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

611–620 of 855 posts

Re: Apple M1 Ultra

#611

Post by hajile »

outworlder wrote: post #367
Geee wrote: post #307

This is insane. They claim that its GPU performance tops the RTX 3090, while using 200W less power. I happen to have this GPU on my PC and not only it costs over $3000, but its also very power-hungry and loud. Currently, you need this kind of GPU performance for high resolution VR gaming at 90 fps, but its just barely enough. This means that the GPU will run very loudly and heat up the room, and running games like HL…

Too bad that historically Apple has not given any attention to Mac gaming.

Apple is the world's biggest and most profitable gaming company. For every AAA gamer, there are a hundred casual gamers (one reason why Nintendo consoles run circles around Sony and Microsoft).

Apple has invested billions into their gaming division. The big thing they need right now is a new version of Metal that gets feature parity with Vulkan or DX.

Also of note, there are very persistent rumors of an upcoming VR headset. Their M1 alone would blow away competition like the Quest. A Pro or max chip with some disabled CPU cores wouldn't cost a ton due to being scavenged cores and would positively stomp the competition.

Re: Apple M1 Ultra

#612

Post by cyber_kinetist »

vimy wrote: post #609
whoisburbansky wrote: post #301

Earlier quoted context omitted.

Cursory look gives you a ~$3500 price tag for a gaming PC with a 3090 [1], vs. at least $4k for a Mac Studio with an M1 Ultra. Roughly the same ballpark, but I wouldn't call the M1 Ultra more affordable given those numbers. 1. https://techguided.com/best-rtx-3090-gaming-pc/#:~:text=With... .

You also need to compare the right cpu. M1 Ultra cpu is the equivalent of the fastest threadripper. Which costs $3990. So a pc with similar performance would be $7500

Not the top-of-the-line threadripper (which can go up to 32c/64t), but probably similar to the 5950x (16c/32t), which costs like 1000$.

But you’re comparing apples to oranges, because the real advantage of M1 chips is the unified memory - almost no CPU-GPU communication overhead, and that the GPU can use ginormous amounts of memory.

Re: Apple M1 Ultra

#613

Post by Tagbert »

kllrnohj wrote: post #360
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]…

The GPU claims on the M1 Pro & Max were, let's say, cherry picked to put it nicely. The M1 Ultra claims already look suspicious since the GPU graph tops out at ~120W & the CPU graph tops out at ~60W yet the M1 Studio is rated for 370W continuous power draw. Since you mention ML specifically, looking at some benchmarks out there (like https://tlkh.dev/benchmarking-the-apple-m1-max#heading-gpu & https://wandb.ai/tcapel…

You are comparing the power source (370W) to the CPU/SOC (120W). The power supply provides power for USB-C/Thunderbolt ports and it’s never a good idea to spec a power supply too low and run it too close to capacity.

Re: Apple M1 Ultra

#614

Post by dev_tty01 »

forgotmyoldacc wrote: post #387
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]…

Neural Engine cores are not accessible for third party developers, so it'll be severely constrained for practical purposes. Currently the M1 Max is no match for even last generation mid-tier Nvidia GPU.

Huh? Neural engine is certainly usable by developers. You just use the CoreML framework.

Re: Apple M1 Ultra

#615

Post by wyager »

Geee wrote: post #307

This is insane. They claim that its GPU performance tops the RTX 3090, while using 200W less power. I happen to have this GPU on my PC and not only it costs over $3000, but its also very power-hungry and loud. Currently, you need this kind of GPU performance for high resolution VR gaming at 90 fps, but its just barely enough. This means that the GPU will run very loudly and heat up the room, and running games like HL…

> you need this kind of GPU performance for high resolution VR gaming at 90 fps, but its just barely enough

I run VR games on the index at 144hz with high settings without issue on a 3060 Ti.

I've been on the market for a 3070-3090, but only because I want a card for which a water block is available, not because I need more power for any extant game.

Re: Apple M1 Ultra

#616

Post by adfgadfgaery »

kllrnohj wrote: post #392
adfgadfgaery wrote: post #281

Earlier quoted context omitted.

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 Penti…

The existing AMD CPUs aren't quite like that. Technically they are all UMA, not NUMA - the L3 cache is distributed, but they are all behind a single memory controller with consistent latencies to all cores. But the Threadripper 1st gen was absolutely like that. Straight up 2+ CPUs connected via infinity fabric pretending to be a single CPU. So is that 56 core Xeon that Intel was bragging about for a while there until…

>Technically they are all UMA, not NUMA - the L3 cache is distributed, but they are all behind a single memory controller with consistent latencies to all cores.

This stuff rapidly starts to make my head spin. I have not studied interconnects and have never written any NUMA-aware software. I will just post this link (read the "Memory Latency" section):

https://www.anandtech.com/show/16529/amd-epyc-milan-review/4

As I understand it, the I/O die is partitioned into four quadrants. Each quadrant has two memory controllers and is attached to two compute dies. CPUs can access memory attached to the same quadrant with lower latency than going to another quadrant. This is a NUMA system that can be configured to appear as one logical NUMA node.

I believe their smaller parts with two or fewer compute dies will be UMA, but with the same non-uniform latency to L3.

>So is that 56 core Xeon that Intel was bragging about for a while there until the 64 core Epycs & Threadrippers embarrassed the hell out of it.

I believe the 64-core Epycs and Threadrippers came first. The 56-core Xeon was a purpose-built part for HPC, so it wasn't quite a marketing gimmick.

Re: Apple M1 Ultra

#617

Post by wyager »

Geee wrote: post #554
tanduv wrote: post #522

Earlier quoted context omitted.

HL Alyx is actually quite well optimized, and you can definitely run it on Ultra with super-sampling on a 3090. [1] https://www.youtube.com/watch?v=kjNaC0-hiPE

I didn't see display resolution (or the model of HMD) mentioned in the video. I'm using Varjo Aero, which has 2880x2720 per eye, which is almost quapruple (per eye) compared to Valve Index. I think this resolution is enough for a good VR experience; pixels are almost invisible, and even small text is readable. However, HL Alyx doesn't run at 90 fps on full resolution.

That HMD looks pretty badass. Didn't realize hardware support for eye tracking/foveation already existed. Does any software support it? That would fix your perf issue.

Re: Apple M1 Ultra

#618

Post by bee_rider »

fomine3 wrote: post #551
bee_rider wrote: post #342

Earlier quoted context omitted.

Could they do a multi-socket board for the Mac Pro?

I expect this for CPU side. Multiple SoC introduces NUMA but it's already done on Dual Xeon based Mac. I wonder how their GPU work for the configuration.

In a world where latency:storage size tradeoffs need to be made for practical reasons (and will, at some point, be required for fundamental physical reasons), we should just embrace NUMA anyway. Death to the lie of uniform access! NUMA is the future!

Ehrm, anyway.

It actually isn't clear to me whether designing a two socket motherboard is fundamentally an easier task than jamming more of the things into a single package (given that they have already embraced some sort of chiplette paradigm).

Re: Apple M1 Ultra

#619

Post by wyager »

quickthrower2 wrote: post #441
Geee wrote: post #307

This is insane. They claim that its GPU performance tops the RTX 3090, while using 200W less power. I happen to have this GPU on my PC and not only it costs over $3000, but its also very power-hungry and loud. Currently, you need this kind of GPU performance for high resolution VR gaming at 90 fps, but its just barely enough. This means that the GPU will run very loudly and heat up the room, and running games like HL…

I wonder how it compares for crypto mining, and if that group would be buying these up

If you're mining crypto you want 95% of your capex to go to graphics cards/ASICs, not to RAM+NICs+CPU+case+etc.

Re: Apple M1 Ultra

#620

Post by noisy_boy »

snowwrestler wrote: post #412
marcan_42 wrote: post #172

Earlier quoted context omitted.

I've been saying 4x M1 Max is not a thing and never will be a thing ever since the week I got my M1 Max and saw that the IRQ controller was only instantiated to support 2 dies, but everyone kept parroting that nonsense the Bloomberg reporter said about a 4-die version regardless... Turns out I was right. The Mac Pro chip will be a different thing/die.

Plus they are running out of M1 superlatives. They’ll have to go to M2 to avoid launching M1 Plaid.

M1 Mark II ala Sony
Post reply on HN