Deep Learning & CUDA Benchmarks On The GeForce GTX 1080 Under Linux
Written by Michael Larabel in Graphics Cards on 11 June 2016. Page 2 of 3. 18 Comments

First up is the Caffe AlexNet deep learning benchmark.

To no surprise at all, the GeForce GTX 1080 Pascal was by far the fastest tested graphics card for running Caffe with CUDA 8.0-RC1 + cuDNN.... Quite a nice performance improvement over even the GTX TITAN X.

For those curious what the CPU-only performance is like with the Xeon E3-1280 v5 Skylake CPU used during testing for this same exact AlexNet workloads.... All of the tested NVIDIA GPUs are many times faster.

Similar to the SHOC OpenCL benchmarks, the SHOC CUDA results show very significant performance improvements for the GeForce GTX 1080 "Pascal" graphics card over Maxwell.

It will certainly be fun to see where the GTX 1070 fits in on Linux with our review next week.


Related Articles

Trending Linux News