Related post: https://hazyresearch.stanford.edu/blog/2025-11-09-amd-brr
HipKittens: Fast and furious AMD kernels
hazyresearch.stanford.edu
HipKittens: Fast and furious AMD kernels
1–10 of 94 posts
Re: HipKittens: Fast and furious AMD kernels
#2One thing I don't understand about Nvidia’s valuation is that right now a small number of algorithms have 'won,' such as Transformers. The data is very important. Compared to the past where customized code was much more common, such as modeling code and HPC, the ecosystem was very important and it was almost impossible to implement all CUDA and related code.
Competitors now only need to optimize for a narrow set of algorithms. If a vendor can run vLLM and Transformers efficiently, a massive market becomes available. Consequently, companies like AMD or Huawei should be able to catch up easily. What, then, is Nvidia’s moat? Is InfiniBand enough?"
Re: HipKittens: Fast and furious AMD kernels
#3One thing I don't understand about Nvidia’s valuation is that right now a small number of algorithms have 'won,' such as Transformers. The data is very important. Compared to the past where customized code was much more common, such as modeling code and HPC, the ecosystem was very important and it was almost impossible to implement all CUDA and related code. Competitors now only need to optimize for a narrow set of a…
Infiniband is being replaced with UEC (and it isn't needed for inference). For inference there is no moat and smart players are buying/renting AMD or Google TPUs.
Re: HipKittens: Fast and furious AMD kernels
#4You'd think AMD would swing in on something like this and fund it with the money needed to succeed. I have no knowledge of it but my guess is no, AMD never misses an opportunity to miss an opportunity - when it comes to GPUs and AI.
Re: HipKittens: Fast and furious AMD kernels
#5Ahh, composable-kernel. The highest offender in the list of software that have produced unrecoverable OOMs in my Gentoo system (it’s actually Clang while compiling CK, which uses upwards of 2.5GB per thread).
Re: HipKittens: Fast and furious AMD kernels
#6You'd think AMD would swing in on something like this and fund it with the money needed to succeed. I have no knowledge of it but my guess is no, AMD never misses an opportunity to miss an opportunity - when it comes to GPUs and AI.
First rule of AMD stock is nobody understands AMD stock. I guess it’s also the same for AMD’s software endeavors.
Re: HipKittens: Fast and furious AMD kernels
#7One thing I don't understand about Nvidia’s valuation is that right now a small number of algorithms have 'won,' such as Transformers. The data is very important. Compared to the past where customized code was much more common, such as modeling code and HPC, the ecosystem was very important and it was almost impossible to implement all CUDA and related code. Competitors now only need to optimize for a narrow set of a…
The vast amount of CUDA libraries for anything you can think of. I think there’s where they have the biggest leverage.
Re: HipKittens: Fast and furious AMD kernels
#8One thing I don't understand about Nvidia’s valuation is that right now a small number of algorithms have 'won,' such as Transformers. The data is very important. Compared to the past where customized code was much more common, such as modeling code and HPC, the ecosystem was very important and it was almost impossible to implement all CUDA and related code. Competitors now only need to optimize for a narrow set of a…
Infiniband is being replaced with UEC (and it isn't needed for inference). For inference there is no moat and smart players are buying/renting AMD or Google TPUs.
I didn't know you can you buy Google TPUs now?
Re: HipKittens: Fast and furious AMD kernels
#9You'd think AMD would swing in on something like this and fund it with the money needed to succeed. I have no knowledge of it but my guess is no, AMD never misses an opportunity to miss an opportunity - when it comes to GPUs and AI.
From the performance comparison table, basically AMD could be NVIDIA right now, but they aren’t because… software?
That’s a complete institutional and leadership failure.
Ironically, building chips is the actual _hard_ part. The software and the compilers are not trivial but the iteration speed is almost infinite by comparison.
It goes to show that some companies just don’t “get” software. Not even AMD!
Re: HipKittens: Fast and furious AMD kernels
#10You'd think AMD would swing in on something like this and fund it with the money needed to succeed. I have no knowledge of it but my guess is no, AMD never misses an opportunity to miss an opportunity - when it comes to GPUs and AI.
It is now funded and working.