An Important Note On The Alder Lake Mobile Power/Performance With Linux 5.19

Xonotic saw its performance nosedive on Linux 5.19 across multiple settings and runs, but it wasn't limited to just the graphics regressing.

Fortunately when comparing the kernels, I was also monitoring the CPU SoC power consumption exposed by the PowerCap sysfs interface. That led to some big insight for what's going on with Linux 5.19... The CPU power consumption in the affected workloads is much lower on 5.19.

But this Linux 5.19 regression wasn't universal / across-the-board but some workloads were performing similar or slightly faster between Linux 5.18 and 5.19 Git. In cases where the performance was unchanged, the CPU power efficiency tended to be slightly better than 5.18.

With glmark2 for example the raw performance was unchanged on Linux 5.19 but now experiencing less fluctuation and peaks in the Core i7 1280P power consumption.

The same went for VKMark with seeing the same performance between these kernels but at least the performance-per-Watt is in better shape with less fluctuation under load.

The CPU temperature of the i7-1280P within this MSI notebook tended to show no measurable difference between kernels.

Meanwhile, even the WireGuard VPN tunnel performance regressed significantly on Linux 5.19...

But it too showed how the CPU power consumption/behavior was strikingly different between the kernels.

Let's continue...