Intel Core i9 10980XE Linux Performance Benchmarks

[birdie]

And no Ryzen 9 3950X in tests? What a bummer.

[Michael]

Michael thanks for the benchmarks. Just a question for clarification on Core i9 10980XE, you say hardware mitigations for l1tf, meltdown and spectre. But what about MDS specifically. As the Intel Xeon E-2288G Coffee Lake managed to get hardware mitigations for l1tf, meltdown, spectre and MDS as well.

Errr yeah MDS too, forgot to mention it, but as shown on the system table inside the article: "- Core i9 10980XE: itlb_multihit: KVM: Mitigation of Split huge pages + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Enhanced IBRS IBPB: conditional RSB filling + tsx_async_abort: Mitigation of TSX disabled"

[eva2000]

All mitigations are enabled at their defaults. The 400+ is only when TSX is enabled and therefore TAA mitigations active, but TSX isn't enabled by default with the Linux 5.4 kernel

Michael thanks for the benchmarks. Just a question for clarification on Core i9 10980XE, you say hardware mitigations for l1tf, meltdown and spectre. But what about MDS specifically. As the Intel Xeon E-2288G Coffee Lake managed to get hardware mitigations for l1tf, meltdown, spectre and MDS as well.

[nuetzel]

And what about Intel's MKL tampering against AMD and others?

https://www.computerbase.de/2019-11/mkl-workaround-erhoeht-leistung-auf-amd-ryzen/

https://en.wikipedia.org/wiki/Math_Kernel_Library

Performance
[-]
This has netted the system a nickname of "cripple AMD" routine since 2009.[10] As of 2019, MKL, which remains the choice of many pre-compiled Mathematical applications on Windows (such as NumPy, SymPy, and MATLAB), still significantly underperforms on AMD CPUs with equivalent instruction sets.[11]

[oibaf]

Thanks for the clarification.

[Michael]

Are the latest mitigations enabled?

This graph shows a value of 120:


But Cascade Lake with latest mitigations enabled should have a much higher value of over 400, as found in latest Phoronix test:

All mitigations are enabled at their defaults. The 400+ is only when TSX is enabled and therefore TAA mitigations active, but TSX isn't enabled by default with the Linux 5.4 kernel

[oibaf]

Are the latest mitigations enabled?

This graph shows a value of 120:


But Cascade Lake with latest mitigations enabled should have a much higher value of over 400, as found in latest Phoronix test:

[mumar1]

For a very good and obvious reason Intel does not want to see their LGA 2066 HEDT-CPUs directly put up against TR 3XXX in a test at release time. For 6!!! hours Intel can sell their stuff without being compared to competition.

[programmerjake]

With the GraphicsMagick imaging benchmarks, the Core i9 10980XE still comes up short of the 2970WX due to it being a 16 core vs. 24 core race albeit at close price points.

Isn't the 10980XE 18 cores?

[Michael]

Michael
Please do the right thing and update this article by including TR 3000 numbers once AMD's embargo expires (a separate article on TR 3000 notwithstanding).

Will we see Windows vs. Linux benchmarks on 10980XE vs. 3960X vs. 3970X? Would be interesting to see if the Microsoft scheduler's NUMA issues have gone away with the new Threadrippers.

kobblestown
Unfortunately, AMD sampling strategy on 3950X is very messed up. That AMD looks worse than they could is entirely their fault.

Windows vs. Linux 3970X numbers at least will be coming in a few days.