On Fri, Jul 29, 2016 at 2:07 AM, Michael Di Domenico
<[email protected]> wrote:
> but in all seriousness, if someone wanted some unbiased benchmarks run
> on something i have or have access to, i would do it.  that was really
> all my original point was.

I don't have published results for the GTX 1080, but take a look at
the spec results I pointed to for the K40 vs s9150. That tells a basic
story, but is however a narrow set of codes. Before SC16 I expect we
may have some Pascal and updated AMD results..

For machine learning, visual and other benchmarks - I have some ideas,
but I'm curious what others would recommend.

The challenge is really how to do best 1:1 comparisons.. The spec
benchmarks are OpenACC for both AMD and NVIDIA - other benchmarks tend
to be OpenCL or CUDA, which doesn't have equal support on both sides
and makes comparisons less clear/fair.
_______________________________________________
Beowulf mailing list, [email protected] sponsored by Penguin Computing
To change your subscription (digest mode or unsubscribe) visit 
http://www.beowulf.org/mailman/listinfo/beowulf

Reply via email to