Intel Announces Knights Mill: A Xeon Phi for Deep Learning

Post by DannyBee » Thu, Aug 18, 2016, 12:19 AM UTC

kartD wrote: post #8

cs702 wrote: post #2

We badly need an alternative to Nvidia/CUDA for deep learning... but realistically, if Intel wants to make headway in the deep learning market, it must offer hardware that can not only compete on performance with Nvidia, but also work out-of-the-box (that is, without requiring lots of one-off tinkering and tweaking) with popular deep/machine learning frameworks like TensorFlow, Caffe, Torch, and Theano. There is a lo…

I agree, AMD is working on making it easy to port CUDA[1], which still doesn't provide a strict alternative, but it's something http://wccftech.com/amd-cuda-compilercompatibility-layer-ann...

Post by zump » Thu, Aug 18, 2016, 12:23 AM UTC

kartD wrote: post #4

Nice, but how is this going to fit with Nervana? Also more than hardware how do Intel's libraries compare with CuDNN? At the end of the day ease of use and software support matter along with the hardware.

Post by valarauca1 » Thu, Aug 18, 2016, 12:36 AM UTC

DannyBee wrote: post #11

kartD wrote: post #8

Earlier quoted context omitted.

I agree, AMD is working on making it easy to port CUDA[1], which still doesn't provide a strict alternative, but it's something http://wccftech.com/amd-cuda-compilercompatibility-layer-ann...

Google also released GPUCC ( http://research.google.com/pubs/pub45226.html ), and got CUDA support into upstream clang. LLVM also has an AMD GPU backend, and says this thing is built on clang/llvm. So i suspect it's based on that support :)

Post by vonnik » Thu, Aug 18, 2016, 12:49 AM UTC

imaginenore wrote: post #12

ThinkBeat wrote: post #9

I wish they had a range that was affordable for a hobbyist. You can buy a cheap Nvidia card to "get your feet wet". I would like to play with these things.

Why not use Google's platform, which runs on their new custom chips (Tensor Processing Unit)?

Post by agibsonccc » Thu, Aug 18, 2016, 3:23 AM UTC

cs702 wrote: post #2

We badly need an alternative to Nvidia/CUDA for deep learning... but realistically, if Intel wants to make headway in the deep learning market, it must offer hardware that can not only compete on performance with Nvidia, but also work out-of-the-box (that is, without requiring lots of one-off tinkering and tweaking) with popular deep/machine learning frameworks like TensorFlow, Caffe, Torch, and Theano. There is a lo…