Probably because Apple recently announced their Private Cloud Compute based on their own hardware that is being used to handle any Apple Intelligence features that don’t run on device.
Normally dual use means it has a possible use as a weapon. Compute power is restricted from export to enemy nations and is considered dual use. There was once a moment when a Mac had enough computing power that it was prohibited from export during a few months until the new limit was established, which was cleverly used as an ad: https://youtu.be/l2ThMmgQdpE?si=xgLsybj86ErWnGvR
I don't think that they're referring to this specific phrase. I had no idea that it has this specific meaning, and I would guess most consumers don't, either. If it was clickbait, it was very bad as clickbait.
The trick Apple plays with their laptops is to use a wide memory bus, which would traditionally require a lot of memory slots (at least one per channel), which usually has a poor cost/benefit ratio on consumer systems that are usually not memory bandwidth bound. But Apple wants to use the same memory for the GPU, which needs more bandwidth, so instead they solder it as if this is necessary (rather than just cheaper).
Servers normally have a lot of memory slots, e.g. Epyc servers have more memory bandwidth than the top end Apple Silicon entirely because the servers have tons of memory channels.
But they also correspondingly don't have any higher memory bandwidth than an ordinary PC. Obviously the "server" chips are going to want more than that. (Apple still solder the memory on the base model laptops though.)
They use a massively stripped down and modified version of iOS for their "Private Cloud Compute". https://security.apple.com/blog/private-cloud-compute/ See also the WWDC 2024 Platforms State of the Union video. At least some of the actual AI/server software is using Swift on Server.
Seems like for an AI data center you’d really want a dedicated AI chip, no? The current M chips have way more CPU and GPU than inference workloads need or want, right?
Still a waste of CPU. Seems hard to believe that workloads would saturate both GPU and NPU. It’s got to be sub-optimal. I’m just not sure by how much. My ill informed spidey sense says by quite a bit. But I don’t know which is why I asked!
Unless your RAM requirements aren’t being met, the M1 is good enough. The M2 is the best of the line so far because the M3 handicapped memory bandwidth for the only use cases that need so much memory.
M4 is tempting and a decent time to upgrade if you’ve had an M1 since 2020 or 2021, but its exact specifications aren’t know for all lines yet.
On CPU/GPU benchmarks the M3 comes out ahead of the M2 (all flavors in like-for-like comparisons, plain, pro, max). However the M3 Pro and some M3 Max's have lower memory bandwidth than their M2 counterparts.
So a plain M3 is better than a plain M2. The top end M3 Max is better than the M2 Max. For the rest, you have to do a comparison and determine based on your intended use-cases, but for most people the M3 {Pro,Max} will be better than the M2 equivalent.
M3 Pro (and some lower-core count M3 Max) parts have lower memory bandwidth than their M2 comparable. If your workload is highly dependent on memory bandwidth, then M2->M3 on Pro or Max might not be an improvement.
Same here... It's still amazing, and I have no legitimate reason to upgrade. Could be looked at as a failure of some sort, that they haven't given us a good reason.
Going from the M1 to M4 is barely worth the upgrade, is it really such a big deal it's been 4 months since the last MacBook Air SoC upgrade and it'll be another ~4 months until the next round releases on its usual schedule anyways? It's not like the iPad stole away time from the MacBook cadence by releasing first.
N3B is such a poorly-yielding node that it’s worth actively pushing up the release cycles and get people onto the next product, which will use N3E instead.
While I agree that it doesn’t make sense to upgrade from M1 to M4 in many cases, imagine you want to purchase a two Mac Studios that are $4k-$5k each, because you need better monitor support and i/o, but you can only purchase a really old M2 model and all the Laptops are M3 perform better but lack some i/o and the iPad is M4.
It’s super nice that they overengineer everything and that’s why I buy Apple products to be honest, but they should roll out their chips in a sensible way and not leave their most expenses products at the tail end to upgrade right before their release their next chip.
Plus I think that it makes sense to wait for the M4, because it will have enough TOPs to support their future AI endeavors and will have a much longer useful life than purchasing and M2 studio today.
Are you telling me that thinner isn't better? What else could you possibly want? Native pdf editors? Better and native applications to write, take notes, draw, and edit? Being able to draw/markup an overlay on the screen regardless of the app? Ability to ssh and use the iPad like a netbook? I think you're asking too much. It was a real great achievement that they made it thinner. Real creative thinking and this provides the most utility to users! The pencil is there for aesthetics anyways. And that's the biggest crime with the new iPad. That the pencil doesn't look different, so no one can tell that I'm using the new one from a distance. Big mistake Apple. Get your priorities right. It's not about making better technology, it is about status and getting rich. Obviously it is impossible to both make better technology and get rich at the same time, and we know which is more important.
>>> Ability to ssh and use the iPad like a netbook
Note the order of the words and the comparison to a netbook. They did not ask about "ssh TO an iPad" (also, no need to shout, you can always use *to* to get to). In the comparison to a netbook, the use-case is ssh'ing from the device to other devices. Netbooks being a category of laptops that were small, generally underpowered, but typically with good battery life, great for mobile access to remote servers and workstations.
> Netbooks being a category of laptops that were small, generally underpowered, but typically with good battery life, great for mobile access to remote servers and workstations.
I had netbook in ~2009 or something, it was with the Linux preinstalled (Ubuntu as I remember) and definitely had ability to log in into it.
I absolutely associate netbook as a device to which I can log to.
I meant from, but I do think it should work bidirectionally. Truthfully my comment was about using the ipad as a netbook. Which a big reason Apple doesn't is because they don't want to cannibalize macbook air sales.
I haven't found the ability to use it much in the same way I use my Air. Perhaps I didn't try hard enough or don't know the right setup. But given the computational power of the iPad and the limited tasks I use my Air for (browsing the web, slide making, latex writing, minor computational tasks, etc), I see no reason why I shouldn't be able to do all this on the iPad just as easily.
I know some will say "well that's not what an iPad is for" and I just have to ask "why not?" I think there are a lot of people that want some machine where there's a detachable keyboard, can work mostly like a computer, but also can work like a tablet. When tablets were first made, they had to make compromises because of the computation and architecture limits. But those limitations don't really exist anymore.
Also that tablet functionality shouldn't be weirdly hindered. Like why can't I draw on the screen? I can do that on my computer with my mouse but not the device where they sell a $100 pencil? Shouldn't that be the "magic" where I can present anything and communicate better? What if I want to pull up a webpage and then we need to highlight or talk about things? Why do I need a ton of apps to do what one of the key points the iPad is literally marketed for? Taking notes? Who doesn't have GoodNotes or Notability? (both of which strangely lack features) That's a low fucking bar and even the new OS doesn't do a great job. But this should have been there on day 1. Seriously: give me a way to make folders and organize (mine does not let me have subfolders! If you want to go all out, allow me to connect notes to PDFs! Even better if I can link note sections to PDF sections!), and just give me a few modifiable templates (blank, lines, grids. Then let me change the two colors and spacing. You can't tell me this isn't trivial to expose to the user. We're programmers, this shit doesn't need to be hard coded, though presets are good to have). And give me an infinite scrolling sheet too. That's all I need. Other things are nice, but this core functionality is missing in the Apple native Notes and Pages (but more exists if I simply open up a PDF directly, though what do I do when I can't write in the margins?!). I don't get where the disconnect is. I often hear these same complaints from other academics and from students. It just feels like the people in charge of making decisions aren't actually using the product.
And the worst part of it all is, why isn't anyone making a different better product? It seems like Microsoft, Google, and Samsung are stuck in the same similar feedback loops.
I don't know about you all, but the reason I started programming and the main reason I still program is because I need to make things that don't yet exist. Sometimes it can be a small specific task. Maybe it is that I need something to function differently than it does because I have unique situations or tastes. It used to be that companies couldn't keep us out from modifying things. But it begs the question as to why keep us out in the first place? Especially Apple. To market themselves as the brand of creativity, yet there is only "The Apple Way." You'd think that those that grew up hacking machines and seeing the power and utility in modifiability would take those lessons and build them into the core of products. I understand protecting dumb users, but you can have advanced menus. But it is that exploration and adaption that is at the core of humanity and creativity. That's why I'm passionate here. Because tech can do so much more.
the interest in cracking open apple is almost entirely spite-driven, not on the basis of freedom of computing and user freedom.
this becomes immediately obvious as soon as consoles get brought up. Not only do people not care about a mandatory 30% cut of revenue and no sideloading (except a limited capability for self-signing that isn’t allowed to utilize the standard APIs or the full performance of the chips) but they actively argue that the console is just a games appliance. And in fact that idea has been enshrined into law, game consoles get a specific carve out within the DMA just so they don’t have to!
it’s always been about spite, the things you’re saying sound nice but they’re just a convenient and temporary port-of-call for people with other objectives, and the things you outline here will be forgotten as soon as it’s convenient.
Just like RCS and whatever: like not only is Google’s proprietary encryption extensions a poison pill, but google itself doesn’t even care enough to implement rcs in its own services. Google voice has less RCS support than apple does in tyool 2024.
What goes around, comes around. When you build a platform predicated on the idea that nobody else can profit off of your product, you better be ready when a regulator tells you that business is off-limits.
This becomes immediately obvious as soon as John-Deere gets brought up. Not only do people realize that outlawing repairs should be illegal, but they actively admit that the world would be a better place if spiteful conduct was made illegal.
Especially when you consider Apple's tactics in all that. You don't have to be tech literate to think something is weird when Apple is sending lawyers and lobbyists to Nebraska to side with John Deere to make it hard for farmers to repair their equipment. If anything, I think it radicalized a lot more people because it was an explicit demonstration of how they don't want you to repair and instead want you to buy new things. I think this was one of the big contributing reasons why the right turned so much on tech in general, because they can always use the message of farmers.
Which I still dislike Apple for and think it is hypocritical that they market themselves so much around recycled electronics and parts. We've all seen those open fires in third world countries. Making their products out of recycled materials is great, but preventing them from being upgradable and repairable undermines that entire message. The problem is that there's no good choices and creating a new competitor is exceptionally difficult. The market is cornered by two players and they have similar incentives. That's not good for consumers, that's not good for innovation, it's not good for the economy, and it isn't even good for those two players (in short term, yes, but not in the long run. But being short sighted is quite frankly what led us here in the first place).
I wonder what the internal experience is like SSHing into one of these Apple Cloud Server boxes.
I mean, sure, it’s similar to what you’d surely get running Terminal.app, but I mean I doubt Apple engineers are doing some internal equivalent of apt when they need something.
There’s brew, but it would be… awkward if that’s what they used in their private cloud fleet.
Apple’s article on how their Private Cloud Compute system works is a good read if you are interested in the details.
> First, we intentionally did not include remote shell or interactive debugging mechanisms on the PCC node. Our Code Signing machinery prevents such mechanisms from loading additional code, but this sort of open-ended access would provide a broad attack surface to subvert the system’s security or privacy. Beyond simply not including a shell, remote or otherwise, PCC nodes cannot enable Developer Mode and do not include the tools needed by debugging workflows.
Yeah, they somewhat hinted at it in their WWDC announcement about private AI. They are running models on apple hardware. If you look up their jobs portal you can see some roles that also seem to hint at that (OS and kernel teams)
I could really see this taking off. Sagemaker is garbage and hugging face interfaces are laggy and buggy. The future is on-prem bespoke models you can run on commodity NPU SoCs. On my M3 MacBook I’ve been running local models like Llama-70B instead of ChatGPT for a while now.
It will be funny to see Apple crack under pressure from the AI bubble.
They have been severely mismanaged for a decade by forcing mandatory yearly upgrades across all of their stacks.
In the past several years the cracks have started to show in their software (bugs lingering for years, features being delayed for months after yearly OS upgrades etc.). Hardware teams seemed to be mostly immune from it, and delivered insane results.
Now it's "omg must have AI don't matter if you're ready for it or no. Upper management saw LLM demos and are now forcing it everywhere. What do you mean you had other plans for M5? It's AI now damn it"
62 comments
[ 2.8 ms ] story [ 121 ms ] threadBut in this specific case it’s utter clickbait.
Dual use: https://en.wiktionary.org/wiki/dual-use
It would be of interest because normally you'd be optimizing for very different characteristics.
Servers normally have a lot of memory slots, e.g. Epyc servers have more memory bandwidth than the top end Apple Silicon entirely because the servers have tons of memory channels.
The stacked logic setup is far more interesting in terms of what it means for die density and efficiency.
I'm still on M1 and I know that M2 and M3 would not change my experience for my web development work.
But yeah, if you're just going off the maker's marketing materials, then yeah, you'd always feel behind.
Granted, no gaming or number-crunching (that goes on a server!), just coding, browsers, and Excel.
M4 is tempting and a decent time to upgrade if you’ve had an M1 since 2020 or 2021, but its exact specifications aren’t know for all lines yet.
So a plain M3 is better than a plain M2. The top end M3 Max is better than the M2 Max. For the rest, you have to do a comparison and determine based on your intended use-cases, but for most people the M3 {Pro,Max} will be better than the M2 equivalent.
See tables here
https://en.wikipedia.org/wiki/Apple_M2#Variants
https://en.wikipedia.org/wiki/Apple_M3#Variants
It’s super nice that they overengineer everything and that’s why I buy Apple products to be honest, but they should roll out their chips in a sensible way and not leave their most expenses products at the tail end to upgrade right before their release their next chip.
Plus I think that it makes sense to wait for the M4, because it will have enough TOPs to support their future AI endeavors and will have a much longer useful life than purchasing and M2 studio today.
I have a Bluetooth keyboard and somehow all f-keys (F1-F12) just won’t register. The same happens in remote desktop from the ipad to a real computer.
So yeah… it works as long as you stay within the limits apple has randomly chosen for you.
>>> Ability to ssh and use the iPad like a netbook
Note the order of the words and the comparison to a netbook. They did not ask about "ssh TO an iPad" (also, no need to shout, you can always use *to* to get to). In the comparison to a netbook, the use-case is ssh'ing from the device to other devices. Netbooks being a category of laptops that were small, generally underpowered, but typically with good battery life, great for mobile access to remote servers and workstations.
I had netbook in ~2009 or something, it was with the Linux preinstalled (Ubuntu as I remember) and definitely had ability to log in into it. I absolutely associate netbook as a device to which I can log to.
I know some will say "well that's not what an iPad is for" and I just have to ask "why not?" I think there are a lot of people that want some machine where there's a detachable keyboard, can work mostly like a computer, but also can work like a tablet. When tablets were first made, they had to make compromises because of the computation and architecture limits. But those limitations don't really exist anymore.
Also that tablet functionality shouldn't be weirdly hindered. Like why can't I draw on the screen? I can do that on my computer with my mouse but not the device where they sell a $100 pencil? Shouldn't that be the "magic" where I can present anything and communicate better? What if I want to pull up a webpage and then we need to highlight or talk about things? Why do I need a ton of apps to do what one of the key points the iPad is literally marketed for? Taking notes? Who doesn't have GoodNotes or Notability? (both of which strangely lack features) That's a low fucking bar and even the new OS doesn't do a great job. But this should have been there on day 1. Seriously: give me a way to make folders and organize (mine does not let me have subfolders! If you want to go all out, allow me to connect notes to PDFs! Even better if I can link note sections to PDF sections!), and just give me a few modifiable templates (blank, lines, grids. Then let me change the two colors and spacing. You can't tell me this isn't trivial to expose to the user. We're programmers, this shit doesn't need to be hard coded, though presets are good to have). And give me an infinite scrolling sheet too. That's all I need. Other things are nice, but this core functionality is missing in the Apple native Notes and Pages (but more exists if I simply open up a PDF directly, though what do I do when I can't write in the margins?!). I don't get where the disconnect is. I often hear these same complaints from other academics and from students. It just feels like the people in charge of making decisions aren't actually using the product.
And the worst part of it all is, why isn't anyone making a different better product? It seems like Microsoft, Google, and Samsung are stuck in the same similar feedback loops.
I don't know about you all, but the reason I started programming and the main reason I still program is because I need to make things that don't yet exist. Sometimes it can be a small specific task. Maybe it is that I need something to function differently than it does because I have unique situations or tastes. It used to be that companies couldn't keep us out from modifying things. But it begs the question as to why keep us out in the first place? Especially Apple. To market themselves as the brand of creativity, yet there is only "The Apple Way." You'd think that those that grew up hacking machines and seeing the power and utility in modifiability would take those lessons and build them into the core of products. I understand protecting dumb users, but you can have advanced menus. But it is that exploration and adaption that is at the core of humanity and creativity. That's why I'm passionate here. Because tech can do so much more.
this becomes immediately obvious as soon as consoles get brought up. Not only do people not care about a mandatory 30% cut of revenue and no sideloading (except a limited capability for self-signing that isn’t allowed to utilize the standard APIs or the full performance of the chips) but they actively argue that the console is just a games appliance. And in fact that idea has been enshrined into law, game consoles get a specific carve out within the DMA just so they don’t have to!
it’s always been about spite, the things you’re saying sound nice but they’re just a convenient and temporary port-of-call for people with other objectives, and the things you outline here will be forgotten as soon as it’s convenient.
Just like RCS and whatever: like not only is Google’s proprietary encryption extensions a poison pill, but google itself doesn’t even care enough to implement rcs in its own services. Google voice has less RCS support than apple does in tyool 2024.
This becomes immediately obvious as soon as John-Deere gets brought up. Not only do people realize that outlawing repairs should be illegal, but they actively admit that the world would be a better place if spiteful conduct was made illegal.
Which I still dislike Apple for and think it is hypocritical that they market themselves so much around recycled electronics and parts. We've all seen those open fires in third world countries. Making their products out of recycled materials is great, but preventing them from being upgradable and repairable undermines that entire message. The problem is that there's no good choices and creating a new competitor is exceptionally difficult. The market is cornered by two players and they have similar incentives. That's not good for consumers, that's not good for innovation, it's not good for the economy, and it isn't even good for those two players (in short term, yes, but not in the long run. But being short sighted is quite frankly what led us here in the first place).
I mean, sure, it’s similar to what you’d surely get running Terminal.app, but I mean I doubt Apple engineers are doing some internal equivalent of apt when they need something.
There’s brew, but it would be… awkward if that’s what they used in their private cloud fleet.
> First, we intentionally did not include remote shell or interactive debugging mechanisms on the PCC node. Our Code Signing machinery prevents such mechanisms from loading additional code, but this sort of open-ended access would provide a broad attack surface to subvert the system’s security or privacy. Beyond simply not including a shell, remote or otherwise, PCC nodes cannot enable Developer Mode and do not include the tools needed by debugging workflows.
— https://security.apple.com/blog/private-cloud-compute/
They have been severely mismanaged for a decade by forcing mandatory yearly upgrades across all of their stacks.
In the past several years the cracks have started to show in their software (bugs lingering for years, features being delayed for months after yearly OS upgrades etc.). Hardware teams seemed to be mostly immune from it, and delivered insane results.
Now it's "omg must have AI don't matter if you're ready for it or no. Upper management saw LLM demos and are now forcing it everywhere. What do you mean you had other plans for M5? It's AI now damn it"