Let's not forget we're also talking pretty much fanless and a lot less power consumption.
I remember reading "why ARM design will rule" (simpler design) and we're seeing it in the flesh.
I keep hoping to see another manufacturer, e.g. Lenovo making their own ARM machine for thinkpads... It would be incredibly popular I feel. But most likely will never happen.
In the mobile space, sure, but not in the desktop space.
Their chips are good, no doubt, but AMD and Intel are not sleeping at the wheel. Even if Apple chips were available as CPUs and Motherboards with full Windows and Linux support, they wouldn't be the obvious choice for all use cases. The M2 Ultra is incredible, but there are workloads where it stands no chance against a threadripper.
They don't really have a solution for desktop. These are still mobile/laptop chips. Don't get me wrong: these are insanely brilliant paradigm-changing chips that are more than enough for the majority of people.
However, despite all the marketing it still remains to be seen whether they can transition into desktop/workstation chips with these. Especially at a reasonable price.
One that millions of people enjoy, and more are getting into every year passing. Young people who were raised with already decent graphics will not downgrade to 10 years older graphic quality and will thus buy decent gaming PC.
It is actually the market that Apple has chosen that is more niche.
About 80% of Mac sales are laptops (a lot of which are status symbol/ rich executive computer type) so it really does not leave a lot of customers that are into computers.
I believe they made that choice for both historical reasons and trying to maximize profits. But I think they are wrong, and it will bite them eventually. There is no reason a powerful desktop couldn't be just as good at video editing. Actually, the gaming desktops tend to be pretty good at video production too, it just does not have the software Apple has...
The Mac Pro, Mac Studio, Mac Mini, and iMac are basically how Apple covers that market. The Mac Pro is pretty much unobtainable on a budget and definitely aimed at a niche market (high end video/3D graphics production). The Mac Studio is actually pretty nice as a high end work station. The mac mini kind of stops where the Mac Studio begins in terms of performance. If you are maxing out the specs on that, you might be better off getting the entry model Mac Studio. And an imac is what you use if you don't care about performance that much.
But the point is that all of these have a market and some people are spending big on this. I talked to one of my friends recently who runs a high tech printing company and they spent some money on a mac studio recently for doing graphics work. Worth the money according to him.
I have a 14" macbook pro (M1). It's great. I do some flight simuluation on it occasionally with X-plane 12. That runs pretty smooth on this thing and it's one of the few things I have that actually gets this laptop warm.
Apple did a great job of getting great performance with modest power budgets. Comparing their products to water cooled monstrosities is not really that interesting.
They are desktop only, but conceptually they're still much more like mobile chips. With M-series CPUs Apple is either targeting limited power consumption instead of going for max performance (which is a characteristic of mobile chips), or the chips just does not scale to high-end x86 territory despite any possible power consumption increase.
Either way, as soon as power consumption does not matter, price-performance (well, even both price _or_ performance, separately) wise M-series is not competitive with top x86 CPUs.
Yes it is based on a mobile optimized architecture but M2 Ultra is scoring ~90% as high as the top model Intel i9-13900KS on Geekbench 6. It's up there with Intels best. And it's keeping up on the CPU side while having a massive GPU bolted on the die.
One thing to note regarding the M2 Ultra (from Apple marketing):
"It has dedicated, hardware-enabled H. 264, HEVC, and ProRes encode and decode, allowing M2 Ultra to play back up to 22 streams of 8K ProRes 422 video — far more than any PC chip can do."
Would need an expensive GPU or hardware accelerator to match that capability on the Intel side. The companies that use these machines probably don't even consider the $500 cost difference versus a Dell Precision workstation or whatever.
Still very, very nice chips, but "only competitor" is far removed from reality. Also, mac mini with m2 pro, 32gb of ram and 1TB of storage is 2500€; the above mentioned amd one is 770€.
So "better at just 3-4 times the price"... color me highly unimpressed.
Are you sure that you're up to speed with the latest developments? The M1 was a pretty good wake-up call, yes, but it's not like AMD are Intel are standing still, there are now very nice chips available from all manufacturers.
I’m really not that impressed by geekbench because JavaScript ran fast enough on the m1’s. I feel like the m1’s are actually fast enough for basic web browsing. I don’t feel the emphasis put on it makes sense.
What I’d love to see is more memory bandwidth but memory speeds have either stagnated for 3 years or regressed. Especially now that Apple is marketing this as a good computer for “transformers”. To me the interesting part of Apple Silicon’s is it’s ability to act as a high-vram low-compute low-bandwidth GPU given the cost.
The one thing I was very impressed by was the m3 max having a 128gb memory limit. I did not expect that. The 128gb spec of the m3 max is very tempting as it feels like a very capable do-it-all computer.
> I’m really not that impressed by geekbench because JavaScript ran fast enough on the m1’s. I feel like the m1’s are actually fast enough for basic web browsing. I don’t feel the emphasis put on it makes sense.
Do you think Geekbench is specifically a JavaScript or browser benchmark, or are you just saying that browser performance is all you care about for laptops and that you don't see any benefit from further performance increases for your use case?
No idea where this idea has formed that the M-series competes tit-for-tat with x86 chips, let alone outcompetes them. It beats some models in Geekbench scores and some 3D benchmarks (for ones with iGPUs), sure; but multispectrum benchmarks have shown it's still generally outcompeted by most x86 chips targeted at the same market, drastically in some cases:
Apple's claims have always been that it performs better per watt, which is true; and that it's more efficient for quite a few common use cases (JavaScript execution, compositing, etc), especially when battery life is a concern.
There really isn't much of a perf difference between X86 and ARM these days.
Generally CISC vs RISC argument this would devolve into but they all have instruction sets that either optimize performance or introduce ways to handle complex or reduced instructions in their own way.
RISC has only the benefit of being simple out of the box if designed that way and can have a lower power draw simply because. CISC is complex out of the box so it has more draw out of the box.
> There really isn't much of a perf difference between X86 and ARM these days.
We're not talking ISAs here or any sort of ideological argument regarding instruction set design. We're talking concrete architectural implementations.
I linked specific benchmarks that show very real differences in a wide set of benchmarks. Particularly more generalized ones than the use cases the M-series is optimized for.
If we take both companies' words to heart that their architectures recieved a 20% uplift, it would carry through that the gap would be similar today. Especially with the Ryzen 8000 around the corner, which is claimed/expected to have an even stronger single-core IPC uplift than 6000->7000.
5XXX, there was no 6XXX series, it was a rebrand of mixed Zen+ Zen 2 and Zen 3 cored CPUs. Bandwidth is the only real issue with AMD CPUs in comparison to Intel, and typically an advantage of Apple vs AMD though. Chiplet disadvantages vs monolithic design, much easier to perfect binning, much harder to get equal cross-core utilization.
I got the M1 MBP a few years ago and it was about twice as fast as my reasonable desktop at the time on simple things like compiling code. And it was silent. It was the silence that wowed me. Didn’t want to turn my desktop on again after that.
When my colleagues got their M1s, they were about as fast as the then 6 years old 4 core i5 I was using at the time.
That is somewhat impressive, but I find it hard to believe that a reasonable desktop was only half as fast as an M1.
I'm thinking if that difference was due to Windows' slow file system rather than the machine itself. For instance there's is a large performance delta between Windows and Linux when compiling JavaScript.
Stop using Geekbench to "prove" how fast your CPU is. It's one aspect of benchmarking, not the totality. It doesn't cover the vast majority of
generic benchmarks. It covers "common use case" ones, which is exactly what the Apple engineers target.
That's not a bad thing, but as mentioned in the OP there are over a hundred things it sucks at, and a few dozen more it barely keeps up at. Even if it's awesome at executing JS code.
Benchmarks (or any timed CPU performance tests) are precisely how we prove how fast CPUs are. No need for quotes on "prove". It's an objective measure, an actual proof. Geekbench 6 is one suite of tests that is reliable and has provided a useful measure for comparing production server VM performance at my job. It accurately predicted the performance of real server work loads.
I would rather use Geekbench or any random benchmarking suite than rely on some guy commenting that he thinks/feels his ancient quad i5 is as fast as the M1 (not even close).
> It's an objective measure, an actual proof. Geekbench 6 is one suite of tests that is reliable
Did I say it wasn't useful? I said it's only useful if you're trying to measure a single set of metrics versus a broad spectrum. As is clear from the 100+ tests it doesn't compare to a generation old Ryzen architecture at. I also clearly said that it was a useful metric for what many M-series users are actually looking for ("common use cases such as JavaScript execution and visual compositing").
You should really learn to comprehend entire thoughts versus the idea/sentence/strawman you make up in your head.
> You should really learn to comprehend entire thoughts
Both of these replies from you are poorly written, hence you're getting downvoted by others. You should learn to form clearer thoughts that are worth communicating, and work on your writing.
I'm being downvoted for being a smarmy ass. I accept that.
Now feel free to introspectively accept your own faults. Particularly your pigheaded inability to argue in good faith and/or comprehend basic English.
PS - you didn't need two messages for that. All it does is reinforce your insecurity and necessity to "be right".
Next time, just edit your post. Or better yet, post and move on. It's an internet comment from an anonymous person, no need to let it live rent free in your head.
But understood, one favors one party so is invalid and the other favors your weird cult, so it's acceptable. I'll keep that in mind in the future for internet "debate": "if the benchmark doesn't favor Apple, it's invalid for <INSERT-SOME-REASON>".
You can move the goalposts or change the argument all you like. I don't really care who wins/loses an internet argument, and I'm quite well versed on both architectures to be able to acknowledge the strengths and weaknesses of each. Which I fully acknowledged multiple times, despite your inability to parse the text of that information.
> Which is a part of the x86 instruction set. As equally as with Apple's own simd instructions:
Yes M1 does not have AVX. That is x86 only. So if the application does not use M1 SIMD then it won't run as quickly. Benchmark results were obvious in many cases. Apple SIMD is simply not 10x slower than x86 SIMD. That is obvious.
Note that the Apple Accelerate API that you linked to is available on MacOS only. The benchmarks were run on Asahi Linux.
> favors your weird cult
LOL. I own an AMD Zen 3 desktop, Intel i5 Surface Pro, Apple M1 laptop. I develop on MacOS, Windows, Linux. Which cult does that put me in? Am I in all three cults?
> As is clear from the 100+ tests it doesn't compare to a generation old Ryzen architecture
I checked the Phoronix test results. Looks like AMD is ahead specifically in DSP / linear algebra tests where the code was designed for AVX instead of NEON. These are massively accelerated by specialized vector processing instructions. The results would be quite different if the code was optimized for both instruction sets.
> It's an objective measure, an actual proof. Geekbench
> Geekbench 6 is one suite of tests that is reliable and has provided a useful measure for comparing production server
Geekbench is notoriously the least-useful of the lot. The big flashy numbers you share from that test are an aggregate score of almost nothing meaningful or similar being compared. Geekbench is less about hardware analysis and more about offhand comparisons to other systems. Using it to compare ARM and x86 is useless, given that both architectures are more efficient at different workloads. Worse yet, we then "average" those scores to try and give people a realistic performance impression based off wildly different CPU architectures.
In some tests, it is entirely probable that an i5 would smoke an M1. Load up an SIMD benchmark and compare AVX2 to ARM NEON. Even a Haswell CPU will come out on top if you stack the benches right.
> ... In some tests, it is entirely probable that an i5 would smoke an M1
I'm aware that there is no such thing as a perfect benchmark. Geekbench has been useful for me and is good enough to get a general idea of performance.
Check it yourself. Compare i5-4690k (fastest i5 from 2014) with my M1 Pro.
M1 is faster in every single-core test in Geekbench 6. Overall 195% for single core and 427% for multi-core.
> an aggregate score of almost nothing meaningful or similar being compared
The dissimilarity between tests is intentional, by design, and is precisely what makes it useful as a general benchmark. The tests seem to cover a range of various real world use cases (C compilation, ray tracing, file compression, image processing, web browsing, etc). As I mentioned earlier, it usefully predicted real server performance at my job (JVM server doing financial batch processing).
> The dissimilarity between tests is intentional, by design, and is precisely what makes it useful as a general benchmark.
That's fine, if those numbers mean something to you then don't let me take that from you. My point is more that Geekbench has always been about cooking the books and avoiding objective comparisons. A number of their tests have been shown to instead rely on memory and SSD bandwidth, which kinda makes the whole "CPU score" concept bunk.
Most benchmarks will indeed rank the 2021 RISC CPU above the 2014 CISC CPU. My point is more that Geekbench has always been a dubious point of comparison for hardware differences. The meaning behind those scores is pretty pointless when you actually see how they're weighed and aggregated.
> A number of their tests have been shown to instead rely on memory and SSD bandwidth
These are simulating common CPU workloads, so of course they will be partly constrained by the processors memory controller bandwidth and latency, as in real workloads. The fact that this is the case makes it even more useful as a general processing benchmark. Modern processors include a memory controller in the package. What you see as "cooking the books" is actually the correct and expected approach for this purpose. It should rely on memory bandwidth for some of the tests. The entire Geekbench data set is ~1GB so would be cached by the OS in RAM on any of systems we care about benchmarking.
If you need a specialized benchmark then run a specialized benchmark. Geekbench is not pretending to be something like a LINPACK benchmark.
Compiling code is something that the M1 excelled at. My guess that a lot of that was excellent memory performance compared to what was normal in desktops at the time.
I don't understand "compiling code" as a metric. Do you mean Swift and Objective-C? Or do you mean general things like Rust and C++?
Based on multiple benchmarks, the Ryzen 6000 series and Alder Lake series should outperform the M2 Ultra pretty drastically (200-300%, which is well above the M3 general uplift of 20%) unless there's a severe disk bottleneck, which is going to be more of a SSD/PCIe issue than a CPU one (and something that can be configured/improved on PCs).
I was talking about when the first generation M1 came out and it was faster than normal contemporary desktops at developer stuff like compiling Java and c/c++ and objectiveC (was this before iOS went swift? I can’t remember). Lots of devs got one and praised it to bits on HN. And my builds were basically twice as fast.
These days, due to various constraints with tooling my current company still uses, I’m on a new x86 MBP, and it feels slower and hotter and noisier. But it’s not a desktop.
> Based on multiple benchmarks, the Ryzen 6000 series and Alder Lake series should outperform the M2 Ultra pretty drastically (200-300%, which is well above the M3 general uplift of 20%) unless there's a severe disk bottleneck, which is going to be more of a SSD/PCIe issue than a CPU one (and something that can be configured/improved on PCs).
Compilation is strongly memory bound, especially large builds. Rust and Haskell compilers regularly clock a upward of 70 Gb/sec of the memory throughput on the parallel builds of some small to most medium size code bases. Large C++ builds should be the same. Synthetic benchmarks are a poor reflection of such workloads.
Having said that, it does not matter how fast is the CPU if it stalls on the memory access, if e.g. the memory bus is not wide enough. Max's 512-bit wide memory bus at 400Gb/sec and Ultra's 1024-bit memory bus at 800Gb/sec can feed the L2 cache much faster resulting in a more even CPU core utilisation and the workload spread.
To add to the confusion, manufacturer quoted TDP has been kinda nonsense for years, too; many Intel chips in particular will happily go significantly over their alleged TDP for fairly normal workloads.
I would love to upgrade my personal intel MacBook.
Unfortunately (mebe) I found a killer deal on a 5950x last year and now I’ve got a beast of a workstation with 128gb ram.
I have actually found it hard to use my laptop anymore. Work laptop is a M1 with 64gb ram. It’s fast. But my workstation feels a lot more powerful. It’s hard to use either laptops and dealing with spinners and waiting.
I have been tracking 5950x prices and anything over $300 seems unreasonable.
And I think I'm just going to get a 7700x and a new MB and 32GB of RAM,
it should be a $500-sh upgrade. It should have M3 like single threaded perf,
multi threaded perf better than 5950x, slightly worse than M3 Max, better memory bandwidth than the 5950x.
You're right about just having a beefier PC, but maybe a baseline M2/M3 Air to SSH into a local machine + a beefier PC is a good combination. Luckly I switched to an all in terminal dev experience with Neovim/Helix/Tmux and the advantage is that one can seamlessly mix ssh/mosh shells in this workflow. Don't even need to set up any VNC/RDPs/Moonlights. Maybe even that is not needed if your languages supports ccache like tools.
I'm mostly interested in compilation speed, maybe running some programs faster on beefier hardware and I don't mind working on Linux.
So to recap, a baseline Air for portability(or some light ryzen laptop) plus a medium range PC seems like a winning combination for me. Given you already have a case/psu/ssd together they can cost as an entry level m3 macbook pro with 8GB of RAM.
wtf, you’re comparing 800W with 20W machine. No shit your workstation is more powerful, it’s a fucking workstation not a battery operated laptop.
If you are troubled by loading on 64GB M1 Max, then congrats, you are doing workload that no laptop will handle.
> If you are troubled by loading on 64GB M1 Max, then congrats, you are doing workload that no laptop will handle.
I can easily buy a laptop with 128GB RAM, eats more energy (100W+) and is more powerful. It's just that it does not have an Apple logo on it. Does it not count as a laptop?
In the literal sense? No. I don’t feel like those kinds of laptops are comfortable on my lap.
To be a bit more serious, yeah sure they are laptops, but you tend to sacrifice portability, battery life and comfort. Which for most people are pretty important metrics for a laptop. “Luggable” is a term sometimes used about this class of products.
I’m sure the poster was being hyperbolic. Of course you can find laptops with more RAM and some use cases that 64GB wouldn’t cover. But for most common tasks it’s more than enough.
What are you even doing? I have refurbished workstation from 2016 (xeon with 32gig ram for 200usd) and the cheapest M1 air. And i am sitting here thinking we went past of what people need for personal/work computer because i can easily do my work on it and would get nothing from faster machines. Yes i dont compile big codebases (where i assume it done outside of my system anyway) but i just wonder what are the HW companies going to do to keep people upgrading their machines.
We'll find a way to use up what they can provide :-)
With 64GB ram and our infrastructure defined with NixOS, I find myself building and spinning it up locally (in test mode) as part of working on it. Having ~10 VMs running locally with 2-3 GB of ram each quickly adds up.
It feels a little dirty, but it's really nice to have the possibility. :-)
I switched from an i7 1065g7, 32GB RAM to an M2 Pro (10 CPU cores), 32GB RAM.
It makes a huge difference to me.
Android Studio and Gradle builds are running a lot faster now.
Before about 2-3 minutes, now less than a minute.
My company builds an Android SDK for payments.
For me the end is not reached and I'm looking forward to an upgrade in 3-5 years.
Yes, I have an M1 Macbook Pro that I rarely use as I have a similar AMD-powered desktop machine as yours, and I'm usually at my desk anyway when working, so I simply don't use the laptop unless I need to travel.
I'm looking to upgrade to the Threadrippers [0] next year however and get even more of a speed advantage.
Do you need that amount of memory on your laptop? My work M1 laptop has only 16gb of ram. If I need more ram I have access to many servers and use my laptop as a SSH console, a jupyter, or a VSCode client.
The servers have up to 6TB of ram each, but I mostly use one with 512GB.
When the M series laptops came out I thought that it made sense to be economical with memory, but LLM's became huge a few years after release and Apple Silicon happens to be local inferencing beasts that can use all the memory you can throw at it.
At release I would have said most people should stick with as little memory as they could get away with due to Apple's pricing and the non-upgradability. Nowadays, I'd either recommend the M3 or M3 Pro and keeping things cheaper, but if you're going to get the Max at all, I sort of feel like you may as well go for the 16c cpu/40c GPU/128gb option at $1500 USD more and play around with stuff like Laama. It will become obsolete in not too long due to memory tech marching forwards, but it's still going to be plenty capable of dabbling with this stuff for awhile. The M3 Max has the memory bandwidth of a 4 dimm DDR5-6400 server which is pretty impressive given the pricing.
Beyond that, I see people speccing high memory for things like VFX and Macs are pretty popular choices for those doing video work.
I have a workstation with a 3950x and 128gigs ram that I mostly use for fluid sims in Houdini and rendering tasks. There's honestly nothing that I'm doing on my work machine (an M2 macbook pro) that I feel would benefit from more cores or more RAM - really compute-intensive stuff gets offloaded to a server anyways.
The only real difference I'm noticing is that emacs for some reason is just slower on osx than on linux, but that's not down to the raw performance, as even my 7 old Thinkpad with a 4-core i5 makes emacs feel way snappier than my macbook.
I have a 5950x when 128GB of RAM, and an M1 Max. I was performance testing some Clojure code a few days ago (pure number crunching in parallel), and I was shocked to see it running faster in every test case on my M1 Mac.
I think it might have more to do with memory bandwidth and thermals than the processing speed and cores.
I travel between three locations, just with my phone and NVMe disk. All three locations have identical setup: monitors, workstation, desk, chair... Computer for me is a just docking station for a harddrive.
I really hated dealing with USB dongles, power suspend and all that stuff.
If I have to work on the go, I have bluetooth keyboard for my phone.
I'm in the other boat, but it's also not great for Apple. My min-spec 2021 14" MBP is so much overkill for the stuff I'm doing (C/C++ coding) that I probably won't need a faster one for a decade or so ;)
Apple silicon is either totally silent, or totally silent/very quiet at idle (depending on settings) and not that loud at load. Apple is using the most power efficient chips on the planet and exhaustively optimised cooling solutions that use the aluminium case as a heatsink.
I've tried to make an ATX gaming build silent, it never worked out satisfactorily after a lot of time and expense. I've done the silent gaming build and Apple Silicon and I cannot recommend Apple silicon enough. I spent many hours researching, building, and tweaking a computer buying premium components and at the end of the day the computer ended up being louder and less performant than the Mac I took out of the box and plugged into the wall. Even if I built that silent gaming PC perfectly, which is a big "if", the two big problems are that ATX gaming PCs are poorly designed for the passive cooling performance of the CPU/GPU/PSU, and that the lower performance/watt means that your gaming PC will likely have more coil whine for the same level of performance.
Three is that the GOOD silent Intel/AMD machines are NOT gaming PCs, they're low powered PCs sold on aliexpress with a custom case that acts as a heatsink that are much cheaper than anything Apple sells.
In terms of less expansion:
- No expandable storage sucks, but it's also not a HUGE deal when 1TB is realistically going to be enough local storage for 90% of devs for the lifespan of the computer.
- No expandable memory sucks, but you know what else sucks? Gaming desktops using dual channel memory running at maybe 100gb/s with the GPU choked off from accessing it by the PCIe bus. The price you pay to get a pooled 128gb memory the GPU/CPU can use at 400gb/s bandwidth from Apple is actually insanely low compared to the competition.
- For ports, Apple is running three TB4 connectors and an HDMI 2.1 port with 168gb/s combined bandwidth with Wifi 6E/BT 5.3. Sure having lots of USB 3.0 ports is cool, but what else is cool is being able to have multiple laptops and setting up a desk which allows any of your laptops to connect to every peripheral via a single cable.
- Finally, Apple Silicon holds its resale value, so selling your current model and buying a different one if needed is really not THAT big of a deal.
Nah... Work laptop is a M2 Pro, 32GB ram. I need the form factor because I travel internationally with it, but it feels as fast as or faster than any other machine I have ever used...
Powerful how? M1 Max has like 8x the memory bandwidth of a 5950x running 4 sticks of memory. 5950x has more compute, and in practice compute heavy tasks are more optimised for the platform, but I don't actually find myself exploiting that all that often since I'm not a Linux Kernel compilation enthusiast.
M3 Max is better than 5950x maybe 90-95% of the time and its shored up its weaknesses in raw compute horsepower. That being said, I'm getting more interested in a workstation myself because the M-series has stagnated in terms of memory bandwidth and a couple of used 3090's are pretty cheap.
Like burgers, make it wider, not taller. More ram not faster cpus. I dream ofna terabyte ram accessible to both gpu and cpu. I promise I will do more than just keep a few more tabs open with it.
104 comments
[ 3.1 ms ] story [ 120 ms ] threadI remember reading "why ARM design will rule" (simpler design) and we're seeing it in the flesh.
I keep hoping to see another manufacturer, e.g. Lenovo making their own ARM machine for thinkpads... It would be incredibly popular I feel. But most likely will never happen.
ARM is not some magic trick
https://chipsandcheese.com/2021/07/13/arm-or-x86-isa-doesnt-...
Their chips are good, no doubt, but AMD and Intel are not sleeping at the wheel. Even if Apple chips were available as CPUs and Motherboards with full Windows and Linux support, they wouldn't be the obvious choice for all use cases. The M2 Ultra is incredible, but there are workloads where it stands no chance against a threadripper.
However, despite all the marketing it still remains to be seen whether they can transition into desktop/workstation chips with these. Especially at a reasonable price.
But the point is that all of these have a market and some people are spending big on this. I talked to one of my friends recently who runs a high tech printing company and they spent some money on a mac studio recently for doing graphics work. Worth the money according to him.
I have a 14" macbook pro (M1). It's great. I do some flight simuluation on it occasionally with X-plane 12. That runs pretty smooth on this thing and it's one of the few things I have that actually gets this laptop warm.
Apple did a great job of getting great performance with modest power budgets. Comparing their products to water cooled monstrosities is not really that interesting.
They released the M1 Ultra last year. Desktop only.
M2 Ultra is also desktop only.
The M3 Ultra is not released yet but will also be a desktop chip.
https://www.apple.com/mac-pro/
Either way, as soon as power consumption does not matter, price-performance (well, even both price _or_ performance, separately) wise M-series is not competitive with top x86 CPUs.
https://browser.geekbench.com/v6/cpu/compare/3353019?baselin...
Also noticed the M2 is faster than Intel on the Clang benchmark. (The Clang workload uses the Clang compiler to compile the Lua interpreter - https://www.geekbench.com/doc/geekbench6-cpu-workloads.pdf).
One thing to note regarding the M2 Ultra (from Apple marketing):
"It has dedicated, hardware-enabled H. 264, HEVC, and ProRes encode and decode, allowing M2 Ultra to play back up to 22 streams of 8K ProRes 422 video — far more than any PC chip can do."
Would need an expensive GPU or hardware accelerator to match that capability on the Intel side. The companies that use these machines probably don't even consider the $500 cost difference versus a Dell Precision workstation or whatever.
https://www.notebookcheck.net/AMD-Ryzen-9-7940HS-analysis-Ze...
Still very, very nice chips, but "only competitor" is far removed from reality. Also, mac mini with m2 pro, 32gb of ram and 1TB of storage is 2500€; the above mentioned amd one is 770€. So "better at just 3-4 times the price"... color me highly unimpressed.
https://browser.geekbench.com/v6/cpu/multicore
What I’d love to see is more memory bandwidth but memory speeds have either stagnated for 3 years or regressed. Especially now that Apple is marketing this as a good computer for “transformers”. To me the interesting part of Apple Silicon’s is it’s ability to act as a high-vram low-compute low-bandwidth GPU given the cost.
The one thing I was very impressed by was the m3 max having a 128gb memory limit. I did not expect that. The 128gb spec of the m3 max is very tempting as it feels like a very capable do-it-all computer.
Do you think Geekbench is specifically a JavaScript or browser benchmark, or are you just saying that browser performance is all you care about for laptops and that you don't see any benefit from further performance increases for your use case?
I was not expecting this type of comment on 2023 on a discussion on premium hardware marketed as the highest performant hardware on the market.
highest performant for what?
According to Apple, their M3 products outperform everything under the sun.
As a reminder, here's a link to Apple's press release.
https://www.apple.com/newsroom/2023/10/apple-unveils-m3-m3-p...
This means that the 8k series with twice the operation width should actually be faster than the M3 Max.
https://browser.geekbench.com/search?utf8=%E2%9C%93&q=7945HX...
My 7840u isn't too far behind on single core perf.
https://www.phoronix.com/review/apple-m2-amd-ryzen
Apple's claims have always been that it performs better per watt, which is true; and that it's more efficient for quite a few common use cases (JavaScript execution, compositing, etc), especially when battery life is a concern.
Generally CISC vs RISC argument this would devolve into but they all have instruction sets that either optimize performance or introduce ways to handle complex or reduced instructions in their own way.
RISC has only the benefit of being simple out of the box if designed that way and can have a lower power draw simply because. CISC is complex out of the box so it has more draw out of the box.
We're not talking ISAs here or any sort of ideological argument regarding instruction set design. We're talking concrete architectural implementations.
I linked specific benchmarks that show very real differences in a wide set of benchmarks. Particularly more generalized ones than the use cases the M-series is optimized for.
If we take both companies' words to heart that their architectures recieved a 20% uplift, it would carry through that the gap would be similar today. Especially with the Ryzen 8000 around the corner, which is claimed/expected to have an even stronger single-core IPC uplift than 6000->7000.
Because depending on your use case it does.
If you're doing anything with media the dedicated codec silicon and Neural Engine makes a huge difference.
Not sure what the point of restating that is.
That is somewhat impressive, but I find it hard to believe that a reasonable desktop was only half as fast as an M1.
I'm thinking if that difference was due to Windows' slow file system rather than the machine itself. For instance there's is a large performance delta between Windows and Linux when compiling JavaScript.
No they were faster than your old quad core i5.
My M1 Pro laptop has a Geekbench 6 score of 12,626.
My Ryzen 5600X desktop (6 cores, faster than a quad i5) has a Geekbench score of 8,583.
That's not a bad thing, but as mentioned in the OP there are over a hundred things it sucks at, and a few dozen more it barely keeps up at. Even if it's awesome at executing JS code.
I would rather use Geekbench or any random benchmarking suite than rely on some guy commenting that he thinks/feels his ancient quad i5 is as fast as the M1 (not even close).
Did I say it wasn't useful? I said it's only useful if you're trying to measure a single set of metrics versus a broad spectrum. As is clear from the 100+ tests it doesn't compare to a generation old Ryzen architecture at. I also clearly said that it was a useful metric for what many M-series users are actually looking for ("common use cases such as JavaScript execution and visual compositing").
You should really learn to comprehend entire thoughts versus the idea/sentence/strawman you make up in your head.
Both of these replies from you are poorly written, hence you're getting downvoted by others. You should learn to form clearer thoughts that are worth communicating, and work on your writing.
Now feel free to introspectively accept your own faults. Particularly your pigheaded inability to argue in good faith and/or comprehend basic English.
PS - you didn't need two messages for that. All it does is reinforce your insecurity and necessity to "be right".
Next time, just edit your post. Or better yet, post and move on. It's an internet comment from an anonymous person, no need to let it live rent free in your head.
> PS - you didn't need two messages for that.
I only wrote one message for that.
Anyway, put that aside. The benchmark suite at Phoronix is biased to AMD due to a lot of tests being optimized for x86 instructions (AVX).
Sick burn, bro.
> I only wrote one message for that.
You returned to write this later:
https://news.ycombinator.com/item?id=38121689
Because a stupid internet argument was living rent free in your head.
> Anyway, put that aside. The benchmark suite at Phoronix is biased to AMD due to a lot of tests being optimized for x86 instructions (AVX).
Which is a part of the x86 instruction set. As equally as with Apple's own simd instructions:
https://developer.apple.com/documentation/accelerate/simd
But understood, one favors one party so is invalid and the other favors your weird cult, so it's acceptable. I'll keep that in mind in the future for internet "debate": "if the benchmark doesn't favor Apple, it's invalid for <INSERT-SOME-REASON>".
You can move the goalposts or change the argument all you like. I don't really care who wins/loses an internet argument, and I'm quite well versed on both architectures to be able to acknowledge the strengths and weaknesses of each. Which I fully acknowledged multiple times, despite your inability to parse the text of that information.
Yes M1 does not have AVX. That is x86 only. So if the application does not use M1 SIMD then it won't run as quickly. Benchmark results were obvious in many cases. Apple SIMD is simply not 10x slower than x86 SIMD. That is obvious.
Note that the Apple Accelerate API that you linked to is available on MacOS only. The benchmarks were run on Asahi Linux.
> favors your weird cult
LOL. I own an AMD Zen 3 desktop, Intel i5 Surface Pro, Apple M1 laptop. I develop on MacOS, Windows, Linux. Which cult does that put me in? Am I in all three cults?
I checked the Phoronix test results. Looks like AMD is ahead specifically in DSP / linear algebra tests where the code was designed for AVX instead of NEON. These are massively accelerated by specialized vector processing instructions. The results would be quite different if the code was optimized for both instruction sets.
> Geekbench 6 is one suite of tests that is reliable and has provided a useful measure for comparing production server
Geekbench is notoriously the least-useful of the lot. The big flashy numbers you share from that test are an aggregate score of almost nothing meaningful or similar being compared. Geekbench is less about hardware analysis and more about offhand comparisons to other systems. Using it to compare ARM and x86 is useless, given that both architectures are more efficient at different workloads. Worse yet, we then "average" those scores to try and give people a realistic performance impression based off wildly different CPU architectures.
In some tests, it is entirely probable that an i5 would smoke an M1. Load up an SIMD benchmark and compare AVX2 to ARM NEON. Even a Haswell CPU will come out on top if you stack the benches right.
I'm aware that there is no such thing as a perfect benchmark. Geekbench has been useful for me and is good enough to get a general idea of performance.
Check it yourself. Compare i5-4690k (fastest i5 from 2014) with my M1 Pro.
https://browser.geekbench.com/v6/cpu/compare/3356017?baselin...
M1 is faster in every single-core test in Geekbench 6. Overall 195% for single core and 427% for multi-core.
> an aggregate score of almost nothing meaningful or similar being compared
The dissimilarity between tests is intentional, by design, and is precisely what makes it useful as a general benchmark. The tests seem to cover a range of various real world use cases (C compilation, ray tracing, file compression, image processing, web browsing, etc). As I mentioned earlier, it usefully predicted real server performance at my job (JVM server doing financial batch processing).
That's fine, if those numbers mean something to you then don't let me take that from you. My point is more that Geekbench has always been about cooking the books and avoiding objective comparisons. A number of their tests have been shown to instead rely on memory and SSD bandwidth, which kinda makes the whole "CPU score" concept bunk.
Most benchmarks will indeed rank the 2021 RISC CPU above the 2014 CISC CPU. My point is more that Geekbench has always been a dubious point of comparison for hardware differences. The meaning behind those scores is pretty pointless when you actually see how they're weighed and aggregated.
These are simulating common CPU workloads, so of course they will be partly constrained by the processors memory controller bandwidth and latency, as in real workloads. The fact that this is the case makes it even more useful as a general processing benchmark. Modern processors include a memory controller in the package. What you see as "cooking the books" is actually the correct and expected approach for this purpose. It should rely on memory bandwidth for some of the tests. The entire Geekbench data set is ~1GB so would be cached by the OS in RAM on any of systems we care about benchmarking.
If you need a specialized benchmark then run a specialized benchmark. Geekbench is not pretending to be something like a LINPACK benchmark.
Performance on Linux on equivalent hardware has always been far superior (on intel macs)
Based on multiple benchmarks, the Ryzen 6000 series and Alder Lake series should outperform the M2 Ultra pretty drastically (200-300%, which is well above the M3 general uplift of 20%) unless there's a severe disk bottleneck, which is going to be more of a SSD/PCIe issue than a CPU one (and something that can be configured/improved on PCs).
These days, due to various constraints with tooling my current company still uses, I’m on a new x86 MBP, and it feels slower and hotter and noisier. But it’s not a desktop.
Compilation is strongly memory bound, especially large builds. Rust and Haskell compilers regularly clock a upward of 70 Gb/sec of the memory throughput on the parallel builds of some small to most medium size code bases. Large C++ builds should be the same. Synthetic benchmarks are a poor reflection of such workloads.
Having said that, it does not matter how fast is the CPU if it stalls on the memory access, if e.g. the memory bus is not wide enough. Max's 512-bit wide memory bus at 400Gb/sec and Ultra's 1024-bit memory bus at 800Gb/sec can feed the L2 cache much faster resulting in a more even CPU core utilisation and the workload spread.
And that may be true. My RPI2 is impressive at performance per watt, still my 5800X or even a 3900X blow it away at almost every workload.
Girlfriend's M1 side to side with any of my desktops is just noticeably slower in everything.
There just isn't an excuse for desktop processors to be as thirsty these days.
EDIT: according to CPU Monkey [2] it's 45W. All these websites don't seem so reliable...
[1] https://nanoreview.net/en/cpu/apple-m3-max
[2] https://www.cpu-monkey.com/en/compare_cpu-apple_m3_max_14_cp...
What about the graphics and neural processors integrated into that chip? How do they compare to the M3 Max?
What’s the state of the Linux drivers and AI workload acceleration for them?
Does it keep its cool under load? How will it perform in a laptop?
How does the latency and bandwidth between the memory, CPU/GPU/APU compare?
Etc
I would love to upgrade my personal intel MacBook.
Unfortunately (mebe) I found a killer deal on a 5950x last year and now I’ve got a beast of a workstation with 128gb ram.
I have actually found it hard to use my laptop anymore. Work laptop is a M1 with 64gb ram. It’s fast. But my workstation feels a lot more powerful. It’s hard to use either laptops and dealing with spinners and waiting.
Anyone else in the same boat?
You're right about just having a beefier PC, but maybe a baseline M2/M3 Air to SSH into a local machine + a beefier PC is a good combination. Luckly I switched to an all in terminal dev experience with Neovim/Helix/Tmux and the advantage is that one can seamlessly mix ssh/mosh shells in this workflow. Don't even need to set up any VNC/RDPs/Moonlights. Maybe even that is not needed if your languages supports ccache like tools.
I'm mostly interested in compilation speed, maybe running some programs faster on beefier hardware and I don't mind working on Linux.
So to recap, a baseline Air for portability(or some light ryzen laptop) plus a medium range PC seems like a winning combination for me. Given you already have a case/psu/ssd together they can cost as an entry level m3 macbook pro with 8GB of RAM.
I can easily buy a laptop with 128GB RAM, eats more energy (100W+) and is more powerful. It's just that it does not have an Apple logo on it. Does it not count as a laptop?
If it's drawing 100W then the poor battery life make it more like a semi-portable.
Also given what we know about male fertility I wouldn't have that anywhere near my lap.
In the literal sense? No. I don’t feel like those kinds of laptops are comfortable on my lap.
To be a bit more serious, yeah sure they are laptops, but you tend to sacrifice portability, battery life and comfort. Which for most people are pretty important metrics for a laptop. “Luggable” is a term sometimes used about this class of products.
I’m sure the poster was being hyperbolic. Of course you can find laptops with more RAM and some use cases that 64GB wouldn’t cover. But for most common tasks it’s more than enough.
Then in my experience, Apple did not make a pro laptop for a very long time until the M series came along.
I remember my MacBook Pro from 2009 getting hot enough to hurt my legs, let alone any from the following decade that ran hotter.
Easy, just drop software support for it and force you to buy a new one... Or switch to Linux
At the same time it looks like great thing for Linux
With 64GB ram and our infrastructure defined with NixOS, I find myself building and spinning it up locally (in test mode) as part of working on it. Having ~10 VMs running locally with 2-3 GB of ram each quickly adds up.
It feels a little dirty, but it's really nice to have the possibility. :-)
For me the end is not reached and I'm looking forward to an upgrade in 3-5 years.
Mooooaaarrrrr javascript.
I'll probably get downvoted... by people using a wrapped Chrome engine. That, friends, is true irony.
I'm looking to upgrade to the Threadrippers [0] next year however and get even more of a speed advantage.
[0] https://news.ycombinator.com/item?id=37942418
The servers have up to 6TB of ram each, but I mostly use one with 512GB.
I see that you can now order machines with 32TB of ram, like this one: https://www.supermicro.com/en/products/system/mp/6u/sys-681e...
At release I would have said most people should stick with as little memory as they could get away with due to Apple's pricing and the non-upgradability. Nowadays, I'd either recommend the M3 or M3 Pro and keeping things cheaper, but if you're going to get the Max at all, I sort of feel like you may as well go for the 16c cpu/40c GPU/128gb option at $1500 USD more and play around with stuff like Laama. It will become obsolete in not too long due to memory tech marching forwards, but it's still going to be plenty capable of dabbling with this stuff for awhile. The M3 Max has the memory bandwidth of a 4 dimm DDR5-6400 server which is pretty impressive given the pricing.
Beyond that, I see people speccing high memory for things like VFX and Macs are pretty popular choices for those doing video work.
The only real difference I'm noticing is that emacs for some reason is just slower on osx than on linux, but that's not down to the raw performance, as even my 7 old Thinkpad with a 4-core i5 makes emacs feel way snappier than my macbook.
brew install emacs-plus@29 --with-native-comp
I have a 5950x when 128GB of RAM, and an M1 Max. I was performance testing some Clojure code a few days ago (pure number crunching in parallel), and I was shocked to see it running faster in every test case on my M1 Mac.
I think it might have more to do with memory bandwidth and thermals than the processing speed and cores.
I travel between three locations, just with my phone and NVMe disk. All three locations have identical setup: monitors, workstation, desk, chair... Computer for me is a just docking station for a harddrive.
I really hated dealing with USB dongles, power suspend and all that stuff.
If I have to work on the go, I have bluetooth keyboard for my phone.
It works quite well, even for AMD CPUs.
I happen to have the same CPU as you and been running my Hackintosh setup smoothly for 1.5 years.
You can see my setup and full open core configuration here;
https://github.com/reimertz/hackintosh-x570i-5950x-6800xt
The only caveat is intel-only cpu instructions that might cause certain apps to not work. The adobe suite is one example.
As someone with a gaming computer, I never understood devs that use a laptop...
its way slower, waaaaaaay worse cooling = more noise and less expansion possible in terms of RAM, storage and ports.
I've tried to make an ATX gaming build silent, it never worked out satisfactorily after a lot of time and expense. I've done the silent gaming build and Apple Silicon and I cannot recommend Apple silicon enough. I spent many hours researching, building, and tweaking a computer buying premium components and at the end of the day the computer ended up being louder and less performant than the Mac I took out of the box and plugged into the wall. Even if I built that silent gaming PC perfectly, which is a big "if", the two big problems are that ATX gaming PCs are poorly designed for the passive cooling performance of the CPU/GPU/PSU, and that the lower performance/watt means that your gaming PC will likely have more coil whine for the same level of performance.
Three is that the GOOD silent Intel/AMD machines are NOT gaming PCs, they're low powered PCs sold on aliexpress with a custom case that acts as a heatsink that are much cheaper than anything Apple sells.
In terms of less expansion:
- No expandable storage sucks, but it's also not a HUGE deal when 1TB is realistically going to be enough local storage for 90% of devs for the lifespan of the computer.
- No expandable memory sucks, but you know what else sucks? Gaming desktops using dual channel memory running at maybe 100gb/s with the GPU choked off from accessing it by the PCIe bus. The price you pay to get a pooled 128gb memory the GPU/CPU can use at 400gb/s bandwidth from Apple is actually insanely low compared to the competition.
- For ports, Apple is running three TB4 connectors and an HDMI 2.1 port with 168gb/s combined bandwidth with Wifi 6E/BT 5.3. Sure having lots of USB 3.0 ports is cool, but what else is cool is being able to have multiple laptops and setting up a desk which allows any of your laptops to connect to every peripheral via a single cable.
- Finally, Apple Silicon holds its resale value, so selling your current model and buying a different one if needed is really not THAT big of a deal.
M3 Max is better than 5950x maybe 90-95% of the time and its shored up its weaknesses in raw compute horsepower. That being said, I'm getting more interested in a workstation myself because the M-series has stagnated in terms of memory bandwidth and a couple of used 3090's are pretty cheap.
Like burgers, make it wider, not taller. More ram not faster cpus. I dream ofna terabyte ram accessible to both gpu and cpu. I promise I will do more than just keep a few more tabs open with it.
More discussion over here: https://news.ycombinator.com/item?id=38108235