Intel's Latest CPU Microcode Update Isn't All That Scary
Two weeks ago Intel published new CPU microcode with mitigations for a few new processor vulnerabilities. As usual, I ran benchmarks testing the new microcode against prior revisions to see if these CPU vulnerability mitigations led to any measurable performance hit.
Across my testing on Xeon Scalable Cascade Lake, Skylake, Tiger Lake, and Rocket Lake the new CPU microcode was rather uneventful... It wasn't worth putting out a follow-up article of all the flat results across dozens of benchmarks.
There is though an article circulating today that Intel CPUs may have "slowed down" from this microcode. Cited is a particular micro-benchmark and the belief is the new CPU microcode disables the hardware zero store optimization.
While that particular micro-benchmark is impacted, just wanted to relay across dozens of benchmarks on multiple systems using a mix of real-world and synthetic kernel benchmarks, I am seeing no measurable change when switching to this latest microcode. That's when running many of the benchmarks where with prior kernel and microcode mitigations we have seen performance differences and other common workloads.
Here for example are some of the Core i9 11900K microcode comparison benchmarks with no real difference. That's in line with the comparable difference I am seeing on other recent Intel CPU generations.
I'll fire up the latest microcode on a few more systems, but at least from the testing since earlier this month, this has actually been quite a boring microcode update from the performance perspective for real-world tests compared to other mitigations in recent years.