AMD Claims World’s Fastest Per-Core Performance with New EPYC Rome 7Fx2 CPUs

Re:AMD Claims World’s Fastest Per-Core Performance with New ... by lliamander on Tuesday April 14, 2020 @09:09PM (#22873387)

Lots of gaming benchmarks out there, but sadly as of yet there is really no dedicated hardware reviews for the software engineering space.

Others have mentioned Phoronix (linux benchmarking) but I would also recommend Level1Techs. They are the only reviewers I've found that actually (occasionally) dig into what kind of performance actually matters to developers.

Recently, they did a video on using the Threadripper 3990X for Unreal Engine game development[0]. Some useful insights came out of that, such as:

- Compilers favor large cache sizes (which favors AMD)

- Usually the biggest performance bottleneck isn't compiling the code, but running automated tests (especially when it involves running a bunch of VMs).

[0] https://www.youtube.com/watch?v=VQa6r6Ci1jg

[ Parent | on HN ]

Re:AMD Claims World’s Fastest Per-Core Performance with New ... by DeathArrow on Wednesday April 15, 2020 @07:54AM (#22876684)

Here's a compilation benchmark: https://www.anandtech.com/bench/CPU-2019/2224

I guess there are more gaming benchmarks because there are much more gamers than software engineers.

[ Parent | on HN ]

Re:AMD Claims World’s Fastest Per-Core Performance with New ... by lliamander on Wednesday April 15, 2020 @12:44PM (#22879909)

Compilation benchmarks are fine, and many review sources are adding them, but there's still a pretty big gulf between "compiling the Linux kernel/chromium" and a typical developer workload.

> I guess there are more gaming benchmarks because there are much more gamers than software engineers.

That may be true, but I don't think that's a sufficient explanation. There are plenty of benchmarks for other professional workloads, such as CAD, 3D rendering, video editing, and mathematical modeling for financial or scientific applications.

[ Parent | on HN ]