The Performance Impact Of MDS / Zombieload Plus The Overall Cost Now Of Spectre/Meltdown/L1TF/MDS

[phoronix]

Phoronix: The Performance Impact Of MDS / Zombieload Plus The Overall Cost Now Of Spectre/Meltdown/L1TF/MDS

The past few days I've begun exploring the performance implications of the new Microarchitectural Data Sampling "MDS" vulnerabilities now known more commonly as Zombieload. As I shared in some initial results, there is a real performance hit to these mitigations. In this article are more MDS/Zombieload mitigation benchmarks on multiple systems as well as comparing the overall performance impact of the Meltdown/Spectre/Foreshadow/Zombieload mitigations on various Intel CPUs and also AMD CPUs where relevant.

http://www.phoronix.com/vr.php?view=27859

[Alex/AT]

Hello. I am extremely confused by the following picture: https://openbenchmarking.org/embed.p...ha=30a8bc6&p=2

The performance impact on 9900k in this test is noticeably bigger than in other tests, and given it has 8 cores instead of 6 that may be an interesting finding.

What I suspect is that buffer flushes require wait till full stop of the execution on all cores. If anyone could test this on some 14-28 core Xeon, it may confirm or deny this.

[Michael]

Hello. I am extremely confused by the following picture: https://openbenchmarking.org/embed.p...ha=30a8bc6&p=2

The performance impact on 9900k in this test is noticeably bigger than in other tests, and given it has 8 cores instead of 6 that may be an interesting finding.

What I suspect is that buffer flushes require wait till full stop of the execution on all cores. If anyone could test this on some 14-28 core Xeon, it may confirm or deny this.

I'll have large core Xeon results ~Monday.

[ChadD]

Thanks for the tests. One thing that has me thinking now... the impact on context switching. I wonder how badly it directly effects general multi tasking. Would be interesting to see a stress type test where a few different standard tests are run concurrently, just to see if the impact is greater then expected if multi tasking type context switching is going on.

[Alex/AT]

Thank you!

[andyprough]

I tried turning off hyperthreading on my laptop a few days ago. No noticeable difference so far. I'll compile a kernel this weekend and see if it's agonizingly slow, although from these benchmarks I don't expect to see a real difference.

[RussianNeuroMancer]

Thanks for PostgreSQL test! Would be interesting to how much all of this is impact virtualization.

[Xaero_Vincent]

Heavy context switching is a disaster for Intel chips now. Ugh.

[geearf]

Those 16% don't even account for Intel disabling TSX on some CPUs too (like mine) :/

[ThoreauHD]

Well, on the bright side, that only wipes out 2 years of performance gains. Could be worse.