Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Apple M1 Ultra

apple.com

Post reply on HN
Print view View original

281–290 of 855 posts

Re: Apple M1 Ultra

#281

Post by adfgadfgaery »

pphysch wrote: post #116

> This enables M1 Ultra to behave and be recognized by software as one chip, so developers don’t need to rewrite code to take advantage of its performance. There’s never been anything like it. Since when did the average developer care about how many sockets a mobo has...? Surely you still have to carefully pin processes and reason about memory access patterns if you want maximum performance.

This line is nonsense and you can safely ignore it. There have been multi-chip-modules that act like a single socket for many years. In particular, pretty much every current AMD CPU works that way. I guarantee you that for the M1 Ultra, just like every CPU before it, the abstraction will be leaky. Programmers will still care about the interconnect when eking out the last few percent of performance.

Remember the Pentium D? Unfortunately, I used to own one.

Re: Apple M1 Ultra

#282

Post by LegitShady »

gowld wrote: post #90
brutal_boi wrote: post #32

Earlier quoted context omitted.

From the article: > Apple even says its new GPU is a match for Nvidia's RTX 3080 mobile chip, though you'll have to take Apple's word for it on that one. We've also reached out to Nvidia to see what it might have to say on the matter. > RTX 3080 mobile chip > mobile chip There's a 50%[1] (!) difference with mobile and non-mobile versions of the chip. So that's hardly a deal breaker. [1] https://www.videocardbenchmark…

The "mobile" scam in GPUs is terrible. Nvidia flat out lies about Mobile performance by giving misleading product names (same as the desktop names).

its beyond that. the same chip might have several tdps and drastic performance differences between models, such that a high tdp 3070 mobile is faster than a low tdp 3080. you end up having to get benchmarks for each particular laptop configuration.

Re: Apple M1 Ultra

#284

Post by savant_penguin »

polyrand wrote: post #258

I think the GPU claims are interesting. According to the graph's footer, the M1 Ultra was compared to an RTX 3090. If the performance/wattage claims are correct, I'm wondering if the Mac Studio could become an "affordable" personal machine learning workstation (which also won't make the electricity bill skyrocket). If Pytorch becomes stable and easy to use on Apple Silicon [0][1], it could be an appealing choice. [0]…

And hopefully not make you deaf with their buzzing fans

Re: Apple M1 Ultra

#285

Post by dathinab »

gjsman-1000 wrote: post #161
zamadatix wrote: post #133

Earlier quoted context omitted.

But the answer to the question is still "no".

Doesn't have to though. A Threadripper 3990X uses barrels of electricity, generates plenty of heat, comes with no GPU, has worse single-threaded performance, and still costs $4000 by itself without any of the parts needed to make it actually work.

Through you would need to compare it to the coming threadripper 5000WX(?) or better the soon coming Ryzen 7000 CPUs (which seen to have integrated graphics).

I mean they all are CPUs coming out this year as far as I know.

Re: Apple M1 Ultra

#286

Post by paulmd »

zitterbewegung wrote: post #54
cube2222 wrote: post #4

Looks like all the people saying "just start fusing those M1 CPU's into bigger ones" were right, that's basically what they did here (fused two M1 Max'es together). And since the presenter mentioned the Mac Pro would come on another day, I wonder if they'll just do 4x M1 Max for that.

They might have to have the unified memory more dense to get to 1.5 TB max of RAM on the machine (also since this would be originally shared with a GPU). Maybe they could stack the RAM on the SoC or just get the RAM at a lower process node.

The M1 Max/Ultra is already extremely dense design for that approach, it's really almost as dense as you can make it. There's packages stacked on top, and around, etc. I guess you could put more memory on the backside but that's not going to do more than double it, assuming it even has the pinout for that (let's say you could run it in clamshell mode like GDDR, no idea if that's actually possible, but just hypothetically).

The thing is they're at 128GB which is way way far from 1.5TB. You're not going to find a way to get 12x the memory while still doing the embedded memory packages.

Maybe I'll be pleasantly surprised but it seems like they're either going to switch to (R/LR)DIMMs for the Mac Pro or else it's going to be a "down" generation. And to be fair that's fine, they'll be making Intel Mac Pros for a while longer (just like with the other product segments), they don't have to have every single metric be better, they can put out something that only does 256GB or 512GB or whatever and that would be fine for a lot of people.

Re: Apple M1 Ultra

#287

Post by Mikeb85 »

Ugh really wish that a non-Apple vendor could make an ARM chip of this calibre. Jealous but cat bring myself to use a proprietary OS and get locked into Apple.

Re: Apple M1 Ultra

#288

Post by kiratp »

fivea wrote: post #261
kiratp wrote: post #230

Earlier quoted context omitted.

My workstation has a 3990x. Our "world" build is slightly faster on my M1 Max. https://twitter.com/kiratpandya/status/1457438725680480257 The 3990x runs a bit faster on the initial compile stage but the linking is single threaded and the M1 Max catches up at that point. I expect the M1 Ultra to crush the 3990x on compile time.

> The 3990x runs a bit faster on the initial compile stage but the linking is single threaded and the M1 Max catches up at that point. Isn't linking IO-bound?

Exposing my limited understanding of that level of the computing stack - it is but Apple seems to have very very good caching strategies - filesystem and L1/2/3.

https://llvm.org/devmtg/2017-10/slides/Ueyama-lld.pdf

There is a breakdown in those slides discussing what parts of lld are single threaded and hard to parallelize so I suspect single thread performance plays a big role too. I generally observe one core pegged during linking.

Re: Apple M1 Ultra

#289

Post by swyx »

was surprised to learn that the CPUs and GPUs on the M1x chips are essentially a single unit, and for the M1 Ultra they basically slapped two M1’s together.

in traditional PC building, the CPU is quite distinct from the GPU. can anyone ELI5 what the benefits are to having the CPU closely integrated with GPU like the M1 has? seems a bit unwieldy but i dont know anything about computer architecture

Re: Apple M1 Ultra

#290

Post by Koshkin »

polyrand wrote: post #258

I think the GPU claims are interesting. According to the graph's footer, the M1 Ultra was compared to an RTX 3090. If the performance/wattage claims are correct, I'm wondering if the Mac Studio could become an "affordable" personal machine learning workstation (which also won't make the electricity bill skyrocket). If Pytorch becomes stable and easy to use on Apple Silicon [0][1], it could be an appealing choice. [0]…

For some definitions of "affordable."
Post reply on HN