GPU utilisation randomly drops and so does the fps

[Brickmazing]

Setup- Ryzen 5 5600x paired with a radeon 7800xt on a msi b550m-A pro motherboard, 32gb of kingston fury cl16 ram in dual channel and Msi mag a750gl psu. 
I have been using it for a month now and the gpu utilisation drop and fps drop is more for hogwarts legacy (gpu utillisation drops to 20%) but it still drops to 60-70% in other games like forza horizon 5 and Gta5. And idk why but when I play cs2 for more than 15 mins my pc just freezes and I have to restart it to get it working again. However cinebench and furmark run fine always without any issues or throttling.

I tried turning the re-bar off and it worked for some time till the problems showed up again. I have tried overclocking, undervolting the gpu but it has no effect. And this problem has been there since the beginning of my build. My temps are fine and always around 65 degrees for both cpu and gpu.

My main suspect is the motherboard as it has just a single 4 pin header for the cpu power and not the traditional 8 pin. Also there is no heatsink for the vrms.

[podkall]

8 hours ago, Brickmazing said:

My main suspect is the motherboard as it has just a single 4 pin header for the cpu power and not the traditional 8 pin. Also there is no heatsink for the vrms

If it was the motherboard though, Cinebench couldn't get throttled much.

 

Does your 5600x get similar Cinebench scores as other? Because if yes then there's really not much throttling.

 

As far as 4pin power goes, it's actually enough for 5600x because that is a 65-105W CPU, more on the lower end of those 65watts.

 

 

As far as motherboard VRMs go, sometimes they can run cold enough even without heatsink, if there isn't too much power heating them up, there isn't too much need for that heat to get quickly dissipated as the heat naturally escapes through the case's airflow.

 

Since Furmark and Cinebench pass it could potentially be some stability problem and for that you could try OCCT test.
 

It can test different kinds of issues, PSU, GPU VRAM, and power, but most importantly the VRAM and others that Furmark doesn't completely saturate.

 

https://www.ocbase.com/

[Brickmazing]

1 hour ago, podkall said:

If it was the motherboard though, Cinebench couldn't get throttled much.

 

Does your 5600x get similar Cinebench scores as other? Because if yes then there's really not much throttling.

 

As far as 4pin power goes, it's actually enough for 5600x because that is a 65-105W CPU, more on the lower end of those 65watts.

 

 

As far as motherboard VRMs go, sometimes they can run cold enough even without heatsink, if there isn't too much power heating them up, there isn't too much need for that heat to get quickly dissipated as the heat naturally escapes through the case's airflow.

 

Since Furmark and Cinebench pass it could potentially be some stability problem and for that you could try OCCT test.
 

It can test different kinds of issues, PSU, GPU VRAM, and power, but most importantly the VRAM and others that Furmark doesn't completely saturate.

 

https://www.ocbase.com/

Sure, I was actually planning on testing by build for stability issues. Someone had suggested using hwinfo64 to me for monitoring the components while gaming. 
So, I guess I will test out my system and try to reach a conclusion by ruling out the good components one by one. 

 

[podkall]

2 minutes ago, Brickmazing said:

Sure, I was actually planning on testing by build for stability issues. Someone had suggested using hwinfo64 to me for monitoring the components while gaming. 
So, I guess I will test out my system and try to reach a conclusion by ruling out the good components one by one. 

 

hwinfo64 is very good

[Brickmazing]

On 4/27/2025 at 2:50 AM, podkall said:

hwinfo64 is very good

Update-

I used hwinfo64 to collect some sensor data.

During the time my gpu utilisation drops down, the cpu core clock drops from 4.4 to 3.8 Ghz. Also the CPU TDC drops at the same time. The GPU clock fluctuates a lot too between 3800 and 4400 Mhz but that does not correspond to the utilisation drop. This is what I could infer from hwinfo.

 

[podkall]

2 hours ago, Brickmazing said:

Update-

I used hwinfo64 to collect some sensor data.

During the time my gpu utilisation drops down, the cpu core clock drops from 4.4 to 3.8 Ghz. Also the CPU TDC drops at the same time. The GPU clock fluctuates a lot too between 3800 and 4400 Mhz but that does not correspond to the utilisation drop. This is what I could infer from hwinfo.

 

And temps?

[Brickmazing]

5 hours ago, podkall said:

And temps?

Temps are all fine.

gpu max at 70

cpu max at 80-85 when tested by occt under 100% load.

Also another weird update- GTA5 crashes 10 mins into the game. My screen goes black and gpu fans go into full blast. I have to turn off my pc manually. And when I power it on my display drivers are somehow not compatible with the amd adrenalin software and I am stuck at 1080p resolution and cannot change it.

So I use DDU to uninstall the drivers and reinstall them again. After that the display is fine but only till I run GTA5 again and it does the same damn thing.

 

[podkall]

18 minutes ago, Brickmazing said:

gpu max at 70

even hotspot temp?

 

18 minutes ago, Brickmazing said:

Also another weird update- GTA5 crashes 10 mins into the game. My screen goes black and gpu fans go into full blast. I have to turn off my pc manually. And when I power it on my display drivers are somehow not compatible with the amd adrenalin software and I am stuck at 1080p resolution and cannot change it.

So I use DDU to uninstall the drivers and reinstall them again. After that the display is fine but only till I run GTA5 again and it does the same damn thing.

Sounds like overheating, but the drivers are bit weird.

[Brickmazing]

40 minutes ago, podkall said:

even hotspot temp?

 

Sounds like overheating, but the drivers are bit weird.

Hotspot is a but higher around 85.

And yeah, the driver issue is weird.

Was playing cs2, the gpu temps were fine this time but after 20 mins, the game crashed, black screen, gpu fans on full speed and I had to manually reboot my pc.

But get this- No driver issue this time.

I mean this is a new gpu, could be a bad unit but that’s not very common right?

[podkall]

5 hours ago, Brickmazing said:

I mean this is a new gpu, could be a bad unit but that’s not very common right?

Yeah, perhaps bad GPU, can you RMA/return it for replacement/cashback?

[Brickmazing]

On 4/29/2025 at 7:59 AM, podkall said:

Yeah, perhaps bad GPU, can you RMA/return it for replacement/cashback?

Update- 

I ran the system out of the case and it did not crash even a single time. I kept xmp and re-bar disabled.

Gta5 did pretty well with 99-100% gpu load on 1440 max settings.

Hogwarts dropped gpu load to 40% in some areas with the stuttering and the cpu load, cpu voltage dropped down too during the same time.

So I reckon it’s not a gpu issue.

Even the cpu and gpu temps were great.

cpu at 70deg max and gpu hotspot at 70 again.

Even inside the case the temps did jot cross 76 that is including the gpu hotspot.

Maybe a ram issue?

For now keeping the xmp disabled seems to provide more stability.

[podkall]

10 hours ago, Brickmazing said:

Update- 

I ran the system out of the case and it did not crash even a single time. I kept xmp and re-bar disabled.

Gta5 did pretty well with 99-100% gpu load on 1440 max settings.

Hogwarts dropped gpu load to 40% in some areas with the stuttering and the cpu load, cpu voltage dropped down too during the same time.

So I reckon it’s not a gpu issue.

Even the cpu and gpu temps were great.

cpu at 70deg max and gpu hotspot at 70 again.

Even inside the case the temps did jot cross 76 that is including the gpu hotspot.

Maybe a ram issue?

For now keeping the xmp disabled seems to provide more stability.

So OCCT doesn't give any errors or problems?

 

Do have a spare USB drive? https://www.memtest86.com/

[Brickmazing]

18 hours ago, podkall said:

So OCCT doesn't give any errors or problems?

 

Do have a spare USB drive? https://www.memtest86.com/

Thanks 

Will try that.

Till now what I can say is it might not be a particular hardware at fault.

Maybe just a compatibility issue or bios issue that is making my pc crash.

Might as well be windows11.

[podkall]

5 hours ago, Brickmazing said:

Thanks 

Will try that.

Till now what I can say is it might not be a particular hardware at fault.

Maybe just a compatibility issue or bios issue that is making my pc crash.

Might as well be windows11.

Freezes can be software issue, but of course they can also be hardware related.

 

Was your PC always having problems, or just last month, if your build is more than month old?

[Brickmazing]

On 5/1/2025 at 11:40 PM, podkall said:

Freezes can be software issue, but of course they can also be hardware related.

 

Was your PC always having problems, or just last month, if your build is more than month old?

Well it had problems from the start like gpu utilisation drops in hogwarts and screen freezing mid game in cs2 but no crashes.

Recently after playing gta it crashed for the first time.

The performance also seems to have taken a hit and it works way worse than before.

Now I have to run everything at stock including the ram which obviously has affected the performance.

Although lighter old games and games like forza horizon run fine but the fps has dropped due to my ram being at default and not at xmp.

[podkall]

28 minutes ago, Brickmazing said:

Well it had problems from the start like gpu utilisation drops in hogwarts and screen freezing mid game in cs2 but no crashes.

Recently after playing gta it crashed for the first time.

The performance also seems to have taken a hit and it works way worse than before.

Now I have to run everything at stock including the ram which obviously has affected the performance.

Although lighter old games and games like forza horizon run fine but the fps has dropped due to my ram being at default and not at xmp.

Then that's hard, because these problems could be from anything, since the entire PC was having problems from the start.

If you have USB, that memtest86 could help see if it's RAM.

 

Alternatively you could just try running 1 stick to see if it's the other RAM then swapping sticks if the problems stay or disappear after taking out one of the sticks.

[Brickmazing]

On 5/3/2025 at 2:13 AM, podkall said:

Then that's hard, because these problems could be from anything, since the entire PC was having problems from the start.

If you have USB, that memtest86 could help see if it's RAM.

 

Alternatively you could just try running 1 stick to see if it's the other RAM then swapping sticks if the problems stay or disappear after taking out one of the sticks.

So I ram memtest overnight and the ram passed.

A peculiar thing I noticed is that my motherboard has a Bios version which was not listed on the msi official support page and had a build date newer than the last bios update released by msi.

I got this board new so I was expecting maybe an outdated bios but this is strange.

Should I flash the last Bios update provided on the online support page or should I keep this one?

I am a bit scared of bricking the board as I live in an area which has power cuts sometimes, especially during summer.

 

[podkall]

8 hours ago, Brickmazing said:

So I ram memtest overnight and the ram passed.

A peculiar thing I noticed is that my motherboard has a Bios version which was not listed on the msi official support page and had a build date newer than the last bios update released by msi.

I got this board new so I was expecting maybe an outdated bios but this is strange.

Should I flash the last Bios update provided on the online support page or should I keep this one?

I am a bit scared of bricking the board as I live in an area which has power cuts sometimes, especially during summer.

 

If you're concerned about power cuts, maybe it's best to try everything else first before updating BIOS.