News for hackers. Stuff that matters.

A comment on MI300X vs. H100 vs. H200 Benchmark Part 1: Training – CUDA Moat Still Alive. Parent comment

Re:MI300X vs. H100 vs. H200 Benchmark Part 1: Training – CUD... by Twirrim on Monday December 23, 2024 @12:21AM (#42492006)
Engineers at hyperscalers are struggling through all the bugs too. It's coming at notable opportunity cost for them, at a time when they also want an end to the monopoly. Do they buy AMD and wade through bug after bug, regression after regression, or do they shell out slightly more money for Nvidia GPUs and have it "just work".

AMD has to get on top of their software quality issues if they're ever going to succeed in this segment, or they need to be producing chips so much faster than Nvidia that it's worth the extra time investment and pain.

[ Parent | on HN ]
Re:MI300X vs. H100 vs. H200 Benchmark Part 1: Training – CUD... by sangnoir on Monday December 23, 2024 @05:42AM (#42493425)
> Engineers at hyperscalers are struggling through all the bugs too

[citation needed]

[ Parent | on HN ]
Re:MI300X vs. H100 vs. H200 Benchmark Part 1: Training – CUD... by almostgotcaught on Monday December 23, 2024 @08:12AM (#42494142)
Lolol 100% accurate. Go trawl through PRs to Triton by FB people to the AMD portion of the codebase.
[ Parent | on HN ]
Re:MI300X vs. H100 vs. H200 Benchmark Part 1: Training – CUD... by Twirrim on Monday December 23, 2024 @08:13AM (#42494147)
Sorry, but NDAs mean I can't say any more.
[ Parent | on HN ]
Re:MI300X vs. H100 vs. H200 Benchmark Part 1: Training – CUD... by mike_hearn on Monday December 23, 2024 @02:48PM (#42497078)
It's in the article. Meta don't use AMD for training and write their own kernels for inference. You can't train with AMD, full stop, because their software stack is so buggy.
[ Parent | on HN ]
Re:MI300X vs. H100 vs. H200 Benchmark Part 1: Training – CUD... by sangnoir on Monday December 23, 2024 @05:46PM (#42498214)
> It's in the article

The same article also states that AMD provided custom bug-fixes written by Principle Engineers to address bugs in a benchmark - this is software that will only become part of the public release in 2 quarters. I ask again, do you think AMD will not expedite non-public bug-fixes for hyperscalers?

> You can't train with AMD, full stop, because their software stack is so buggy.

Point 7 from the article:

>> The MI300X has a lower total cost of ownership (TCO) compared to the H100/H200, but training performance per TCO is worse on the MI300X on public stable releases of AMD software. This changes if one uses custom development builds of AMD software.

I have an inkling that Meta does not obtain MI300 drivers from https://download.amd.com

[ Parent | on HN ]
Re:MI300X vs. H100 vs. H200 Benchmark Part 1: Training – CUD... by mike_hearn on Tuesday December 24, 2024 @06:42AM (#42501293)
AI labs don't want to train models using a stack build some guy hacked up on his desktop last night that's been through no proper QA process. The cost of a training run that fails or results in a garbage model due to numerical errors are huge.

Which is why, as they say clearly, nobody is training models on AMD. Only inference, at most. I'm not sure why you keep claiming they are training using private drivers. They clearly aren't.

[ Parent | on HN ]
Re:MI300X vs. H100 vs. H200 Benchmark Part 1: Training – CUD... by sangnoir on Tuesday December 24, 2024 @05:54PM (#42505536)
Now I i see how where talking past each other. I 100% agree that none of if the hyperscalers are currently (publicly) training on AMD silicon. I disagree with forward-looking statements like it "can't" happen, because I can guarantee you several of them are actively working on making it possible to train on AMD chips - that's just too juicy a target for a bonus packets all the way up to directors: "Our team saved the org $x0 million in TCO vy enabling training on MI300/MI400X in our new clusters"
[ Parent | on HN ]
Re:MI300X vs. H100 vs. H200 Benchmark Part 1: Training – CUD... by mike_hearn on Thursday December 26, 2024 @06:11AM (#42514521)
Oh, I see. "Can't" means in the present tense in my previous sentence, it wasn't meant to be a definitive statement about the entire future.
[ Parent | on HN ]