The Performance Cost Of Spectre / Meltdown / Foreshadow Mitigations On Linux 4.19

[Michael]

In addition to the relative values posted here, are the absolute values from each test also available?

Yep, http://openbenchmarking.org/result/1...WN82&obr_imw=y As always things are transparent with Phoronix Test Suite and OpenBenchmarking.org.

[captainohair]

In addition to the relative values posted here, are the absolute values from each test also available?

[geearf]

You should do a new test with the old intel-microcode 20171117.1, which was the last release without any spectre patches. The up-to-date microcode also lows the performance.

Yes that would be nice to also include in the tests with or without microcode fixes.

[davidbepo]

thanks for the benchmarking michael

[ipsirc]

You should do a new test with the old intel-microcode 20171117.1, which was the last release without any spectre patches. The up-to-date microcode also lows the performance.

[torsionbar28]

Thanks for doing these comparisons!
To me it confirms that AMD is the wiser choice :-)

For sure, especially in the datacenter. Low single-digit percent impact on AMD, with up to 20% impact on intel. Ouch. EPYC is shaping up to be a real winner for AMD.

[fhuberts]

Thanks for doing these comparisons!
To me it confirms that AMD is the wiser choice :-)

[Michael]

Typo:



Also, why are the results pictures rather than SVG graphs?

(DejaVu Sans is the worst font)

Fixed, thanks.

Working on some graph code/infrastructure improvements and while those WIP changes were being made, easier just doing this article as pictures.

[tildearrow]

Typo:

7~16% lower than an umitigated kernel

Also, why are the results pictures rather than SVG graphs?

(DejaVu Sans is the worst font)

[phoronix]

Phoronix: The Performance Cost Of Spectre / Meltdown / Foreshadow Mitigations On Linux 4.19

One of the most frequent test requests recently has been to look at the overall performance cost of Meltdown/Spectre mitigations on the latest Linux kernel and now with L1TF/Foreshadow work tossed into the mix. With the Linux 4.19 kernel that just kicked off development this month has been continued churn in the Spectre/Meltdown space, just not for x86_64 but also for POWER/s390/ARM where applicable. For getting an overall look at the performance impact of these mitigation techniques I tested three Intel Xeon systems and two AMD EPYC systems as well as a virtual machine on each side for seeing how the default Linux 4.19 kernel performance -- with relevant mitigations applied -- to that of an unmitigated kernel.

http://www.phoronix.com/vr.php?view=26809