VMWare reports performance regression up to -70% on Kernel 5.19 because of Retbleed patches - did anybody here experience this?

https://lkml.iu.edu/hypermail/linux/kernel/2209.1/02248.html

[…] we have evaluated the performance
of Linux kernel 5.19 against the 5.18 release and we have
noticed performance regressions in Linux VMs on ESXi

In other words: They have tested the performance in a virtual environment. Not sure if their findings can be translated to a regular PC running Linux natively.

I did a quick benchmark comparison with geekbench. I compared kernel 5.19 with 5.15. I did not want to install 5.18.
In this benchmark the difference between 5.19 and 5.15 is neglectable, within the error limits.

https://browser.geekbench.com/v5/cpu/compare/17234187?baseline=17234102

1 Like

I also tested Linux hosted in a vmware. And I do not see significant difference there either:

https://browser.geekbench.com/v5/cpu/compare/17235178?baseline=17235113

1 Like

They may use Intel CPU that has some security vulnerability.

Intel retbleed/IBRS commit: 16.29 secs (absolute diff ~2 secs)
Intel retbleed/IBRS commit: 23.824secs (absolute diff ~2 secs)

The Kernel prevents the vulnerability and dramatically decreases performance, that is my guess.