Benchmarking AMD FX vs. Intel Sandy/Ivy Bridge CPUs Following Spectre, Meltdown, L1TF, Zombieload

While a synthetic test, the Hackbench test developed by Facebook for stressing the Linux kernel's scheduler and socket/pipe performance is one of the most profoundly impacted tests from the CPU mitigations we've seen. The scheduler performance on the Core i5 CPUs went from completing this test quicker than the AMD FX CPUs to slower, even with those CPUs not having Hyper Threading. The Core i7 2700K when losing HT also was slower than the AMD FX CPUs. The Core i7 3770K Ivybridge system took 2.3x longer to run this test under the complete mitigations or 1.6x the time if just using the default mitigations.

The Perl interpreted performance has been slightly slower we've seen multiple times as a result of these mitigations while the AMD performance was unchanged.

Perl can actually perform better if HT is disabled to avoid the chances it's stuck running on a virtual core.

Many GNU C Library (Glibc) operations do run slower as a result of the Spectre / Meltdown / Foreshadow / Zombieload mitigations while the AMD Spectre impact was minimal.

The Intel mitigations do make some of these Glibc operations now slower than the AMD Vishera CPUs.