3 ms·
The big feature is CUDA 9 and cuDNN 7 support, which promises double-speed training on Volta GPUs/FP16. (it should be noted that TF 1.5 does not support CUDA 9.
by minimaxir 9y ago
The big feature is CUDA 9 and cuDNN 7 support, which promises double-speed training on Volta GPUs/FP16. (it should be noted that TF 1.5 does not support CUDA 9.1 yet, which I found out the hard way)
I updated my Keras container with the TF 1.5 RC, CUDA 9, and cuDNN 7 (https://github.com/minimaxir/keras-cntk-docker https://github.com/minimaxir/keras-cntk-docker), but did not notice a significant speed increase on a K80 GPU (I'm unsure if Keras makes use of FP16 yet either).
- jorgemf 9y agoSpeed in FP16 under k80 is almost as FP32, the architecture doesn't work well with FP16. As you said, you need Volta to notice improvements with FP16. The other two main features are: Eager execution and TensorFlow Lite
- make3 9y agok80 is pretty old now, 1080ti have x5 perf of k80 for one tenth of the price
- minimaxir 9y agoWhen I say K80, I mean running a GPU in the cloud, not running it locally (and as an aside, due to the crypto boom, buying a physical GPU for cheap is more difficult)
- make3 9y agooh yeah. what i meant was that as k80 s are old, they don't have the fancy features like being fast with fp16, which came later. k80 come from slightly before the deep learning explosion afaik
- zeptomu 9y agoA nice compromise is the GTX1080 that you can rent at Hetzner for 100e/month [0]. I am not sure how it compares to other GPUs but for experimenting it works quite well. [0] https://www.hetzner.de/dedicated-rootserver/matrix-ex https://www.hetzner.de/dedicated-rootserver/matrix-ex
- make3 9y agothat sounds really nice, thanks for the link
- make3 9y agoah, 100e. it's still really good
- TheAlchemist 9y agoIt's actually 118e/month (the price on the webiste are excluding VAT). I'm using it, and I can only recommend - 0 problems, pretty decent box for a low price. Great solution, and waiting for the crypto bubble to burst and GPU prices to fall !
- puzzle 9y agoYou need Pascal hardware or later for FP16. Amazon's G3 Maxwell instances, which are newer than K80s, don't support it, either.
- adrianmacneil 9y agoP3 instances have Volta (V100) GPUs.