I use current TW on a Lenovo P72 laptop with a Core i7-8850H CPU.
Some time ago, I decided it to compare Geekbench scores between kernel with (this is the default) and without vulnerability mitigations disabled.
To disable mitigations, I used these kernel parameters:
noibrs noibpb nopti https://browser.geekbench.com/v4/cpu/compare/12738676?baseline=12738264 nospectre_v1
I was astonished to see that the single-core score mitigated was 4579 vs 5843 the score with mitigations disabled, resulting in 21.6% performance loss ! Multi-core performance is unaffected.
I finally isolated to nospectre_v2, which if specified is causing that huge bump in performance. So the spectre v2 mitigation is causing that huge single core performance loss.
Geekbench comparison: default kernel vs kernel with nospectre_v2: https://browser.geekbench.com/v4/cpu/compare/12738751?baseline=12738676
Today, I tried Geekbench on Fedora 30 beta that got recently released. Without changing any kernel parameter it has the same performance than TW with nospectre_v2 !:
TW default vs Fedora default: https://browser.geekbench.com/v4/cpu/compare/12738751?baseline=12738264
TW nospectre_v2 vs Fedora default https://browser.geekbench.com/v4/cpu/compare/12738676?baseline=12738264
So basically, TW has about 20% less single core performance by default, which in my opinion is some kind of bug. This seems to me a good candidate to explain why TW trailing in Phoronix recent benchmarks. And maybe an indication that there may be a performance issue with spectre v2 mitigation only in TW.
For reference, here is the output of spectre-metldown-checker.sh on all systems compared:
TW default: https://termbin.com/8p7z TW nospectre_v2: https://termbin.com/bb6t Fedora default: http://termbin.com/0u7o