Hacker News new | ask | show | jobs
by blitzar 321 days ago
> widely-available H100 GPUs

Just looked in the parts drawer at home and dont seem to have a $25,000 GPU for some inexplicable reason.

8 comments

Does it even make sense calling them 'GPUs' (I just checked NVIDIA product page for the H100 and it is indeed so)?

There should be a quicker way to differentiate between 'consumer-grade hardware that is mainly meant to be used for gaming and can also run LLMs inference in a limited way' and 'business-grade hardware whose main purpose is AI training or running inference for LLMs".

We are fast approaching the return of the math coprocessor. In fashion they say that trends tend to reappear roughly every two decades, its overdue.
Yeah I would love for Nvidia to introduce faster update cycle to their hardware, so that we'll have models like "H201", "H220", etc.

I think it will also make sense to replace "H" with a brand number, sort of like they already do for customer GPUs.

So then maybe one day we'll have a math coprocessor called "Nvidia 80287".

I remember the building hugh end workstations for a summer job in the 2000s, where I had to fit Tesla cards in the machines. I don't remember what their device names were, we just called them tesla cards.

"Accelerator card" makes a lot of sense to me.

It's called a tensorcore and it's in most GPUs
"GPGPU" was something from over a decade ago; for general purpose GPU computing
Yeah, Crysis came out in 2007 and could run physics on the GPU.
I think apple calls them NPUs and Broadcom calls them XPUs. Given they’re basically the number 2 and 3 accelerator manufacturers one of those probably works.
By the way I wonder, what has more performance, a $25 000 professional GPU or a bunch of cheaper consumer GPUs costing $25 000 in total?
Consumer GPUs in theory and by a large margin (10 5090s will eat an H100 lunch with 6 times the bandwidth, 3x VRAM and a relatively similar compute ratio), but your bottleneck is the interconnect and that is intentionally crippled to avoid beowulf GPU clusters eating into their datacenter market.

Last consumer GPU with NVLink was the RTX 3090. Even the workstation-grade GPUs lost it.

https://forums.developer.nvidia.com/t/rtx-a6000-ada-no-more-...

H100s also has custom async WGMMA instructions among other things. From what I understand, at least the async instructions formalize the notion of pipelining, which engineers were already implicitly using because to optimize memory accesses you're effectively trying to overlap them in that kind of optimal parallel manner.
I just specify SXM (node) when I want to differentiate from PCIe. We have H100s in both.
We could call the consumer ones GFX cards, and keep GPU for the matrix multiplying ones.
GPU stands for "graphics processing unit" so I'm not sure how your suggestion solves it.

Maybe renaming the device to an MPU, where the M stands for "matrix/math/mips" would make it more semantically correct?

I think that G was changed to "general", so now it's "general processing unit".
This doesn't seem to be true at all. It's a highly specialized chip for doing highly parallel operations. There's nothing general about it.

I looked around briefly and could find no evidence that it's been renamed. Do you have a source?

CPU is already the general (computing) processing unit so that wouldn't make sense
Well, does it come with graphics connectors?
Nope, doesn't have any of the required hardware to even process graphics iirc
Although the RTX Pro 6000 is not consumer-grade, it does come with graphics ports (four Displayports) and does render graphics like a consumer card :) So seems the difference between the segments is becoming smaller, not bigger.
That’s because it’s intended as a workstation GPU not one used in servers
Sure, but it still sits in the 'business-grade hardware whose main purpose is AI training or running inference for LLMs" segment parent mentioned, yet have graphics connectors so the only thing I'm saying is that just looking at that won't help you understand what segment the GPU goes into.
With Ollama i got the 20B model running on 8 TitanX cards (2015). Ollama distributed the model so that the 15GB of vram required was split evenly accross the 8 cards. The tok/s were faster than reading speed.
For the price of 8 decade old Titan X cards, someone could pick up a single modern GPU with 16GB or more of RAM.
They’re widely available to rent.

Unless you’re running it 24/7 for multiple years, it’s not going to be cost effective to buy the GPU instead of renting a hosted one.

For personal use you wouldn’t get a recent generation data center card anyway. You’d get something like a Mac Studio or Strix Halo and deal with the slower speed.

I rented H100 for training a couple of times and I found that they couldn't do training at all. Same code worked fine on Mac M1 or RTX 5080, but on H100 I was getting completely different results.

So I wonder what I could be doing wrong. In the end I just use RTX 5080 as my models fit neatly in the available RAM.

* by not working at all, I mean the scripts worked, but results were wrong. As if H100 couldn't do maths properly.

This comment made my day ty! Yeah definitely speaking from a datacenter perspective -- fastest piece of hardware I have in the parts drawer is probably my old iPhone 8.
>Just looked in the parts drawer at home and dont seem to have a $25,000 GPU for some inexplicable reason.

It just means you CAN buy one if you want, as in they're in stock and "available", not that you can necessarily afford one.

you can rent them for less then $2/h in a lot of places (maybe not in the drawer)
You might find $2.50 in change to use one for an hour though
available != cheap
available /əˈveɪləbl/

adjective: available

able to be used or obtained; at someone's disposal

You can rent one from most cloud providers for a few bucks an hour.
Might as well just use openai api
thats not the same thing at all
That depends on your intentions.