AMD and OpenCL have fallen desperately behind NVidia and CUDA in terms of industry adoption. CIDA is so far the industry standard for machine learning on GPUs. For example tensor flow had a GPU backend using CUDA. Last time I looked OpenCL support was in alpha stages.
This software platform dominance fueled a lot of hype around NVidia, but I'm not sure how much it contributed to revenue. Research and specialized data centers may still be a small percentage of sales compared to mass market. But Intel and AMD both understand that if they don't catch up they will never recover. They are gearing up for a future where every customer device needs some efficient machine learning inference hardware. Then data centers will need to purchase mind boggling amounts of throughput computing for training from streams of incoming data. I'm sure each company is planning hardware derived from current GPUs but more power efficient for the type of number crunching ML algorithms.
This is all more or less true with the current public code. We're building our own standalone stack that will interop with TF, Keras, PyTorch, etc. Basically we are fixing the problem for AMD and Intel on desktop but also embedded chip vendors as well. Currently our stack runs Google's Xception net in Keras slightly faster than TF 1.2 + cuDNN 5.1 on Tesla K80 so there's good evidence OpenCL itself is not that slow.
This is an admirable effort and thanks for doing it.
Have you experienced problems related to the lack of active support from AMD or Intel for OpenCL drivers? Thats been my experience with OpenCL several years ago. On a related note drivers for OpenGL on Linux and Windows from AMD and Intel have been a huge problem for developer for years. Somehow only NVidia cares and invests enough in non-windows driver support. This may be changing with Vulkan however and more focus on GPU compute.
We have found bugs in both AMD and NVIDIA drivers but nothing that didn't get fixed or that we couldn't work around. More difficult are the "special" quirks in some embedded GPUs. I keep an eye on Vulkan etc but right now OpenCL looks like the right choice for portability.
This software platform dominance fueled a lot of hype around NVidia, but I'm not sure how much it contributed to revenue. Research and specialized data centers may still be a small percentage of sales compared to mass market. But Intel and AMD both understand that if they don't catch up they will never recover. They are gearing up for a future where every customer device needs some efficient machine learning inference hardware. Then data centers will need to purchase mind boggling amounts of throughput computing for training from streams of incoming data. I'm sure each company is planning hardware derived from current GPUs but more power efficient for the type of number crunching ML algorithms.