Question Low GPU Clock and Memory Speed with HD8870M card

penton

Reputable
Jan 9, 2016
4
0
4,510
0
Here's my system, for starters—
LAPTOP Dell Inspiron 5737
core I7-4500U 1.8 GHz
Radeon HD8870M dGPU (detected by Windows as R9 M200X)
Intel HD4400 integrated GPU
8GB DDR3

The problem started a few months ago, whereby, at some points when playing, the screen would start to stutter with lower FPS than usual, while using the 20.4.2 Catalyst software.
After a small interval during which the card resumed behaving normally, I kept getting lower FPS in games more quickly and for a longer time.

While exploring this issue with GPU-Z, I've noticed that the GPU clock speed gets stuck at 400 MHz, and the memory clock speed gets stuck at 300 MHz, with no way to change it by loading games or by using FurMark. This happens when after using the GPU for a short time (~20 seconds) and when the GPU temperature is about 64 degrees Celsius. While the clock speed is at 400/300 MHz, the GPU cannot be used at it's maximum speed. After this “low clock-speed” state occurs, after a while when not using the GPU (~5 minutes), the clock speed changes to an even lower speed 300/150 Mhz, and the GPU can once again be used at its normal usage level of 775/1000 MHz, and the cycle with the frozen clock speed happens again.
I've had periods where my GPU could be used by games until it reached 92 degrees C, without it changing the clock speed. Curiously, more recently, if I leave the game open and do not interact with it, I get 775/1000 Mhz normal speed. As soon as I interact with a game object—and in multiple games for that matter—the GPU almost certainly, with a few exceptions, changes to 400/300 Mhz.

The following solutions have been tried:
  1. Restarting the computer and disabling and reenabling the dGPU
  2. Reinstalling the 20.4.2 driver after DDU cleanup
  3. Reinstalling the dGPU 20.4.2 driver in the usual sequence: install chipset driver → install integrated GPU driver → install dGPU driver to prevent fatal freezing during dGPU driver installation after DDU cleanup
  4. Reverting to the recommended manufacturer dGPU Catalyst driver version 15.201 from 2015 after DDU cleanup
  5. Reverting to a previous Catalyst driver version 18.1.1, which I've had for the longest time installed on my system, after DDU cleanup
  6. Tweaking various Gaming display options in the Catalyst game settings (such as resetting the shader cache, when I thought that it was a game issue)
  7. Modifying the PP_GPUPowerDownEnabled registry setting to disable it
  8. Installing MSI Afterburner to disable the Ultra-low Power State option (ULPS)
  9. Disabling hardware acceleration on Chrome.
  10. Disabling the integrated GPU.
Nothing has had any effect.

With this dGPU, I don't have access to power usage indicators or to the “PerfCap Reason” indicator.

What advice can you give me regarding this issue, since I think it's rather unlikely that the GPU is “terminally broken” because I could get it to heat up to 92 degrees without it shutting down, only until I interacted with the games in question (e.g. moved the mouse over a button or opened an in-game menu).

It seems to be a software issue, but I've ran out of ideas on how to deal with the problem.


Thanks!
 
Last edited:
Jun 29, 2020
25
1
35
0
I would test another GPU, I had a GTX 960 that I could benchmark and stress test the hell out of and it threw no errors and performed terrific, but then when randomly playing games or watching YouTube it would give me the blank screen and no signal - I went through so many hoops before I tried a spare GPU and the issues went away - the issue with hardware is it's not always obvious, even with looking at the logs and stress testing the GPU there was no obvious errors but somewhere along the card itself was an issue that would cause the problem, my point being because a GPU seems to be working fine under normal conditions - and even extraordinary conditions doesn't mean it's working properly, your best bet would be to test a new GPU if you have a spare one and see what that does.
 

penton

Reputable
Jan 9, 2016
4
0
4,510
0
I would test another GPU, I had a GTX 960 that I could benchmark and stress test the hell out of and it threw no errors and performed terrific, but then when randomly playing games or watching YouTube it would give me the blank screen and no signal - I went through so many hoops before I tried a spare GPU and the issues went away - the issue with hardware is it's not always obvious, even with looking at the logs and stress testing the GPU there was no obvious errors but somewhere along the card itself was an issue that would cause the problem, my point being because a GPU seems to be working fine under normal conditions - and even extraordinary conditions doesn't mean it's working properly, your best bet would be to test a new GPU if you have a spare one and see what that does.
Unfortunately, I don't have that option in my current laptop—I hadn't stated clearly that it's actually a laptop.
 

ASK THE COMMUNITY