/speed? I don't know when I would notice a difference (on a laptop) between 600 or 1500 Mbit/s (just as an example). Can't download much anyhow as the storage will get full in minutes.
wifi 6 can’t do shit given that there’s crazy overlap if you go for 80MHz channels. Most situations tap out around 700Mbps over usually a 40MHz channel.
Wifi 6e/7 6GHz band with a bunch of non overlapping 160MHz channels is where the juice is at. But even then a lot of devices are limited to smaller channel widths.
> It looks like few people only are using Wifi 7 for now.
Machines can last and be used for years, and it would be a presumably very simple way to 'future proof' things.
And though the IEEE spec hasn't officially been ratified as I type this, it is set to be by the end of 2024. Network vendors are also shipping APs with the functionality, so in coming years we'll see a larger and larger infrastructure footprint going forward.
Laptops/desktops (with 16GB+ of memory) could make use of the faster speed/more bandwidth aspects of WiFi7 better than smartphones (with 8GB of memory).
Hm why? Is 6E really so much worse than 7 in practice that 7 can replace wired for you but 6E can't? That's honestly really weird to me. What's the practical difference in latency, bandwidth or reliability you've experienced between 6E and 7?
I don’t have any 6E device so I cannot really tell for sure but from what I read, 6E gets you to a bit over 1Gbit in real world scenario. 7 should be able to replace my 2.5Gbe dongle or at least get much closer to it. I already have routers WiFi 7 Eeros on a 2.5Gbe wired backbone.
I guess it makes sense if what you do is extremely throughput-focused... I always saw consistency/reliability and latency as the benefits of wired compared to wireless, the actual average throughput has felt fast enough for a while on WiFi but I guess other people may have different needs
I was quite surprised by this discrepancy as well (my new iPhone has 7, but the new MBP does not).
I had just assumed that for sure this would be the year I upgrade my M1 Max MBP to an M4 Max. I will not be doing so knowing that it lacks WiFi 7; as one of the child comments notes, I count on getting a solid 3 years out of my machine, so future-proofing carries some value (and I already have WiFi7 access points), and I download terabytes of data in some weeks for the work I do, and not having to Ethernet in at a fixed desk to do so efficiently will be a big enough win that I will wait another year before shelling out $6k “off-cycle”.
Big bummer for me. I was looking forward to performance gains next Friday.
I find it very odd that the new iMac has WiFi 7 but this does not... Also it is so aggravating they compare to 3 generations ago and not the previous generation in the marketing stats. It makes the entire post nearly useless.
It is very aggravating, but if they advertised a comparison to last year's model and showed you small performance gains you might not want to buy it.
A more charitable interpretation is that Apple only thinks that people with computers a few years old need to upgrade, and they aren't advertising to people with a <1 year old MacBook Pro.
I mean you can easily create your own fully meshed P2P group video chat in your browser just using a little bit of JS that would support everyone running 4k, but it will fail the moment you get more than 3-8 people as each persons video stream is eating 25mbps for every side of a peer connection (or 2x per edge in the graph.)
A huge part of group video chat is still "hacks" like downsampling non-speaking participants so the bandwidth doesn't kill the connection.
As we get fatter pipes and faster GPUs streaming will become better.
edit: I mean... I could see a future where realtime video feeds never get super high resolution and everything effectively becomes a relatively seemless AI recreation where only facial movement data is transmitted similar to how game engines work now.
When I have a full team of people with 1080p webcams and a solid connection I can notice the quality. Most of the time not everyone fulfills those requirements and the orchestrator system has to make do
4k for videoconferencing is nuts. The new camera should be an improvement over the old. Plus, being able to show your actual, physical desktop can be Andy too. Using your iPhone as the webcam will still probably give you the best quality especially if you are in a lower light situation.
So far I’m only reading comments here about people wow’d by a lot of things it seemed that M3 pretty much also had. Not seeing anything new besides “little bit better specs”
The M4 is architecturally better than the M3, especially on GPU features IIRC, but you’re right it’s not a total blow out.
Not all products got the M3, so in some lines this week is the first update in quite a while. In others like MBP it’s just the yearly bump. A good performing one, but the yearly bump.
It does and it gets even worse when you realize those stats are only true under very specific circumstances, not typical computer usage. If you benchmarked based on typical computer usage, I think you'd only see gains of 5% or less.
Anyone know of articles that deep dive into "snappiness" or "feel" computer experiences?
Everyone knows SSDs made a big difference in user experience. For the CPU, normally if you aren't gaming at high settings or "crunching" something (compiling or processing video etc.) then it's not obvious why CPU upgrades should be making much difference even vs. years-old Intel chips, in terms of that feel.
There is the issue of running heavy JS sites in browsers but I can avoid those.
The main issue seems to be how the OS itself is optimized for snappiness, and how well it's caching/preloading things. I've noticed Windows 10 file system caching seems to be not very sophisticated for example... it goes to disk too often for things I've accessed recently-but-not-immediately-prior.
Similarly when it comes to generating heat, if laptops are getting hot even while doing undemanding office tasks with huge periods of idle time then basically it points to stupid software -- or let's say poorly balanced (likely aimed purely at benchmark numbers than user experience).
If your goal is to sell more MBPs (and this is marketing presentation) then, judging by the number of comments that have the phrase "my M1" and the top comment, it seems like M1 vs M4 is the right comparison to make. Too many people are sticking with their M1 machines. Including me.
It's actually interesting to think about. Is there a speed multiplier that would get me off this machine? I'm not sure there is. For my use case the machine performance is not my productivity bottleneck. HN on the otherhand... That one needs to be attenuated. :)
I have to admit, 4 generations in, 1.8x is decent but slightly disappointing all the same.
I'd really like to justify upgrading, but a $4k+ spend needs to hit greater than 2x for me to feel it's justified. 1.8x is still "kind of the same" as what I have already.
The adjectives in the linked article are nausiating. Apple’s marketing team fail as decent humans writting such drivel.
Give us data, tell us whats new, and skip the nonsense buzz filling adjectives.
To quote Russell Brand, just say he sat down, not that he placed his luscious ass in silk covered trousers on a velvetly smooth chair, experiencing pleasure as the strained thigh muscles received respite after gruelling on their feet watching a lush sunset in a cool summers evening breeze.
> Up to 4.6x faster build performance when compiling code in Xcode when compared to the 16‑inch MacBook Pro with Intel Core i9, and up to 2.2x faster when compared to the 16‑inch MacBook Pro with M1 Max.
OK, that's finally a reason to upgrade from my M1.
I'm pleased that the Pro's base memory starts at 16 GB, but surprised they top out at 32 GB:
> ...the new MacBook Pro starts with 16GB of faster unified memory with support for up to 32GB, along with 120GB/s of memory bandwidth...
I haven't been an Apple user since 2012 when I graduated from college and retired my first computer, a mid-2007 Core2 Duo Macbook Pro, which I'd upgraded with a 2.5" SSD and 6GB of RAM with DDR2 SODIMMs. I switched to Dell Precision and Lenovo P-series workstations with user-upgradeable storage and memory... but I've got 64GB of RAM in the old 2019 Thinkpad P53 I'm using right now. A unified memory space is neat, but is it worth sacrificing that much space? I typically have a VM or two running, and in the host OS and VMs, today's software is hungry for RAM and it's typically cheap and upgradeable outside of the Apple ecosystem.
I haven't done measurements on this, but my Macbook Pro feels much faster at swapping than any Linux or Windows device I've used. I've never used an M.2 SSD so maybe that would be comparable, but swapping is pretty much seamless. There's also some kind of memory compression going on according to Activity Monitor, not sure if that's normal on other OSes.
Yes, other M.2 SSDs have comparable performance when swapping, and other operating systems compress memory, too — though I believe not as much as MacOS.
Although machines with Apple Silicon swap flawlessly, I worry about degrading the SSD, which is non-replaceable. So ultimately I pay for more RAM and not need swapping at all.
Degrading the SSD is a good point. This is thankfully a work laptop so I don't care if it lives or dies, but it's something I'll have to consider when I eventually get my own Mac.
It seems you need the M4 Max with the 40-core GPU to go over 36GB.
The M4 Pro with 14‑core CPU & 20‑core GPU can do 48GB.
If you're looking for ~>36-48GB memory, here's the options:
$2,800 = 48GB, Apple M4 Pro chip with 14‑core CPU, 20‑core GPU
$3,200 = 36GB, Apple M4 Max chip with 14‑core CPU, 32‑core GPU
$3,600 = 48GB, Apple M4 Max chip with 16‑core CPU, 40‑core GPU
So the M4 Pro could get you a lot of memory, but less GPU cores. Not sure how much those GPU cores factor in to performance, I only really hear complaints about the memory limits... Something to consider if looking to buy in this range of memory.
Of course, a lot of people here probably consider it not a big deal to throw an extra 3 grand on hardware, but I'm a hobbyist in academia when it comes to AI, I don't big 6-figure salaries :-)
Somehow I got downvoted for pointing this out, but it's weird that you have to spend an extra $800 USD just to be able to surpass 48gb, and "upgrading" to the base level Max chip decreases your ram limit, especially when the M4 Pro on the Mac Mini goes up to 64gb. Like... that's a shit load of cash to put out if you need more ram but don't care for more cores. I was really hoping to finally upgrade to something with 64gb, or maybe 96 or 128 if it decreased in price, but it's they removed the 96 and kept 64 and 128 severely out of reach.
Do I get 2 extra CPU cores, build a budget gaming PC, or subscribe to creative suite for 2.5 years!?
It looks like different versions of the ‘Pro’ based on core count and memory bandwidth. Im assuming the 12c Mini M4 Pro has the same memory bandwidth/channels enabled as the 14c MBP M4 Pro, enabling the 64GB. My guess would be related to binning and or TDP.
I really like these new devices, but I’ve found that the latest MacBook Air (M3) is sufficient for my needs as a manager and casual developer. My MacBook Pro M1 Max has essentially become a desktop due to its support for multiple monitors, but since the Mac Mini M4 Pro can also support up to three external displays, I’m considering selling the MacBook Pro and switching to the Mini. I’ve also noticed that the MacBook Pro’s battery, as a portable device, is less efficient in terms of performance/battery (for my usage) compared to the MacBook Air.
Regarding LLMs, the hottest topic here nowadays, I plan to either use the cloud or return to a bare-metal PC.
Most retailers have had the older models on closeout for a few weeks now. Best Buy, Amazon and Costco have had the M3 models for a few hundred off depending on models.
The M-series macbooks depreciate in value far slower than any of the Intel models. M1 base models can still sell for nearly $1k. It's difficult to find a really good deal.
Watch SlickDeals. I think it was this time last year where lots of refurbs/2 generation old machines were going for massive discounts. Granted they were M1 machines, but some had 64GB RAM and 4TB drives for like $2700. Microcenter and B&H are good ones to watch as well.
1,366 comments
[ 3.0 ms ] story [ 361 ms ] threadI hope that the response times have improved, because it has been quite poor for a 120 Hz panel.
(I tend to feel if you want something specialized, you gotta pay for the expensive model)
The so-called "antiglare" option wasn't true matte. You'd really have to go back to 2008.
For my current laptop, I finally broke down and bought a tempered glass screen protector. It adds a bit of glare, but wipes clean — and for the first time I have a one-year-old MacBook that still looks as good as new.
Will that break the display like it did for tape over the webcam?
No space grey?!
A specced out Mac Studio (M2 being the latest model as of today) isn't cheap, but it can run 180B models, run them fast for the price, and use <300W of power doing it. It idles below 10W as well.
> MacBook Air: The World’s Most Popular Laptop Now Starts at 16GB
> MacBook Air is the world’s most popular laptop, and with Apple Intelligence, it’s even better. Now, models with M2 and M3 double the starting memory to 16GB, while keeping the starting price at just $999 — a terrific value for the world’s best-selling laptop.
Only in US it seems. India got a price increase by $120.
I’ve been using the exact model for about a year and I rarely find limitations for my typical office type work. The only time I’ve managed to thermally throttle it has been with some super suboptimal Excel Macros.
I believe the rumor is that the MacBook Air will get the update to M4 in early spring 2025, February/March timeline.
That said, I believe you. Some press gets a hands-on on Wednesday (today) so unless they plan to pre-announce something (unlikely) or announce software only stuff, I think today is it.
Also, Studio and Pro are hanging there.
The streaming apps virtually all support downloading for offline viewing on iPhone, but the Apple TV just becomes a paperweight when the internet goes out, because I'm not allowed to use the 128GB of storage for anything.
If they're not going to let you use the onboard storage, then it seems unlikely for them to let you use USB storage. So, first, I would like them to change their app policies regarding internal storage, which is one of the purely software improvements I would like to see.
But there are some cases like e.g. watching high-res high-FPS fractal zoom videos (e.g. https://www.youtube.com/watch?v=8cgp2WNNKmQ) where even brief random skipped frames from other things trying to use WiFi at the same time can be really noticeable and annoying.
Unfortunately Apple won’t tell you until the day they sell the machines.
And yes, with enough RAM, it is a surprisingly good dev laptop.
And it's still only 512GB! The M4 version coming in the new year will surely bump this up to something more sensible.
On my Mac I don't have any of these things, it's mostly for programming and some packages. I'm almost always connected to Wi-Fi (except on planes) so I don't really need any photos or videos.
The only people that I see have high storage requirements on Macs are probably video/media creators? As a programmer I'm totally fine with 512GB, but could probably live with 256GB if I wanted to be super lean.
Makes me wonder what else will be updated this week (Studio or Mac Pro)?
Just kidding! As an Apple Shareholder I feel like you should take what Apple gives you and accept the price. ;)
I get that small form factor requires soldered ram, but RAM is also very cheap these days.
> All Apple Silicon Macs are in scope, as well as future generations as development time permits. We currently have support for most machines of the M1 and M2 generations.[^1][^2]
[^1]: https://asahilinux.org/about/
[^2]: https://asahilinux.org/fedora/#device-support
It's 2024, and I still see most Windows users carrying a mouse to use with their laptop.
https://softwareengineeringdaily.com/2024/10/15/linux-apple-...
1. Nested virtualization doesn't work in most virtualization software, so if your workflow involves running stuff in VMs it is not going to work from within another VM. The exception is apparently now the beta version of UTM with the Apple Virtualization backend, but that's highly experimental.
2. Trackpad scrolling is emulated as discrete mouse wheel clicks, which is really annoying for anyone used to the smooth scrolling on macOS. So what I do is use macOS for most browsing and other non-technical stuff but do all my coding in the Linux VM.
https://developer.apple.com/documentation/virtualization/vzg...
This is the sad situation on my M2 MacBook Pro :(
Bandwidth (GB/s) = (Data Rate (MT/s) * Channel Width (bits) * Number of Channels) / 8 / 1000
(8800 MT/s * 64 bits * 8 channels) / 8 / 1000 = 563.2 GB/s
This is still half the speed of a consumer NVidia card, but the large amounts of memory is great, if you don't mind running things more slowly and with fewer libraries.
A top tier hosted model is fast and 100.
Past what specialized models can do, it's about a mixture/agentic approach and next level, nuclear power scale. Having a computer with lots of relatively fast RAM is not magic.
Most laptops will be 2 DIMMS (probably soldered).
The vast majority of any x86 laptop or desktops are 128 bits wide. Often 2x64 bit channels up till last year or so, now 4x32 bit DDR5 in the last year or so. There are some benefits to 4 channels over 2, but generally you are still limited by 128 bits unless you buy a Xeon, Epyc, or Threadripper (or Intel equiv) that are expensive, hot, and don't fit in SFFs or laptops.
So basically the PC world is crazy behind the 256, 512, and 1024 bit wide memory busses apple has offered since the M1 arrived.
Was this example intended to describe any particular device? Because I'm not aware of anything that operates at 8800 MT/s, especially not with 64-bit channels.
I believe I saw somewhere that the actual chips used are LPDDR5X-8533.
Effectively the parents formula describes the M4 max, give or take 5%.
But it has more than 2x longer battery life and a better keyboard than a GPU card ;)
An RTX4090 or H100 has memory extremely close to the processor but I don't think you would call it unified memory.
A huge part of optimizing code for discrete GPUs is making sure that data is streamed into GPU memory before the GPU actually needs it, because pushing or pulling data over PCIe on-demand decimates performance.
If you’re forking out for H100’s you’ll usually be putting them on a bus with much higher throughput, 200GB/s or more.
https://en.wikipedia.org/wiki/DDR5_SDRAM (info from the first section):
> DDR5 is capable of 8GT/s which translates to 64 GB/s (8 gigatransfers/second * 64-bit width / 8 bits/byte = 64 GB/s) of bandwidth per DIMM.
So for example if you have a server with 16 DDR5 DIMMs (sticks) it equates to 1,024 GB/s of total bandwidth.
DDR4 clocks in at 3.2GT/s and the fastest DDR3 at 2.1GT/s.
DDR5 is an impressive jump. HBM is totally bonkers at 128GB/s per DIMM (HBM is the memory used in the top end Nvidia datacenter cards).
Cheers.
Not quite as it depends on number of channels and not on the number of DIMMs. An extreme example: put all 16 DIMMs on single channel, you will get performance of a single channel.
This means that in practice, consumer x86 CPUs have only 128GiB/s of DDR5 memory bandwidth available (regardless of the number of DIMM slots in the system), because the vast majority of them only offer two memory channels. Server CPUs can offer 4, 8, 12, or even more channels, but you can't just install 16 DIMMs and expect to get 1024GiB/s of bandwidth, unless you've verified that your CPU has 16 memory channels.
Happy Halloween!
Unrelated but unified memory is a strange buzzword being used by Apple. Their memory is no different than other computers. In fact, every computer without a discrete GPU uses a unified memory model these days!
And yes, the impressive part is that this kind of bandwidth is hard to get on laptops. I suppose I should have been a bit more specific in my remark.
Many integrated graphics segregate the memory into CPU owned and GPU owned, so that even if data is on the same DIMM, a copy still needs to be performed for one side to use what the other side already has.
This means that the drivers, etc, all have to understand the unified memory model, etc. it’s not just hardware sharing DIMMs.
APUs with shared everything are not a new concept, they are actually older than programmable graphics coprocessors…
https://www.heise.de/news/Gamescom-Playstation-4-bietet-Unif...
So apple manages decent GPU performance, a tiny package, and great battery life. It's much harder on the PC side because every laptop/desktop chip from Intel and AMD use a 128 bit memory bus. You have to take a huge step up in price, power, and size with something like a thread ripper, xeon, or epyc to get more than 128 bit wide memory, none of which are available in a laptop or mac mini size SFF.
Memory interface width of modern CPUs is 64-bit (DDR4) and 32+32 (DDR5).
No CPU uses 128b memory bus as it results in overfetch of data, i.e., 128B per access, or two cache lines.
AFAIK Apple uses 128B cache lines, so they can do much better design and customization of memory subsystem as they do not have to use DIMMs -- they simply solder DRAM to the motherboard, hence memory interface is whatever they want.
Sure, per channel. PCs have 2x64 bit or 4x32 bit memory channels.
Not sure I get your point, yes PCs have 64 bit cache lines and apple uses 128. I wouldn't expect any noticeable difference because of this. Generally cache miss is sent to a single memory channel and result in a wait of 50-100ns, then you get 4 or 8 bytes per cycle at whatever memory clock speed you have. So apple gets twice the bytes per cache line miss, but the value of those extra bytes is low in most cases.
Other bigger differences is that apple has a larger page size (16KB vs 4KB) and arm supports a looser memory model, which makes it easier to reach a large fraction of peak memory bandwidth.
However, I don't see any relationship between Apple and PCs as far as DIMMS. Both Apple and PCs can (and do) solder dram chips directly to the motherboard, normally on thin/light laptops. The big difference between Apple and PC is that apple supports 128, 256, and 512 bit wide memory on laptops and 1024 bit on the studio (a bit bigger than most SFFs). To get more than 128 bits with a PC that means no laptops, no SFFs, generally large workstations with Xeon, Threadrippers, or Epyc with substantial airflow and power requirements
Also important to consider that the RTX 4090 has a relatively tiny 384-bit memory bus. Smaller than the M1 Max's 512-bit bus. But the RTX 4090 has 1 TB/s bandwidth and significantly more compute power available to make use of that bandwidth.
The M4 max is definitely not a 4090 killer, does not match it in any way. It can however work on larger models than the 4090 and have a battery that can last all day.
My memory is a bit fuzzy, but I believe the m3 max did decent on some games compared to the laptop Nvidia 4070 (which is not the same as the desktop 4070). But highly depended on if the game was x86-64 (requiring emulation) and if it was DX11 or apple native. I believe apple claims improvements in metal (the Apple's GPU lib) and that the m4 GPUs have better FP for ray tracing, but no significant changes in rasterized performance.
I look forward to the 3rd party benchmarks for LLM and gaming on the m4 max.
It's not really a new idea, just unusual in computers. The custom SOCs that AMD makes for Playstation and Xbox have wide (up to 384-bit) unified memory buses, very similar to what Apple is doing, with the main distinction being Apples use of low-power LPDDR instead of the faster but power hungrier GDDR used in the consoles.
Servers do have many channels but they run relatively slower memory
* Specifically, it being on-die
Also, DRAM is never on-die. On-package, yes, for Apple's SoCs and various other products throughout the industry, but DRAM manufacturing happens in entirely different fabs than those used for logic chips.
https://en.wikipedia.org/wiki/EDRAM
And no, manaskarekar, the M4 Max does 546 GB/s not GBps (which would be 8x less!).
GB/s and GBps mean the same thing, though GB/s is the more common way to express it. Gb/s and Gbps are the units that are 8x less: bits vs Bytes.
GB/s is the same thing as GBps
The "ps" means "per second"
On PC desktops I always recommend getting a mid-range tower server precisely for that reason. My oldest one is about 8 years old and only now it's showing signs of age (as in not being faster than the average laptop).
Or you could buy a M3 max laptop for $4k, get 10+ hour battery life, have it fit in a thin/light laptop, and still get 546GB/sec. However those are peak numbers. Apple uses longer cache lines (double), large page sizes (quadruple), and a looser memory model. Generally I'd expect nearly every memory bandwidth measure to win on Apple over AMD's turin.
At full tilt an M3 Max will consume 50 to 75 watts, meaning you get 1 to 2 hours of runtime at best, if you use the thing full tilt.
That's the thing I find funny about the Apple Silicon MBP craze, sure they are efficient but if you use the thing as a workstation, battery life is still not good enough to really use unplugged.
Most claiming insane battery life are using the thing effectively as an information appliance or a media machine. At this game the PC laptops might not be as efficient but the runtime is not THAT different provided the same battery capacity.
There are, in my experience, professionals who want to use the best tools someone else builds for them, and professionals who want to keep iterating on their tools to make them the best they can be. It's the difference between, say, a violin and a Eurorack. Neither's better or worse, they're just different kinds of tools.
I was sorely tempted by the Mac studio, but ended up with a 96GB ram Ryzen 7900 (12 core) + Radeon 7800 XT (16GB vram). It was a fraction of the price and easy to add storage. The Mac M2 studio was tempting, but wasn't refreshed for the M3 generation. It really bothered me that the storage was A) expensive, B) proprietary, C) tightly controlled, and D) you can't boot without internal storage.
Even moving storage between Apple studios can be iffy. Would I be able to replace the storage if it died in 5 years? Or expand it?
As tempting as the size, efficiency, and bandwidth were I just couldn't justify top $ without knowing how long it would be useful. Sad they just didn't add two NVMe ports or make some kind of raw storage (NVMe flash, but without the smarts).
This was really driven home to me by my recent purchase of an Optane 905p, a drive that is both very fast and has an MTBF measured in the hundreds of years. Short of a power surge or (in California) an earthquake, it's not going to die in my lifetime -- why should I not keep using it for a long time?
Many kinds of professionals are completely fine with having their Optanes and what not only be plugged in externally, though, even though it may mean their boot drive will likely die at some point. That's completely okay I think.
This is one of the reasons the "3D vcache" stuff with the giant L3 cache is so effective.
On PCs some other hardware (notably the SSD) comes with its own memory. But here it's shared with the main DRAM too.
This is not necessarily a performance improvement, it can avoid copies but also means less is available to the CPU.
EDIT: wtf what's so bad about this comment that it deserves being downvoted so much
Calling something "shared" makes you think: "there's not enough of it, so it has to be shared".
Calling something "unified" makes you think: "they are good engineers, they managed to unify two previously separate things, for my benefit".
For Apple to have come up with using the term "unified memory" to describe this kind of architecture, they would've needed to come up with it at least before 2016, meaning A9 chip or earlier. I have paid some attention to Apple's SoC launches through the years and can't recall them touting it as a feature in marketing materials before the M1. Do you have something which shows them using the term before 2016?
To be clear, it wouldn't surprise me if it has been used by others before Intel did in 2015 as well, but it's a starting point: if Apple hasn't used the term before then, we know for sure that they didn't come up with it, while if Apple did use it to describe A9 or earlier, we'll have to go digging for older documents to determine whether Apple came up with it
Good news for anyone who wants to run 405B LMs locally...
Smart move by Apple
Right?
I think it’s super interesting to know real life workflows and performance of different LLMs and hardware, in case you can direct me to other resources. Thanks !
An M2 is according to a reddit post around 27 tflops
So < 1/10 the performance of just computation. let alone the memory.
What workflow would use something like this?
Memory and memory bandwidth matters most for inferencing. 819.2 GB/s for M2 Ultra is less than half that of A100, but having 192GB of RAM instead of 80gb means they can run inference on models that would require THREE of those A100s and the only real cost is that it takes longer for the AI to respond.
3 A100 at $5300/mo each for the past 2 years is over $380,000. Considering it worked for them, I'd consider it a massive success.
From another perspective though, they could have bought 72 of those Ultra machines for that much money and had most devs on their own private instance.
The simple fact is that Nvidia GPUs are massively overpriced. Nvidia should worry a LOT that Apple's private AI cloud is going to eat their lunch.
Small correction: the M2 Ultra isn't found in laptops, its in the Studio.
:P
Somewhat niche case, I know.
- Apple: all the capacity and bandwidth, but no compute to utilize it
- AMD/Nvidia: all the compute and bandwidth, but no capacity to load anything
- DDR5: all the capacity, but no compute or bandwidth (cheap tho)
I wonder if that has changed or is about to change as Apple pivots their devices to better serve AI workflows as well.
The M4-Max I just ordered comes with 128GB of RAM.
If you need to do some work offline, or for some reason the place you work blocks access to cloud providers, it's not a bad way to go, really. Note that if you're on battery, heavy LLM use can kill your battery in an hour.
The linked Apple Store page says "MacBook Pro blasts forward with the M3, M3 Pro, and M3 Max chips" so it seems like the old version of the page still?
Looking at how long the 8gb lasted it's a pretty sure bet that now you won't need to upgrade for a good few years.
I mean, I have a MacBook air with 16gb of ram and it's honestly working pretty well to this day. I don't do "much" on it though but not many people do.
I'd say the one incentive a MacBook Pro has over the air is the better a screens and better speakers. Not sure if it's worth the money.
> I mean, I have a MacBook air with 16gb of ram and it's honestly working pretty well to this day. I don't do "much" on it though but not many people do.
If an HN user can get along with 16gb on their MacBook Air for the last X years, most users were able to get by with 8gb.
They are supposed to be "green" but they encourage obsolescence.
8GB is fine for most use cases. Part of my gig is managing a huge global enterprise with six figures of devices. Metrics demonstrate that the lower quartile is ok with 8GB, even now. Those devices are being retired as part of the normal lifecycle with 16GB, which is better.
Laptops are 2-6 year devices. Higher end devices always get replaced sooner - you buy a high end device because the productivity is worth spending $. Low end tend to live longer.
Or you could get a framework and you could actually upgrade parts that are worth upgrading - instead of upgrade as in buying a new one
It's fine, but the issue is linux sleep/hibernate - battery drain. To use the laptop after a few days, I have to plug it in and wait for it to charge a little bit because the battery dies. I have to shut it down (not just close the screen) before flying or my backpack becomes a heater and the laptop dies. To use a macbook that's been closed for months I just open it and it works. I'll pay double for that experience. If I want a computer that needs to be plugged in to work I have a desktop for that already. The battery life is not good either.
Maybe it's better now if I take the time to research what to upgrade, but I don't have the time to tinker with hardware/linux config like I did a few years ago.
That's what I loved about my 2013 MBP but didn't have the same experience with the newer models anymore either.
I think "months" or even weeks is pushing it anyways, though a good proper suspend would be much appreciated.
Personally I use my laptop every other day so don't really have the same battery issue there
I don't really see a world where this machine doesn't last me a few more years. If there's anything i'd service would be the battery, but eh. It lasts more than a few hours and I don't go out much.
Anyways to each their own, I also had things break and repairability isn't really a thing with Apple hardware (while it easily could be if they wanted - even if difficult and by certified technicians instead of myself)
This is misleading:
https://news.ycombinator.com/item?id=25074959
"macOS sends hashes of every opened executable to some server of theirs"
To be fair, the link in this story is to a press release. Arguably there are probably many things in it that can be considered "misleading" in certain contexts.
Wish I could spin up a Linux OS on the hardware though. Not a bright spot for me.
https://asahilinux.org
It won't have all the niceties / hardware support of MacOS, but it seamlessly coexists with MacOS, can handle the GPU/CPU/RAM with no issues, and can provide you a good GNU/Linux environment.
I guess you could have a physical MBP in your house and connect it to some bring-your-own-infrastructure CI setup, but most people wouldn't want to do that.
It's a pretty hard problem to partially automate for setups with an engineer in the room. It doesn't sound at all feasible for an unattended data center setup that's designed to host Xcode for compiling apps under macOS.
(This isn't a dig on the Asahi project btw, I think it's great).
However, it doesn't support snapshots for Linux, so you need to power down each session.
I insist my 2020 Macbook M1 was the best purchase I ever made
The battery performance is incredible too.
M1 series machines are going to be fine for years to come.
Actually wasn't M1 itself an evolution / upscale of their A series CPUS that by now they've been working on since... before 2010, the iPhone 4 was the first one with their own CPU, although the design was from Samsung + Intrinsity, it was only the A6 that they claimed was custom designed by Apple.
Bought a reasonably well-specced Intel Air for $1700ish. The M1s came out a few months later. I briefly thought about the implication of taking a hit on my "investment", figured I might as well cry once rather than suffer endlessly. Sold my $1700 Intel Air for $1200ish on craigslist (if I recall correctly), picked up an M1 Air for about that same $1200 pricepoint, and I'm typing this on that machine now.
That money was lost as soon as I made the wrong decision, I'm glad I just recognized the loss up front rather than stewing about it.
That said...scummy move by Apple. They tend to be a little more thoughtful in their refresh schedule, so I was caught off guard.
- More RAM, primarily for local LLM usage through Ollama (a bit more overhead for bigger models would be nice)
- A bit niche, but I often run multiple external displays. DisplayLink works fine for this, but I also use live captions heavily and Apple's live captions don't work when any form of screen sharing/recording is enabled... which is how Displaylink works. :(
Not quite sold yet, but definitely thinking about it.
I'm definitely still happy with it, but job offers upgrade to M4 so... why not?
I've never kept any laptop as long as I've kept the M1. I was more or less upgrading yearly in the past because the speed increases (both in the G4 and then Intel generations) were so significant. This M1 has exceeded my expectations in every category, it's faster quieter and cooler than any laptop i've ever owned.
I've had this laptop since release in 2020 and I have nearly 0 complaints with it.
I wouldn't upgrade except the increase in memory is great, I don't want to have to shut down apps to be able to load some huge LLMs, and, I ding'ed the top case a few months ago and now there's a shadow on the screen in that spot in some lighting conditions which is very annoying.
I hope (and expect) the M4 to last just as long as my M1 did.
My 2015 MBP would like to have a word.
It’s the only laptop purchase I’ve made. I still use it to this day, though not as regularly.
I will likely get a new MBP one of these days.
I also have a M1 from work that is absolutely wonderful, but I think it's time for me to upgrade the 2015 with one of these new M4s.
The longevity of Macbooks is insanely good.
Longevity is not only a thing of MBPs. OTOH, IIRC, some 2017-2019 MBPs (before the Mx switch) were terrible for longevity, given their problematic keyboard.
Last one with upgrade capabilities, now it has two fast SSDs and maximum Ram. I changed the battery once.
Only shame is that it doesn’t get major MacOS upgrades anymore.
Still good enough to browse the web, do office productivity and web development.
12 years of good use, I am not sure I can get so much value anywhere now
Thanks for reminding me that everything is possible, I may try Opencore to keep it even longer !
Sold mine last year for $100 to some dude who claimed to have some software that only runs on that specific laptop. I didn't question it.
I think this M-series macbook airs are a worthy successor to the 2012 MBP. I fully intend to use this laptop for at least the same amount of time, ideally more. The lack of replaceable battery will probably be the eventual killer, which is a shame.
Rebuilding a bunch of Docker images on an older intel mac is quite the slow experience if you're doing it multiple times per day.
Anything you can buy online ships with all required screw drivers and dozen of Youtube videos or ifixit will give you step by step instructions.
10-15 minutes and you'll have the old battery replaced all by yourself.
It's that simple.
The 2012 MBP 15" Retina was probably the only machine I bought where the performance actually got better over the years, as the OS got more optimized for it (the early OS revisions had very slow graphics drivers dealing with the retina display)
The M1 Pro on the other hand, that was a true upgrade. Just a completely different experience to any Apple Intel laptop.
It's been too long, I guess I had blocked out just just how terrible those Intel MacBooks were.
For it's time, the 2015 model was a fantastic package: reliable and robust in form and function.
Would've kept going on it had Apple silicon and 14 inch not come around.
Barring super niche LLM use cases, I don't see why one would need to upgrade.
When you upgrade, prepare to be astonished.
The performance improvement is difficult to convey. It's akin to traveling by horse and buggy. And then hopping into a modern jetliner, flying first class.
It's not just speed. Display quality, build quality, sound quality, keyboard quality, trackpad, ports, etc., have all improved considerably.
A family 2018 Macbook Air got a second life with a battery replacement. Cheap kit from Amazon, screwdrivers included, extremely easy to do. Still in use, no problems.
When I migrated all my laptops to SSDs (lenovos at the time, so it was drop-dead simple), I thought to myself, "this is a once-in-a-generation feeling". I didn't think I would ever be impressed by a laptop's speed ever again. It was nice to be wrong.
Traded it for an M1 Air in 2021 and was astonished at how much faster it was. It even blew away my 2019 16" from work.
You're going to be even more blown away!
I recently replaced it with a used MBA M1, 16GB, 2TB.
It's insane how much faster it is, how long the battery lasts and how cool and silent it is. Completely different worlds.
It has gotten significantly slower the last 2 years, but the more obvious issue is the sound, inability to virtual background, and now lack of software updates.
But if you had told me I'd need to replace it in 2022 I wouldn't believe you
I was considering upgrading to an M3 up until about a month ago when Apple replaced my battery, keyboard, top case, and trackpad completely for free. An upgrade would be nice as it no longer supports the latest MacOS, but at this point, I may just load Ubuntu on the thing and keep using it for another few years. What a machine.
Can't get the latest macOS on it though, but otherwise it still works perfectly well.
Kinda considering upgrading it to a used M1/M2 one at some point.
I have either assembled my own desktop computers or purchased ex corporate Lenovo over the years with a mix of Windows (for gaming obviously) and Linux and only recently (4 years ago) been given a MBP by work as they (IT) cannot manage Linux machines like they do with MacOS and Windows.
I have moved from an intel i5 MBP to a M3 Pro (?) and it makes me want to throw away my dependable ThinkPad/Fedora machine I still uses for personal projects.
Could you get more money by selling it? Sure. But it's hard to be the convenience. They ship you a box. You seal up the old device and drop it off at UPS.
I also build my desktop computers with a mix of Windows and Linux. But those are upgraded over the years, not regularly.
My laptop is my work life and my personal life.
I spend easily 100 hours a week using it not-as-balanced-as-it-should-be between the two.
I don't buy them because I need something new, I buy them because in the G4/Intel era, the iterations were massive and even a 20 or 30% increase in speed (which could be memory, CPU, disk -- they all make things faster) results in me being more productive. It's worth it for me to upgrade immediately when apple releases something new, as long as I have issues with my current device and the upgrade is enough of a delta.
M1 -> M2 wasn't much of a delta and my M1 was fine. M1 -> M3 was a decent delta, but, my M1 was still fine. M1 -> M4 is a huge delta (almost double) and my screen is dented to where it's annoying to sit outside and use the laptop (bright sun makes the defect worse), so, I'm upgrading. If I hadn't dented the screen the choice would be /a lot/ harder.
I love ThinkPads too. Really can take a beating and keep on going. The post-IBM era ones are even better in some regards too. I keep one around running Debian for Linux-emergencies.
Me too. Only one complaint. After I accidentally spilled a cup of water into it on an airplane, it didn't work.
(However AppleCare fixed it for $300 and I had a very recent backup. :) )
What’s more annoying is that I’d jus to get a new one and recycle this one, but the SSD is soldered on. Good on you for having a backup.
Do not own a Mac unless you bought it used or have AppleCare.
And yeah, this incident reminded me of why it's important to back up as close to daily as you can, or even more often during periods when you're doing important work and want to be sure you have the intermediate steps.
Never had AppleCare or any other extended warranty program.
Did just fine up to now.
1) a slight overall savings, though I'm not sure about that. 2) a lack of stress when something breaks. Even if there isn't an overall savings, for me it's been worth it because of that.
Certainly, my recent Mac repair would have cost $1500 and I only paid $300, and I think I've had the machine for about 3 years, so there's a savings there but considerably less recent stress. That's similar to the experience I've had all along, although this recent expense would have probably been my most-expensive repair ever.
As I've noted in a sibling comment, I'll probably stop purchasing mobile Macs until the repair story on Macbooks is improved -- the risk for accidents and repairs is simply much higher on portable machines. That's only going to happen through third-party repair (which I think would simultaneously lead Apple to lower their first-party repair costs, too).
I’ve underwritten my own Mac ownership since the very first Intel MacBook Pro and just like you I’ve been just fine.
Apple is putting raw NAND chips on the board (and yes soldering them) and the controller for the SSD is part of the M-series chip. Yes, apple could use NVMe here if you ignore the physical constraints and ignore fact that it wouldn't be quite as fast and ignore the fact that it would increase their BOM cost.
I'm not saying Apple is definitively correct here, but, it's good to have choice and Apple is the only company with this kind of deeply integrated design. If you want a fully modular laptop, go buy a framework (they are great too!) and if you want something semi-modular, go buy a ThinkPad (also great!).
Day to day I don't mind but when needs change or something breaks it's unfortunate to have to replace the whole machine to fix it.
I don't truly mind that they solder on the SSD, embed the controller into the processor -- you're right that it's great we have choice here. I mind the exuberant repair cost _on top of_ Apple's war on third party repair. Apple is the one preventing me to have choice here, I have to do the repair through them, or wait until schematics are smuggled out of China and used/broken logic boards are available so that the repair costs what it should: $300 to replace 2 chips on my logic board (still mostly labor, but totally a fair price).
I love Apple for their privacy focus and will continue to support them because I need to do Mac and iOS development, but I will likely stop buying mobile workstations from them for this reason, the risk of repair is simply much higher and not worth this situation.
I actually use my laptop on my lap commonly and I think the i9 was going to sterilize me.
Now on M2 MBP and will probably be using it for a very long time.
What different lives we live. This first M1 was in November 2020. Not even four years old. I’ve never had a [personal] computer for _less_ time than that. (Work, yes, due to changing jobs or company-dictated changes/upgrades)
That’s a feature, not a bug, for some. When I upgraded to an M series chip MacBook, I had to turn up the heat because I no longer had my mini space heater.
Other example - I'm by no means rich, but I have a $300 mechanical keyboard - it doesn't make me type faster and it doesn't have additional functionality to a regular $30 Logitech one - but typing on it feels so nice and I spend so much of my life doing it, that to me it's completely justified and needed to have this one then.
I still have a running Thinkpad R60 from 2007, a running Thinkpad T510 from 2012, and a modified running Thinkpad X61 (which I re-built as an X62 using the kit from 51nb in 2017 with a i7-5600U processor, 32 GB of RAM and a new display) in regular use. The latter required new batteries every 2 years, but was my main machine until 2 weeks ago when I replaced it with a ThinkCentre. During their time as my main machine, each of these laptops was actively used around 100 hours per week, and was often running for weeks without shutdown or reboot. The only thing that every broke was the display of the R60 which started to show several green vertical bars after 6 years, but replacement was easy.
This is the best machine I have ever owned. It is so completely perfect in every way. I can't imagine replacing it for many many years.
Best way I've discovered to find those sorts of deals is to use Slickdeals and set up alerts.
(of course, everyone else has a macbook too, there's always someone that can lend me a charger. Bonus points that the newer macbooks support both magsafe and USB-C charging. Added bonus points that they brought back magsafe and HDMI ports)
Personally, I practically never use MagSafe, because the convenience of USB C charging cables all over the house outweighs the advantages of MagSafe for me.
With whisky i feel like id never need anything else. That said, the benchmark jump in the m4 has me thinking i should save up and grab a refurb in a year or two
I don't understand how people are enamored with those things, sure it's better in some way than what it was before but it's also very compromised for the price.
The only thing to keep in mind, is that the M1 was the first CPU in the transition from Intel CPUs (+ AMD GPUs) to Apple Silicon. The M1 was still missing a bunch of things from earlier CPUs, which Apple over time added via the M1 Pro and other CPUs. Especially the graphics part was sufficient for a small laptop, but not for much beyond. Better GPUs and media engines were developed later. Today, the M3 in a Macbook Air or the M4 in the Macbook Pro have all of that.
For me the biggest surprise was how well the M1 Macbook Air actually worked. Apple did an outstanding job in the software & hardware transition.
Are they going to claim that 16GB RAM is equivalent to 32GB on Intel laptops? (/sarc)
> Up to 23.8x faster basecalling for DNA sequencing in Oxford Nanopore MinKNOW when compared to the 16-inch MacBook Pro with Core i9, and up to 1.8x faster when compared to the 16-inch MacBook Pro with M1 Pro.
The only downside is that your computer will no longer double as a space heater :p
If it affects your earning power to that extent, you should probably pony up and save in the long run, probably just a few years until you see returns.
Caste system usually can't be bypassed by paying a monthly subscription fee.
I will note that making it a subscription will tend to increase the overall costs, not decrease. In an environment with ready access to credit, I think offering on a subscription basis is worse for consumers?
If it matters that much to you just sell the old one and buy the new. That's your subscription.
Another observation; I've travelled the world and rarely see people who could use robust, secure products the most (vulnerable people) using Apple products. They're all packing second-tier Samsung or LG Androids and old Windows notebooks (there are decent Samsung, LG, Android, Windows products, but that's not what they have access to).
don’t think it’s wise though, i bought a base m1 pro mbp when it launched and don’t feel a need to upgrade at all yet. i’m holding off for a few more years to grab whenever the next major increase in local llm capability and battery life comes.
No Wifi 7. So you get access to the 6 GHz band, but not some of the other features (preamble punching, OFDMA):
* https://en.wikipedia.org/wiki/Wi-Fi_7
* https://en.wikipedia.org/wiki/Wi-Fi_6E
The iPhone 16s do have Wifi 7. Curious to know why they skipped it (and I wonder if the chipsets perhaps do support it, but it's a firmware/software-not-yet-ready thing).
The real use is transferring huge files within the LAN.
Wifi 6e/7 6GHz band with a bunch of non overlapping 160MHz channels is where the juice is at. But even then a lot of devices are limited to smaller channel widths.
One of the features is preamble punching, which is useful in more dense environments:
* https://community.fs.com/article/how-preamble-puncturing-boo...
* https://www.ruckusnetworks.com/blog/2023/wi-fi-7-and-punctur...
MLO helps with resiliency and the improved OFDMA helps with spectrum efficiency as well. It's not just about speed.
Machines can last and be used for years, and it would be a presumably very simple way to 'future proof' things.
And though the IEEE spec hasn't officially been ratified as I type this, it is set to be by the end of 2024. Network vendors are also shipping APs with the functionality, so in coming years we'll see a larger and larger infrastructure footprint going forward.
https://www.tomsguide.com/face-off/wi-fi-6e-vs-wi-fi-7-whats...
Laptops/desktops (with 16GB+ of memory) could make use of the faster speed/more bandwidth aspects of WiFi7 better than smartphones (with 8GB of memory).
I had just assumed that for sure this would be the year I upgrade my M1 Max MBP to an M4 Max. I will not be doing so knowing that it lacks WiFi 7; as one of the child comments notes, I count on getting a solid 3 years out of my machine, so future-proofing carries some value (and I already have WiFi7 access points), and I download terabytes of data in some weeks for the work I do, and not having to Ethernet in at a fixed desk to do so efficiently will be a big enough win that I will wait another year before shelling out $6k “off-cycle”.
Big bummer for me. I was looking forward to performance gains next Friday.
A more charitable interpretation is that Apple only thinks that people with computers a few years old need to upgrade, and they aren't advertising to people with a <1 year old MacBook Pro.
I look at my local source vs the recording, and I am baffled.
After a decade of online meeting software, we still stream 480p quality it seems.
A huge part of group video chat is still "hacks" like downsampling non-speaking participants so the bandwidth doesn't kill the connection.
As we get fatter pipes and faster GPUs streaming will become better.
edit: I mean... I could see a future where realtime video feeds never get super high resolution and everything effectively becomes a relatively seemless AI recreation where only facial movement data is transmitted similar to how game engines work now.
I am asking for good 720p... With how static cam footage is it would be less than 8mbps probably.
It sounds more exciting than M4 is 12.5% faster than M3.
Not all products got the M3, so in some lines this week is the first update in quite a while. In others like MBP it’s just the yearly bump. A good performing one, but the yearly bump.
Back when Moore's law was still working they didn't skip generations like this.
Everyone knows SSDs made a big difference in user experience. For the CPU, normally if you aren't gaming at high settings or "crunching" something (compiling or processing video etc.) then it's not obvious why CPU upgrades should be making much difference even vs. years-old Intel chips, in terms of that feel.
There is the issue of running heavy JS sites in browsers but I can avoid those.
The main issue seems to be how the OS itself is optimized for snappiness, and how well it's caching/preloading things. I've noticed Windows 10 file system caching seems to be not very sophisticated for example... it goes to disk too often for things I've accessed recently-but-not-immediately-prior.
Similarly when it comes to generating heat, if laptops are getting hot even while doing undemanding office tasks with huge periods of idle time then basically it points to stupid software -- or let's say poorly balanced (likely aimed purely at benchmark numbers than user experience).
https://nanoreview.net/en/cpu-compare/apple-m1-vs-amd-ryzen-...
M4 is built with TSMC's 2nd Gen 3nm process. M3 is on the 1st gen 3nm.
For the base M3 vs base M4:
- the CPU (4P+4E) & GPU (8) core counts are the same
- NPU perf is slightly better for M4, I think, (M4's 38TOPS @ INT8 vs M3's 18TOPS @ INT16)
- Memory Bandwidth is higher for M4 (120 GB/s vs 102.4 GB/s)
- M4 has a higher TDP (22W vs 20W)
- M4 has higher transistor count (28B vs 25B)
It's actually interesting to think about. Is there a speed multiplier that would get me off this machine? I'm not sure there is. For my use case the machine performance is not my productivity bottleneck. HN on the otherhand... That one needs to be attenuated. :)
I've just ordered an (almost) top-of-the-range MBP Max, my current machine is an MBP M1-max, so the comparisons are pretty much spot-on for me.
Selling the M1 Ultra Studio to help pay for the M4 MBP Max, I don't think I need the Studio any more, with the M4 being so much faster.
I'd really like to justify upgrading, but a $4k+ spend needs to hit greater than 2x for me to feel it's justified. 1.8x is still "kind of the same" as what I have already.
Give us data, tell us whats new, and skip the nonsense buzz filling adjectives.
To quote Russell Brand, just say he sat down, not that he placed his luscious ass in silk covered trousers on a velvetly smooth chair, experiencing pleasure as the strained thigh muscles received respite after gruelling on their feet watching a lush sunset in a cool summers evening breeze.
> Up to 4.6x faster build performance when compiling code in Xcode when compared to the 16‑inch MacBook Pro with Intel Core i9, and up to 2.2x faster when compared to the 16‑inch MacBook Pro with M1 Max.
OK, that's finally a reason to upgrade from my M1.
https://youtu.be/G0cmfY7qdmY?si=vbgIr8zn9EzB2Xam
> ...the new MacBook Pro starts with 16GB of faster unified memory with support for up to 32GB, along with 120GB/s of memory bandwidth...
I haven't been an Apple user since 2012 when I graduated from college and retired my first computer, a mid-2007 Core2 Duo Macbook Pro, which I'd upgraded with a 2.5" SSD and 6GB of RAM with DDR2 SODIMMs. I switched to Dell Precision and Lenovo P-series workstations with user-upgradeable storage and memory... but I've got 64GB of RAM in the old 2019 Thinkpad P53 I'm using right now. A unified memory space is neat, but is it worth sacrificing that much space? I typically have a VM or two running, and in the host OS and VMs, today's software is hungry for RAM and it's typically cheap and upgradeable outside of the Apple ecosystem.
That's an architectural limitation of the base M4 chip, if you go up to the M4 Pro version you can get up to 48GB, and the M4 Max goes up to 128GB.
Although machines with Apple Silicon swap flawlessly, I worry about degrading the SSD, which is non-replaceable. So ultimately I pay for more RAM and not need swapping at all.
Apple has hardware accelerated compressed swapping.
Windows has compressed swapping.
And Linux is a mess. You have to manually configure a non-resizable compressed zram, or use it without compression on a non-resizable swap partition.
The M4 Pro goes up to 48 GB
The M4 Max can have up to 128 GB
M4 Max 14 core has a single option of 36GB.
M4 Max 16 core lets you go up to 128GB.
So you can actually get more ram with the Pro than the base level Max.
The M4 Pro with 14‑core CPU & 20‑core GPU can do 48GB.
If you're looking for ~>36-48GB memory, here's the options:
$2,800 = 48GB, Apple M4 Pro chip with 14‑core CPU, 20‑core GPU
$3,200 = 36GB, Apple M4 Max chip with 14‑core CPU, 32‑core GPU
$3,600 = 48GB, Apple M4 Max chip with 16‑core CPU, 40‑core GPU
So the M4 Pro could get you a lot of memory, but less GPU cores. Not sure how much those GPU cores factor in to performance, I only really hear complaints about the memory limits... Something to consider if looking to buy in this range of memory.
Of course, a lot of people here probably consider it not a big deal to throw an extra 3 grand on hardware, but I'm a hobbyist in academia when it comes to AI, I don't big 6-figure salaries :-)
Do I get 2 extra CPU cores, build a budget gaming PC, or subscribe to creative suite for 2.5 years!?
Regarding LLMs, the hottest topic here nowadays, I plan to either use the cloud or return to a bare-metal PC.