Duration: 0:00

30 views

0

Six benchmarks I did, each after making a trivial change I could do by echoing values to files and one with enabling random chromium flags I thought were interesting.
I should mention that this 5600X machine is not mine and I usually use it via network control. I hardly get access to it's UEFI BIOS and I rather have u-boot over proprietary big tech UEFI. Since those desktops are mostly used for their GPU compute, with gaming and CPU being secondary and the CPU performance mostly used for caching. The memory is of course, DDR4, and it runs at 2666MHz, which is not the "best" this CPU/Mobo can support as it goes up to 3200MHz.
As for performance? I don't know at every time how the BIOS is configured, but maximum performance? Since it was complained we are spending $200 a month in power bills on the servers, I say probably not configured to maximize performance.

Now if I remember what was done (can't remember exact values).
1 was done with just switching governor to performance.
2 was done enabling Ozone and running Chromium on Wayland instead of X11.
3 locked the frequency of the cores to the max, made a bit of difference.
4 I switched the scheduler (I/O) to another one...
5 I triggered a few flags in Chromium, which led to a regression.
6 I reset the flags and stopped OpenSM (The infiniband router service) to see if that would increase scores.
Take it with a grain of salt, I'm sure most of the differences and effect is how quickly I can close my sunshine client so it stops using CPU resources after I click Benchmark.