I mean, silicon aside, isn't it all basically CUDA lock-in?
CUDA is the defacto parallel computing platform, and exciting tech moves VERY fast with early mover advantage - so no researcher\computer scientist is going to bother learning another platform or risk\waste time\effort on their stuff breaking when ported to OpenCL\AMD.
Unless a miracle happens, any new parallel-computing requirement that seizes the popular imagination in the next decade (or more) will done in CUDA.
The sad part is that most of the people are using a higher level API like PyTorch. So AMD or Intel just need a simple low level API to access their HW, and then write and tune some kernels for PyTorch (and I suspect the community or AI will gladly help with the tuning). So basically there does not seem to be a real CUDA lock-in, it's just the fact that the competition still seems unable to do even this.
Apple was able to break the CUDA lock-in with their Metal computing api (MPS), in no time at all. Within months, now the major AI libraries like Pytorch and Tensorflow all support Apple GPU without a hitch.
What's taking AMD so long? They just can't do software I guess?
> I mean, silicon aside, isn't it all basically CUDA lock-in?
Everything is and always will be some form of lock-in. Look at what’s happened with CentOS or even Ubuntu. Even what was solid open source choices before have turned into some form of exploited lock-in. I’ve come to the realization that it’s impossible to optimize out of lock-in, instead it’s better to cost optimize for current best practices while continuously running test applications on different platforms and technologies when you have the resources to do so.
The amount of money being spent on ML compute is pretty high. If it stays high, cost will be a relevant axis that others can compete on.
GPUs are also not super available atm. We'll see if this is a temporary issue or one that persists for years. If you can't get A100s, you'll try the AMD/TPU competitors.
CUDA is the defacto parallel computing platform, and exciting tech moves VERY fast with early mover advantage - so no researcher\computer scientist is going to bother learning another platform or risk\waste time\effort on their stuff breaking when ported to OpenCL\AMD.
Unless a miracle happens, any new parallel-computing requirement that seizes the popular imagination in the next decade (or more) will done in CUDA.