News for hackers. Stuff that matters.

A comment on Cavium Is an Arm Server Contender.

Re:Cavium Is an Arm Server Contender by Symmetry on Monday November 27, 2017 @05:47PM (#15792367)
I'd sort of been expecting that ARM would begin making HPC inroads when SVE[1] was ready to go. But I guess those two Neon units provide enough horsepower?

[1]https://community.arm.com/processors/b/blog/posts/technology...

[ Parent | on HN ]
Re:Cavium Is an Arm Server Contender by ajtulloch on Monday November 27, 2017 @09:38PM (#15793822)
They don't for fp32 or int8 workloads until SVE IMO. High end Skylake as an example has 2 AVX-512-F ports (and not considering AVX-512-4FMAPS), which give you do 64 FLOP/cycle (512 / 32 * 2 * 2). Even 2 NEON units, assuming FMA's execute in 1 cycle/unit, gives only 16 FLOP/cycle. For ML inference workloads you can even get away with lower precision and even do away with floating point (which helps motivate AVX-512-4VNNIW).
[ Parent | on HN ]
Re:Cavium Is an Arm Server Contender by jabl on Tuesday November 28, 2017 @02:23AM (#15795150)
From a HPC perspective, I think the current crop is interesting mostly [1] for tire kicking, software porting etc. Real production deployments will start to happen once SVE is available.

[1] Though they could be interesting for applications that are memory BW bound and don't vectorize that well.

[ Parent | on HN ]
Re:Cavium Is an Arm Server Contender by wmf on Monday November 27, 2017 @07:16PM (#15793087)
It seems like Intel/ARM processors are just needed to launch CUDA kernels these days.
[ Parent | on HN ]