Loss performance when resuming from long sleep

thepenguin · 2021-08-06 09:24:08

Hi there!
So, i need to make this sort of a long post, so you guys can have a better understanding on what the heck is going on here. English is not my main language, so bare with me.

issue: When resuming from a long sleep (i would say probably 1 hour at least), the laptop slows down (noticeable on gaming). The fix is rebooting everytime (or never suspend the laptop)

System info: https://pastebin.com/Fz90MEuU
Laptop: Hp Omen 15
OS: updated Arch, with kde
nvidia drivers: 470.57.02
Kernel: 5.10.56-1-lts (also tried the current kernel)

I tried many distros where this issue happens. Including: Pop!_OS nvidia version, ubuntu...etc and i figured i should try Arch to better understand what's going on. The DE doesn't matter too

Relevant info:

- Currently undervolted my CPU with "undervolt" (to avoid hitting 95ºC+ temps)
- Also underclocked to limit at 3.1ghz with "tlp"
- Using feral gamemode to launch games
- If i close the lid and wait for it to suspend, then minutes later wake it up.. the issue most of the times won't occur. I figured this happens in longer sleeps.

I should also inform you, i tried disabling the above utilities, the results are the same.
Right now, as i'm typing, the laptop is in that state. If i launch a game, it's as slow as if the intel turbo boost wasn't turning ON. The problem is, i check frequency, and it's clocking at 31.ghz ...

I checked system monitor, nothing unusual, i check the logs, can't find anything. I'm lost here.

This is a log from resuming today (from overnight) and trying to launch a game: https://pastebin.com/dXDdgFDZ (used journalctl --since "2 hours ago" to generate.
Another curious fact: When trying to play the game in this state, the temperatures are low. This usually is an indicator of the cpu not being forced to do much work (like unplugged from the charger).

What i did try to fix this:

- Disabled all the utilities described above.
- Tried to force the cpu to prioritize the game process (didn't work, and shouldn't be needed anyway)
- Tried to unplug/plug the charger ( the FPS remain exactly the same, which puzzles me)
- After rebooting i try the same experiment: plugged: fps normal. unplugged: low fps as expected.
- Tried 2 different games, to exclude being game related
- Tried using openGL instead of vulkan, to try to isolate if this is a vulkan exclusive problem.

Honestly, my linux knowledge ends here, i don't know what else to do to try to understand this issue that occurs across distros.

Last edited by thepenguin (2021-08-06 09:30:13)

ewaller · 2021-08-06 17:36:36

thepenguin wrote:

English is not my main language, so bare with me.

Bare --> naked ; bear --> an animal, or to carry or endure.

issue: When resuming from a long sleep (i would say probably 1 hour at least), the laptop slows down (noticeable on gaming). The fix is rebooting everytime (or never suspend the laptop)
- Currently undervolted my CPU with "undervolt" (to avoid hitting 95ºC+ temps)

It was not clear whether you tried this without undervolting. These are odd observations, so am am just thinking out loud here. Is it possible that as the system cools down it becomes more sensitive to under voltage -- and perhaps is having issues keeping memory properly refreshed. If you have not tried removing the under volt condition, you might try.

thepenguin · 2021-08-06 18:57:03

ewaller wrote:

thepenguin wrote:
English is not my main language, so bare with me.
Bare --> naked ; bear --> an animal, or to carry or endure.
issue: When resuming from a long sleep (i would say probably 1 hour at least), the laptop slows down (noticeable on gaming). The fix is rebooting everytime (or never suspend the laptop)
- Currently undervolted my CPU with "undervolt" (to avoid hitting 95ºC+ temps)
It was not clear whether you tried this without undervolting. These are odd observations, so am am just thinking out loud here. Is it possible that as the system cools down it becomes more sensitive to under voltage -- and perhaps is having issues keeping memory properly refreshed. If you have not tried removing the under volt condition, you might try.

Hi! Thanks for replying (and sorry about the bare situation eheh)

I also thought about that, so i tried with a lower undervolt value. Didn't work either.
Since i'm dealing with this problem for so many time, i guess i already tested with no undervolt on my experiments. However, since i'm not sure, i'll try it again.
I have to say however, that this doesn't make much sense, because if i don't suspend the laptop...i can play all day ( and i do play in my days off work ), with no issues. So the temps are cool during lunch break , and doesn't affect in any way when i restart playing.

Today while i was waiting for a reply here, i added the "Nomodeset" to my kernel parameters, following the arch wiki instructions. Then i went to play a bit, all good. Then i suspend while taking a bath, about 15 min later tried again, and worked fine. This is just insane, maybe it really needs to sleep more time for a proper test

Last edited by thepenguin (2021-08-06 19:05:07)

progandy · 2021-08-06 19:25:02

thepenguin wrote:

- Tried to unplug/plug the charger ( the FPS remain exactly the same, which puzzles me)

That would have been my suggestion. My laptop likes to get stuck on the lowest cpu frequency after suspending with the charger plugged in and reconnecting it fixes that.

You could look at the pstate settings with e.g. pstate-frequency (AUR).

Last edited by progandy (2021-08-06 19:28:03)

seth · 2021-08-06 19:47:25

Get undervol/tlp/gamemoded etc. out of the way during the tests (less variables, easier deduction ;-)

Have you established that it's about the CPU w/ an isolated stress/performance test?
https://wiki.archlinux.org/title/Stress_testing#stress
https://wiki.archlinux.org/title/Benchmarking#peakperf

If i close the lid and wait for it to suspend, then minutes later wake it up.. the issue most of the times won't occur. I figured this happens in longer sleeps.

Time based decay is a known issue the the nvidia VRAM contents, but idk. how that would directly impact performance…
https://download.nvidia.com/XFree86/Lin … ement.html

thepenguin · 2021-08-06 20:05:58

progandy wrote:

thepenguin wrote:
- Tried to unplug/plug the charger ( the FPS remain exactly the same, which puzzles me)
That would have been my suggestion. My laptop likes to get stuck on the lowest cpu frequency after suspending with the charger plugged in and reconnecting it fixes that.
You could look at the pstate settings with e.g. pstate-frequency (AUR).

Yeah, when i unplugg the frequency drops to 2.2ghz , replugging increases to 3.2ghz , i watch it while the game is minimized. That should impact the FPS, but it doesn't.
That alone make me think about the issue being related to nvidia? because if something was wrong with the clock or with the turbo, the frequency wouldn't be able to hit 3.2ghz

thepenguin · 2021-08-06 20:06:38

seth wrote:

Get undervol/tlp/gamemoded etc. out of the way during the tests (less variables, easier deduction ;-)
Have you established that it's about the CPU w/ an isolated stress/performance test?
https://wiki.archlinux.org/title/Stress_testing#stress
https://wiki.archlinux.org/title/Benchmarking#peakperf
If i close the lid and wait for it to suspend, then minutes later wake it up.. the issue most of the times won't occur. I figured this happens in longer sleeps.
Time based decay is a known issue the the nvidia VRAM contents, but idk. how that would directly impact performance…
https://download.nvidia.com/XFree86/Lin … ement.html

Thanks Seth, some material there to work with. As soon as i have a bit of time, i'll test that. cheers

Arch Linux

#1 2021-08-06 09:24:08

Loss performance when resuming from long sleep

#2 2021-08-06 17:36:36

Re: Loss performance when resuming from long sleep

#3 2021-08-06 18:57:03

Re: Loss performance when resuming from long sleep

#4 2021-08-06 19:25:02

Re: Loss performance when resuming from long sleep

#5 2021-08-06 19:47:25

Re: Loss performance when resuming from long sleep

#6 2021-08-06 20:05:58

Re: Loss performance when resuming from long sleep

#7 2021-08-06 20:06:38

Re: Loss performance when resuming from long sleep

Board footer