3080 GPU Fans/Temperature Monitoring Going Crazy After Driver 576.02

[Kilroy747]

GPU: GIGABYTE GeForce RTX 3080 GAMING OC 12G

 

TLDR: Same issue JayzTwoCents had with his 5060TI driver, but on a 3080 with the 576.02.

 

 

 

After the 576.02 driver release, I've had issues of with GPU fans suddenly ramping to 100% for a couples seconds before going back down to idle.

Opened up Fan Control, HWinfo, and Task Manager  and oh boy does it look fun.

The spike in GPU activity in tasks manager was at 95.5 C, when the fans finally decided to activate at 100% and bring the temps down to ~81.2 C.

The temperature then slowly climbs back up to 95.5C until the fans decide to spike to 100% again. Rinse and repeat.

 

I recalibrated my GPU fans with Fan Control and am still getting the error seen above.

Anything under 38% fan speed results in a speed of 0 during manual control.

Temperature monitoring is obviously not working properly in Task manager or Fan Control.

Only HWinfo is giving me accurate results.

 

Troubleshooting:

I installed the drivers about a day ago, but this only started occuring after recently putting my computer to sleep and woke it back up. 

(So maybe it is that Win0 issue people were discussing, but I'm not sure?)
Have run MWB rootkit scan just in case, and have not found anything. Was half expecting a virus/crypto miner, but nothing.
Of note, I did recently set up a PiHole for my network, but I dont see how that would affect my local GPU temperature monitoring.

As stated before, recalibrated the GPU cans in Fan Control, but they were then met with a "?" next to the fan % indicator, before the indicator fell to 0.

Saw Jayz video and realized I had the same error on a 3080, not a 5060TI: 

 

While any normal person would use DDU and revert to a more stable set of drivers, I want to keep my computer in this state to see if there are any alternative solutions to try.

Open to any ideas/solutions people have to offer.

I dont have all the time in the world to troubleshoot, but will be open to trying suggestiosn when I have the time.

 

[DreamCat04]

11 minutes ago, Kilroy747 said:

GPU: GIGABYTE GeForce RTX 3080 GAMING OC 12G

 

TLDR: Same issue JayzTwoCents had with his 5060TI driver, but on a 3080 with the 576.02.

 

 

 

After the 576.02 driver release, I've had issues of with GPU fans suddenly ramping to 100% for a couples seconds before going back down to idle.

Opened up Fan Control, HWinfo, and Task Manager  and oh boy does it look fun.

The spike in GPU activity in tasks manager was at 95.5 C, when the fans finally decided to activate at 100% and bring the temps down to ~81.2 C.

The temperature then slowly climbs back up to 95.5C until the fans decide to spike to 100% again. Rinse and repeat.

 

I recalibrated my GPU fans with Fan Control and am still getting the error seen above.

Anything under 38% fan speed results in a speed of 0 during manual control.

Temperature monitoring is obviously not working properly in Task manager or Fan Control.

Only HWinfo is giving me accurate results.

 

Troubleshooting:

I installed the drivers about a day ago, but this only started occuring after recently putting my computer to sleep and woke it back up. 

(So maybe it is that Win0 issue people were discussing, but I'm not sure?)
Have run MWB rootkit scan just in case, and have not found anything. Was half expecting a virus/crypto miner, but nothing.
Of note, I did recently set up a PiHole for my network, but I dont see how that would affect my local GPU temperature monitoring.

As stated before, recalibrated the GPU cans in Fan Control, but they were then met with a "?" next to the fan % indicator, before the indicator fell to 0.

Saw Jayz video and realized I had the same error on a 3080, not a 5060TI: 

 

While any normal person would use DDU and revert to a more stable set of drivers, I want to keep my computer in this state to see if there are any alternative solutions to try.

Open to any ideas/solutions people have to offer.

I dont have all the time in the world to troubleshoot, but will be open to trying suggestiosn when I have the time.

Does the stock fan curve also not respond to temperatures? With the new drivers, if your temperatures freeze up from sleep or hibernation (or out of nowhere), the stock fan curves will still work perfectly fine while any custom fan curves will refuse to work

[rippy4500]

Use DDU and then just use version 566.36, 572.xx and beyond are inherently broken and there's nothing you can do. And in my experience, the absolute latest drivers aren't as important as people will have you believe so you're not missing out on much by using 566.

I've personally had issues with newer drivers, 572 caused black screen on startup occasionally, the gpu video decoding not working sometimes, and other weird random issues. Everyone seems to have different issues and NVIDIA is somehow making it worse.

[leclod]

Shut your computer off instead of sleep/hibernation

[Kilroy747]

30 minutes ago, leclod said:

Shut your computer off instead of sleep/hibernation

I left my computer unattended for about an hour and the display output had shut off on its own. The GPU fans were maxed out.
Turned the computer off/on and Fan Control/Task Manager are reading temperatures normally. Fan Control also had an update.
Fan Control can now control the GPU fans again, but when the fan % control drops below 38%, the GPU fans shut off entirely.

This is what my GPU fan curve is effectively stuck at:

 

It seems the driver is forcing 38% fan speed minimums or something?

 

[Kilroy747]

1 hour ago, DreamCat04 said:

Does the stock fan curve also not respond to temperatures? With the new drivers, if your temperatures freeze up from sleep or hibernation (or out of nowhere), the stock fan curves will still work perfectly fine while any custom fan curves will refuse to work

Stock Fan curves did appear to be functioning.

[porina]

https://nvidia.custhelp.com/app/answers/detail/a_id/5650

There's a hotfix driver update including a fix for the issue with incorrectly reported temperatures after resuming from sleep.

[Kilroy747]

30 minutes ago, porina said:

https://nvidia.custhelp.com/app/answers/detail/a_id/5650

There's a hotfix driver update including a fix for the issue with incorrectly reported temperatures after resuming from sleep.

Not for 30 series cards.

[Kilroy747]

1 hour ago, rippy4500 said:

Use DDU and then just use version 566.36, 572.xx and beyond are inherently broken and there's nothing you can do. And in my experience, the absolute latest drivers aren't as important as people will have you believe so you're not missing out on much by using 566.

I've personally had issues with newer drivers, 572 caused black screen on startup occasionally, the gpu video decoding not working sometimes, and other weird random issues. Everyone seems to have different issues and NVIDIA is somehow making it worse.

> "While any normal person would use DDU and revert to a more stable set of drivers, I want to keep my computer in this state to see if there are any alternative solutions to try. "

 

One of the reasosn I'm doing this up is that Jay's video mainly covered the 5060TI and 50 series. Nvidia's patch "fixing" this issue hasnt been released for any other series other then the 50 series. I think its worth bringing up that the issue is more widespread than initially reported and I'm happy to be a bit of a test bench in the meantime.

I do appreciate you including the last stable version of the drivers for anyone who needs to switch back.

[porina]

14 minutes ago, Kilroy747 said:

Not for 30 series cards.

Quote

GPU monitoring utilities may stop reporting the GPU temperature after PC wakes from sleep [5231307]

It is not limited to 50 series. As a hotfix driver it is not listed on the normal download page, only at the link I provided earlier.

[Kilroy747]

21 minutes ago, porina said:

It is not limited to 50 series. As a hotfix driver it is not listed on the normal download page, only at the link I provided earlier.

Apologies, was not aware that was how nvidia did things.

 

Installed it, noticed it froze all temperatue monitoring software except HWinfo @~60C

Rebooted and it resolved the temperature monitoring freeze.

Still have the same problems I had after the first reboot.

Manual fan curve below 38% fan speed drops fan speeds to 0 RPM, even after recalibration in Fan Control.

Edit
Patch does appaear to have solved the "Sleep" induced issues.

[leclod]

3 hours ago, Kilroy747 said:

It seems the driver is forcing 38% fan speed minimums or something?

I think so, the fans on my 3070 are stuck at 1000rpm minimum if I use a custom curve (which is fine by me)

Edited April 22 by leclod

[Kilroy747]

1 hour ago, leclod said:

I think so, the fans on my 3070 are stuck at 1000rpm minimum if I use a custom curve (which is fine by me)

Mine is starting to get weirder the more I look at it.

Going into Fan Control after automatic calibration:

At 100% RPM is ~3100

At "40%" RPM is ~450, which is ~15% of the max.

GPU is idling at ~60C

 

Here is what the Autocal points look like:

It semes like limiting the fans to 38% minimum before going to 0 RPM is causing issues with the calibration, and manual editing cant fix it.

What the values should be for reference:

 

Going to edit my curve to account for this new issue for the time being.

 

[leclod]

3 hours ago, Kilroy747 said:

At "40%" RPM is ~450, which is ~15% of the max.

If I understand you correctly, below 38% the fans stop and your card gets hot.

So can't you make a curve that stays above 38% (let's say above 1000rpm) ? which should be alright noisewise and temperaturewise

Edited April 22 by leclod

[Kilroy747]

3 hours ago, leclod said:

If I understand you correctly, below 38% the fans stop and your card gets hot.

So can't you make a curve that stays above 38% (let's say above 1000rpm) ? which should be alright noisewise and temperaturewise

This is indeed what I did.

Since % is meaningless, I based this curve off of approximate RPM values.

[rippy4500]

13 hours ago, Kilroy747 said:

> "While any normal person would use DDU and revert to a more stable set of drivers, I want to keep my computer in this state to see if there are any alternative solutions to try. "

 

One of the reasosn I'm doing this up is that Jay's video mainly covered the 5060TI and 50 series. Nvidia's patch "fixing" this issue hasnt been released for any other series other then the 50 series. I think its worth bringing up that the issue is more widespread than initially reported and I'm happy to be a bit of a test bench in the meantime.

I do appreciate you including the last stable version of the drivers for anyone who needs to switch back.

I meant it as a response to that sentence, it's just not worth trying to fix it IMO when you don't really lose anything important by using the older version.

Have you tried using MSI afterburner for fan control?

[Kilroy747]

Just thought I'd update this thread now that I've reverted back to 566.36 drivers with DDU.

Followed all steps that DDU recomends to ensure a clean install of 566.36.
The only exception exception was me unplugging the ethernet and wifi antenna only for the computer still wirelessly connecting to the network (should have checked and disconnected). Used safe mode, etc. to ensure the install was clean.

 

The GPU fan issue is still persisting.

Below 38%, the GPU drops the fan speed to 0% speeds in Fan Control.

I recalibrated multiple times, ensured Fan Control was up to date, but the issue still persists.

My adjusted fan curve that takes into account the actual fan RPM is still working thankfully.

Coworkers of mine have reported similar issues with the NZXT/Gigabyte fan control solutions on their personal rigs.

 

 

This was not normal behavior before the 576.02 update.

Its almost like they updated the gpu firmware or something?