AMD EPYC 7F72 vs. Intel Xeon Gold 6258R - Latest EPYC Rome vs. Xeon Cascade Lake Benchmarks

[phoronix]

Phoronix: AMD EPYC 7F72 vs. Intel Xeon Gold 6258R - Latest EPYC Rome vs. Xeon Cascade Lake Benchmarks

Following the Xeon Gold 6250 vs. EPYC 7F32 benchmarks from earlier this month, here is a look at the latest x86_64 server CPUs we have our hands on with the EPYC 7F72 and Xeon Gold 6258R being benchmarked against a lineup of other competing AMD EPYC and Intel Xeon processors under the new Ubuntu 20.04 LTS.

http://www.phoronix.com/vr.php?view=29220

[wagaf]

Once again, awesome work, thanks :-)

It would be nice to be able to click on a link on the graph to go to a related openbenchmarking.org page..
Also the last graph has too many items to distinguish them well with colors, it would be great if it was possible to see the CPU name with a dynamic tooltip by hovering a curve.
Also it would be nice to be able to hide/show some results by clicking on the related legend, and (cherry on the cake)to order the result by brand then perf instead of just perf.

[Volta]

Benchmarks without full mitigations are meaningless. Intel's true performance is much lower (even 7% of vulnerable CPU without mitigations). Nobody sane will buy Intel or winblows 10 todays.

[Michael]

Once again, awesome work, thanks :-)

It would be nice to be able to click on a link on the graph to go to a related openbenchmarking.org page..
Also the last graph has too many items to distinguish them well with colors, it would be great if it was possible to see the CPU name with a dynamic tooltip by hovering a curve.
Also it would be nice to be able to hide/show some results by clicking on the related legend, and (cherry on the cake)to order the result by brand then perf instead of just perf.

Clicking on the OpenBenchmarking.org logo on the graphs should take you to the page.

There used to be tooltips but due to how the browsers are currently handling them when embedding SVGs cross-domain, they don't end up working anymore, especially with some browsers just treating the SVG as an image.

[Michael]

Benchmarks without full mitigations are meaningless. Intel's true performance is much lower (even 7% of vulnerable CPU without mitigations). Nobody sane will buy Intel or winblows 10 todays.

Default mitigations / out-of-the-box are always used for CPU reviews. Separate articles look at mitigation impact.

[tildearrow]

Typos:

With the PyPerformance Pytjhon benchmarks,

Meanwhile the EPYC Rome line-up remains incredibly strong with the the MrBayes molecular biology software package

[Volta]

Default mitigations / out-of-the-box are always used for CPU reviews. Separate articles look at mitigation impact.

I know, but it doesn't tell a thing about Intel performance which is mediocre with proper settings.

[JustRob]

Amazon (for example, and other Providers) offer to assist their users with performance issues resulting from security patches, and even offer a download for Amazon Linux 2 (which could be a better test platform than Ubuntu, disclosure: I'm Debian) for on-premise use under the most popular hypervisors: https://aws.amazon.com/amazon-linux-...n-premises_Use - along with permanently free instances (provided you meet their conditions).

Users do report excessive slowdowns with some applications, after patching. In some cases it's possible to run patch-free (especially on-premise) but in most cases you'll want to apply some of them, running naked obviously leaves you exposed; and doesn't make for quite an apples-to-apples comparison if one server can run securely and it's compared to another not set up similarly. It's no different than comparing two CPUs coupled to wildly different memory and cache sizes running at different speeds, declaring one a winner over the other.

[pkese]

The article is missing a table explaining:
1) CPU name, 2) number of cores, 3) frequency, 4) year of release (or architecture), 5) price
...and possibly memory speed
for each contestant.