Hacker News Reader

The whole thing is imaginary: "In this paper, we propose Chiplet Cloud, a chiplet-based ASIC AI-supercomputer architecture that optimizes total cost of ownership (TCO) per generated token for serving large generative language models to reduce the overall cost to deploy and run these applica- tions in the real world."

So they are comparing actual implementations with a theoretical implementation. Never mind that they got the A100 figures wrong, they are still in the 'wouldn't it be nice if we had 'x'' stage. This looks like a paper whose sole purpose is to raise funds for a research project that will probably ultimately go nowhere and they needed a reason that looks good on paper to increase their chances of getting funded. A100 can already be had for $0.87/hour so even their theoretical advantage is under significant pressure and assuming they got everything else right by the time the project has run the market will have overtaken them. This is what usually happens to CPUs that are application specific.

The $0.87/hour price you gave is theoretical and also we know any price in a paper for compute is wrong by the time of publication.

Pragmatically the prices are closer to $2/hr according to this recent post here on Hacker News: https://llm-utils.org/Nvidia+H100+and+A100+GPUs+-+comparing+...

Although again prices change on a daily basis on spot providers.

> The $0.87/hour price you gave is theoretical

https://cloud.google.com/blog/products/compute/a2-vms-with-n...

That's as close as I got to verifying that price.

Another list that shows pricing both constant and spot. The best GCP spot price is $1.1, but Jarvis seems to say its spot for the 40GB A100 is $0.69:

https://fullstackdeeplearning.com/cloud-gpus/

I feel there are more fair criticisms of that paper than its inclusion of the snapshot price of variable priced compute resource.

Sure, to me it more of an extra item than the main one but it is one that you can readily verify because most of the other claims are far more vague. If they're willing to fudge on that one then I have much less confidence in the rest of their claims.

Seems like HN comments have determined that the cost number is not fudged..

take 3 ideas that are hot: chiplets, cloud, and LLM - remix them into the title of a paper that describes a hypothetical machine.. academia playing catch up and trying to stay relevant in my cynical eye.

Using ChatGPT

I asked gpt for giggles and the comparison is much more thorough, it has written also power per watt improvements, benefits of denser packing, and sustainability of moving toward more energy efficient solution.

I did the same exact mind exercise using ChatGPT but I haven't produced a paper out of the chat session.

Yeah, it is an architectural simulation study, this is what is usually done right at the beginning before resources are allocated to go deep on idea. So in that sense it is imaginary; but this is how new ideas get incubated.

I think the costs are from the Moonwalk model, which is a pretty good reference for estimating costs, although it might be low if you use all Google engineers to build the HW. =P

Looks like figure 8 of paper [1] says it is 18 tokens/s

Doesn’t that graph have a toks/sec of 18? Or am I reading it wrong

It shows 18 tokens per second but that's how fast tokens are generated I think. The number of tokens generated is that times the batch size, which appears to be 12? The graph is quite unclear and I didn't feel like reading the paper more in-depth.

Seems to me the 18 tokens per second from [1] is the throughput and includes the batch size, so I don't think they misread the Deepspeed inference paper. So the chiplet ASIC supercomputer paper would seem to show a decent performance/TCO benefit.

Of course, it's a first architectural study to illustrate the promise of the idea, lots more details to work out in a physical implementation and the final realized benefit is likely to be lower. But even a 3X is huge in this space.

If you look at the "metrics" section on Page 9, it says:

> 2) Metrics: We use three performance metrics: (i) latency, i.e., end-to-end output generation time for a batch of input prompts, (ii) token throughput, i.e., tokens-per-second processed, and (iii) compute throughput, i.e., TFLOPS per GPU.

This is somewhat confusing to me because at least two of these three definitions should be essentially the same thing, but I don't think there's any way to interpret their claim of ~74 teraflops achieved other than ~211 tokens/second of throughput.

Put another way, 18 tokens per second is 2% flops utilization, which we are obviously capable of doing better than for bulk inference.

3x is not huge in this space because just using a 4090 instead of an A100 is a 5x gain.

Maybe 74 tflops is the best they've achieved, but not all 16 GPUs can consistently hit that number? Just guessing.. The 211 tokens/sec throughput on GPU is just insane, it's even better than what TPU can do on PaLM 540B.

Well LM-175B is 540/175=3.08x smaller, so it makes sense you would get better performance. Also, in Table D.4 it takes them 9.614s to process (28 input + 8 output tokens = 136 tok * 256 batches = 34,816 tokens with 24 A100s, which is ~150 tok/s/A100. It feels totally plausible that they could hit 211 tok/s with a bigger model. I think 211 tok/s is in fact a pretty poor showing from them and you could do significantly better.

The 4090 has half the memory bandwidth, so it could not get a 5X gain, it would actually run slower on a memory bound LLM like this.

5x gain per dollar

One possible explanation is that they hit the teraflops number during the prefill stage, where you can process all tokens at once, and are generally more operationally intensive, so you can use more compute. Utilization usually drops during the token generation stage. The utilization of the TPU during token generation is 3% when batch size 16. (https://arxiv.org/pdf/2211.05102.pdf, Table on the last page).

I mean maybe? This seems unlikely. I agree that decode is much more expensive and tok/s depends a lot on what your ratio of decode tokens to prefill tokens is.

This table was very helpful by the way, I didn't see that before. To me it clearly shows that 211 tok/s/A100 is very plausible and in fact kind of a poor showing because if you look at table D.4 and specifically the results for BS=256 PP3/TP8 they achieve ~150 tok/s/A100 on a model that's 3x larger than GPT-3.