Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Working on Chips for New Macs and AI Servers
- Thread starter MacRumors
- Start date
- Sort by reaction score
Absolutely. It was obvious at the time that it was a three pronged loiss at the time, Just like not spending the money for the Nvidia AI chips for work on Apple Intelligence when the team was begging for them, and announcing it before it was ready and making fools of themselves.
Eventually Tim and other top execs will run out of people to blame.
I have yet to get Genmoji to actually give me either a) anything, or b) something that represents what I asked for.
Latest -- "Clock showing time at 3'oclock". What it produced was a clock with 3 hands pointing at 10,12, and 2 o'clock (plus a seconds hand!).
Most other attempts result in nothing. Another useless AI feature on my iPhone 16 PM.
My point was more on it's easy to say this in hindsight, but at the time, there was no way to know what the future holds and AI was not that widely popular then. And their CPU's didn't beat Apple CPU's, their claims were based on giving them full power when testing and then testing the battery longevity on lower power, but their chips were excellent and really made Intel and AMD increase their efforts. But
...but Apple don't make servers, so why would they want to build server chips?
Apple's core market is in mobile devices and ultra-portable laptops with cool designs, easy-to-use GUIs and seamless integration of apps and services (how well they're delivering on that is for another thread). Those aren't really important in the server market.
Apple Silicon still seems to be comfortably ahead of the game vs. Qualcomm Snapdragon in that respect. Even high-end personal workstations like the Mac Pro are probably a dead-end - more and more can be done with laptops and small-form-factor systems at one end of that scale, and more and more can be done in the cloud (dominated by a couple of fairly inassailable large players) at the other.
Yes, they made XServe. It was good for a few years. Then they dropped it because it ceased to have a reason to exist. Back when the XServe was launched, Macs needed proprietary services for file sharing etc. and the PC industry was using its own proprietary systems like Windows Server & Novell Netware which cost a fortune in per-seat licensing, which MacOS X could undercut. By the time it was dropped, everybody - including Macs - was happily using SMB and open Internet/W3C protocols - and Windows Server was facing huge competition from you-can't-beat-free Linux. Sleek design and user-friendly GUIs really don't matter for a server sitting in a rack running predominantly Unix/Linux server software (...and configuring server software via a GUI is like washing your feet with socks on). The XServe was just a.n.other generic rack-mount x86 *nix server box - not something that could command the soft of premium that consumer Macs & iDevices do.
Today, we're not even talking about the sort of multi-ARM-core/large I/O bandwidth Xeon-killer "server" chip that Ampere and AWS were pushing 10 years ago. We're talking about chips for AI services - i.e. neural engines and non-for-graphics GPUs. Apple Silicon is actually pretty good as a delivery platform for AI services (even if Apple's own implementations of AI software are a bit limp) and maybe Apple Silicon-derived technology has something to offer there... esp. on power efficiency (a selling point who's value waxes and wanes with the oil price...) but it's really not the tech - or even the same service - that we were talking about a decade ago.
...also, the "so-called AI" (aka party tricks with LLMs and diffusion) industry is a massive bubble liable to burst at any moment. That doesn't mean that AI is going away (any more than the web went away after the dot com crash) - but a lot of shirts are going to be lost.
Do we know anything new about M5 chips? I assume they’ve been working on M6 and M7 but, what do we know about the chips that are already being produced? No new leaks about the base M5?
My understanding that that since Apple dropped Mac servers several decades ago they rely exclusively on x86 servers running linux and programmed in C++. Apple definitely does not eat their own dog food. And that also explains why there are so many bugs in Apple non-C++ arm software.
Not that I am aware of. I don't think there has even been a codename leak for M5. The big M5 rumor that seems solid, now from at least two different industry sources, is that the M5 Pro/Max will use SoIC instead of SoC.
So M5 and M5 Pro/Max/Ultra could be the first generation of Apple silicon with that distinction, where the M5 Pro/Max/Ultra have an integrated SoIC structure, while the A19, A19 Pro, and base M5 all have a monolithic SoC design.
Last edited:
Yeah, I am aware of such rumors regarding the M5 Pro/Max. But apparently there are no leaks about the base M5 SoC. We don’t know if they will improve the Neural Engine adding more cores, or improve the GPU with a new architecture or more cores. Or if it will be an overclocked M4 (likely) with bigger cachés (less likely) and faster RAM (even less likely)
Ability to make Mac Pro's chip will directly important for them to create server and super computer just like Nvidia did. If you ever checked Nvidia, they provided a workstation with 4x GPU for a personal user like DGX. Dont forget that Apple also made a server based on Mac Pro.
Then where is Mac Pro's chip? Apple cant even make sever and super computer grade chip unlike Nvidia. Cant and Can is a huge difference.Errr.. Nvida?
Apple's solution is much more rigidly "Homogenous Unified Memory" but others are on the MCM and/or multiple die path also. There is no DIMMs slots for mega-modular memory there.
A Quick Introduction to the NVIDIA GH200 aka Grace Hopper
The NVIDIA GH200 or "Grace Hopper" is far from a single product. We have a quick guide so when someone says "GH200" you know what to look forwww.servethehome.com
I said Apple Silicon, NOT Nvidia.
Doesn't change the fact that Apple Silicon sucks for GPU and AI resources. That's also why they are struggling with their own AI server which only has 50,000 made out of more than 5 years old GPU, Cant even make a powerful desktop chip, and heavily limited by SoC as they need to make a huge chip.
Power Efficiency? Yes, it's great but when the max performance and expandability is IMPOSSIBLE, it's totally pointless. So far, they only can use as many Mac mini/Studio as possible to create a farm but that would VERY inefficient especially compared to Nvidia as they are being a monopoly by 90% in GPU market and being inevitable as they are the only solution for AI.
If you truly believe that Apple Silicon is the best, think again. There are reasons why Nvidia is dominating the entire
AI markets.
Yes there are, though your hating on Apple has clouded your summary.
Nvidia succeeded because:
1) Intel has been flopping around, for several years, like a dying fish on a wharf;
2) Export bans have kept the cutting edge photolithography out of the hands of the PRC, therefor Chinese startups in their IC industry are hobbled;
3) the major ARM licensees, such as Apple and Samsung, were not in that business.
Being logical is hating? Wow. If you think that way, tell me if there are any Apple Silicon based super computer to train and study AI for their own company before you even say other things.
Apple have Metal for their own but still, it's such a tiny market especially since Apple lacks their own software. But once Apple can make Mac Pro grade Apple Silicon chip, then they can at least try.
I get beefing up iCloud in order to better share and delineate the personal AI/compute from the cloud. They struggled with that. Even that's been on the roadmap, so where's the news here? Besides, there's been a sense of doing this out-of-order for a few years. Build the reliable stack and scale. Apple hasn't built the first part yet, so this is all putting the cart before the horse. It's not like Apple hasn't made bad bets before (trash can Mac Pro).
The problem is that AI models are basically implemented by researchers, and researchers chose CUDA a long time ago, and they won’t change (an unwise choice if you ask me, because it’s vendor specific, but most researchers are not smart when it comes to real implementation).
Right now, it seems nVidia cuda and tensor cores are the star attraction for tensor operations. Even AMDs ROCm can't compete on the same playfield. I would love for Apple to break nVidia's stranglehold, but I don't have much faith in Apple's AI prowess right now, and is the only reason I still keep a PC kicking around.