You are not logged in.

#1 2023-01-27 20:16:41

mauruco
Member
Registered: 2023-01-27
Posts: 2

bad unigine-heaven score after kernel upgrade

Hello guys.
After updating the kernel the score of my AMD Radeon RX 6650 XT on unigine-heaven dropped significantly.
If I revert the score back to normal. Can anyone help me? Is it a problem in the kernel itself or some new setting?
Thank you for your help.

Settings:
renice -n -20 -p "${unigine_pid}"
cpupower frequency-set --governor performance

Score:
Linux archmauruco 6.1.8-arch1-1 score = 4533
Linux archmauruco 6.1.6-arch1-1 score = 5812

It looks like the system isn't even trying to reach 3.7GHz performance as reported by /sys/devices/system/cpu/cpufreq/policy*/scaling_max_freq

~ $ uname -a
Linux archmauruco 6.1.8-arch1-1 #1 SMP PREEMPT_DYNAMIC Tue, 24 Jan 2023 21:07:04 +0000 x86_64 GNU/Linux

~ $ cpupower frequency-info
analyzing CPU 0:
  driver: acpi-cpufreq
  CPUs which run at the same hardware frequency: 0
  CPUs which need to have their frequency coordinated by software: 0
  maximum transition latency:  Cannot determine or is not supported.
  hardware limits: 1.40 GHz - 4.31 GHz
  available frequency steps:  3.70 GHz, 1.70 GHz, 1.40 GHz
  available cpufreq governors: conservative ondemand userspace powersave performance schedutil
  current policy: frequency should be within 1.40 GHz and 3.70 GHz.
                  The governor "ondemand" may decide which speed to use
                  within this range.
  current CPU frequency: Unable to call hardware
  current CPU frequency: 1.37 GHz (asserted by call to kernel)
  boost state support:
    Supported: yes
    Active: no

~ $ sudo cpupower idle-info
CPUidle driver: acpi_idle
CPUidle governor: menu
analyzing CPU 0:

Number of idle states: 4
Available idle states: POLL C1 C2 C3
POLL:
Flags/Description: CPUIDLE CORE POLL IDLE
Latency: 0
Usage: 1910
Duration: 55569
C1:
Flags/Description: ACPI FFH MWAIT 0x0
Latency: 1
Usage: 184536
Duration: 7758546
C2:
Flags/Description: ACPI IOPORT 0x414
Latency: 18
Usage: 104634
Duration: 31685382
C3:
Flags/Description: ACPI IOPORT 0x415
Latency: 350
Usage: 118221
Duration: 334142810

~ $ cat `/sys/devices/system/cpu/cpufreq/policy*/scaling_max_freq `
3700000
3700000
3700000
3700000
3700000
3700000
3700000
3700000
3700000
3700000
3700000
3700000

# governor ondemand
~ $ watch
Every 2.0s: cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq /sys/d...  archmauruco: Mon Jan 30 17:23:45 2023

1696802
1696789
1696793
1696629
1493130
1211852
1390812
1696753
1696723
1304575
1211959
1696800

# governor performance
~ $ watch
Every 2.0s: cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq /sys/devices/system/cpu/cpu10/cpufreq/scaling_cur_freq /sys/devices/system/cpu/cpu11/cpufreq/scaling_cur_freq /sys/devices/sys...  archmauruco: Mon Jan 30 18:03:42 2023

3665214
3700000
3700000
2994209
3700000
3700000
2991181
3700000
3700000
3274200
2979311
2999213

Last edited by mauruco (2023-01-31 00:09:07)

Offline

#2 2023-02-06 16:21:25

mauruco
Member
Registered: 2023-01-27
Posts: 2

Re: bad unigine-heaven score after kernel upgrade

I wanted to leave my final opinion here and asked to close the thread.
I tested several benchmarks on several different kernel versions. There is indeed a difference from one to the other but they are minimal and are in most cases within the margin of error.
My initial assumption of it being the "CPU Governor" was wrong, the problem is unigine-heaven itself.
My mistake was relying on an old benchmark, maybe it would be good to remove them from the wiki or mark them as outdated.

Benchmarks that I found reliable at the moment are:
unixbench
glmark2
vkmark
unigine-superposition

Offline

#3 2023-02-06 19:11:39

ewaller
Administrator
From: Pasadena, CA
Registered: 2009-07-13
Posts: 19,739

Re: bad unigine-heaven score after kernel upgrade

Closed, by request,  In the future, you can make such requests using the report link of a thread -- it will draw the attention of a moderator rather than waiting for one of us to stumble over the request.  Also, you can use the report link should you want a thread reopened.


Nothing is too wonderful to be true, if it be consistent with the laws of nature -- Michael Faraday
Sometimes it is the people no one can imagine anything of who do the things no one can imagine. -- Alan Turing
---
How to Ask Questions the Smart Way

Offline

Board footer

Powered by FluxBB