[SOLVED] GPU is dropping to 0% usage for 3 seconds, and games stutter then come back again ?

[iStungu]

Hello brilliant minds everywhere would you please help me for that issue as i have the below specs ,

MOBO : MSI TOMAHAWK B550
CPU : RYZEN 9 3900XT @ stock speeds
COOLER : Noctua NH-U14S
GPU : Aorus Master RTX 3080 10 GB rev. 3
Memory : 32GB 16X2 T-FORCE @3200MHZ
SSD : 2X nvme gen 4 in both Pci-e slots (WD SN 750 SE)
HDD : 1X 5TB HDD
PSU : SEASONIC FOCUS 850w GOLD
Case : XPG Battlecruiser

Thermals are perfect
CPU @ idle is 29-35c
@Load is 57 - 65 c

Gpu @ is 33c
@Load is 78c

The problem is i had a 3070 before this setup with a psu Cooler master 750 w bronze , the pc while playing video games sometimes while browsing internet and even sometimes in desktop while doing some work the mouse pointer stutters for 3 seconds and i tracked gpu usage it reaches 0% for approx. 3 seconds or 4 seconds then comes back again in normal condition

and it happens randomly you can't predict it , some has advised me that i shouldn't rely on bronze PSU while running the 3070 with 3900XT so i changed the PSU to 850 a seasonic one and gold certified happened again then some told me the GPU might be faulty i bought the RTX 3080 since i wanted to upgrade anyways and sold my 3070 and to be honest the pc was working find for first 3 days like a charm nothing happened but today while playing SQUAD happened again it's like a stutter not a freeze like frames drop to a 2 fps and the whole gpu usage drops to 0% for 3 Seconds

I have installed a fresh windows 10 updated to latest updates
Uninstalled GPU drivers via DDU and reinstalled drivers again
changed cooler of cpu which was water cooled by the horrible MSI MAG 240 and now it's cooled by Noctua NH-U14S
changed PSU to a golden certified one
changed GPU "thought the old RTX 3070 from zotac had a problem of undervolting it self "
now i don't really know what to do can someone please tell me is there any other way that i can fix this irritating problem.

Thanks.

 

[Murissokah]

Hi, iStungu!

The GPU reaching 0% might be a symptom and not a cause. Maybe something else is stalling, like CPU/Mem or disk and because of it the GPU has nothing to process. I would start by collecting metrics, try using perfmon (Windows Performance Monitor) to track some relevant stats, like:

CPU Usage (% processor time)
Memory (Committed bytes %)
Physical Disk (Current Queue Length)

There are also GPU metrics you can track there.

What we want to know here is whether some other anomalous behaviour happens simultaneously with the GPU usage drop.

I find it unlikely that the issue will be related to power since this would affect stability more than anything else. The idea that something will underperform or have "bottlenecks" because of power doesn't make much sense, it either has the power within parameters or it doesn't. A bad power suppy won't be able to sustain its output parameters if the load is too high, meaning you'll have voltage fluctuations and power-offs from either the power supply itself or the protection circuitry on the motherboard and other components.

Another thing that migh be helpful is using GPU-Z to look at the gpu sensors when the issue happens. You'll be able to see how the power parameters are on the VGA and "perfcap" will tell you if the GPU is being held back for any reason.

 

[iStungu]

the thing is i have tracked this incident more than once and i found that power consumption falls drastically when this happens means the GPU is undervolting for some reason and the incident it self happens very very randomly not on a frequent or fixed pattern that's why i could be waiting for maybe couple of days to track this issue down again , but i posted this issue here hoping this issue would be common to you guys and maybe there are few suspects we could investigate .

But i guess i will keep investigating using Perfmon and Gpu-z and i will post any result might be useful to you guys.

 

[Ralston18]

And also remember Resource Monitor and Task Manager. Use both tools but only one at a time.

Process Explorer (Microsoft, free) may also help narrow down the problem.

https://docs.microsoft.com/en-us/sysinternals/downloads/process-explorer

 

[Murissokah]

the thing is i have tracked this incident more than once and i found that power consumption falls drastically when this happens means the GPU is undervolting for some reason and the incident it self happens very very randomly not on a frequent or fixed pattern that's why i could be waiting for maybe couple of days to track this issue down again , but i posted this issue here hoping this issue would be common to you guys and maybe there are few suspects we could investigate .

But i guess i will keep investigating using Perfmon and Gpu-z and i will post any result might be useful to you guys.

I see. Doesn't seem like any of the more obvious issues I've had before. GPU-z will be your best bet using perfcap to determine if the GPU is having an issue. I'm still not convinced it's a power issue because it makes sense for the GPU to drop voltage and power if it's waiting for something. If it helps, my own 3080 idles at 0.73V and about 9% TDP. Another thought that occurred to me is that your PCI-E bandwidth might be getting saturated, so maybe test for that too. You could run some graphics intensive load while running a disk stress test simultaneously.

 

[iStungu]

I see. Doesn't seem like any of the more obvious issues I've had before. GPU-z will be your best bet using perfcap to determine if the GPU is having an issue. I'm still not convinced it's a power issue because it makes sense for the GPU to drop voltage and power if it's waiting for something. If it helps, my own 3080 idles at 0.73V and about 9% TDP. Another thought that occurred to me is that your PCI-E bandwidth might be getting saturated, so maybe test for that too. You could run some graphics intensive load while running a disk stress test simultaneously.

Today i was playing SQUAD and suddenly for 3 seconds it happened , luckily i got it on GPU-Z and that's what happened please proceed to the link for viewing the image

https://ibb.co/XJNFv8n

 

[Murissokah]

Yeah, notice how perfcap reason is grey. The GPU was idling, that's why everything throttles down. Doesn't look like a power issue or a GPU issue. Did you leave perfmon running as well? If not, please do, I bet something else is happening when it throttles down.

 

[iStungu]

ok i will leave perfmon open next time but could you estimate what is giving it the throttle down , i mean what part would cause such mess ?

 

[Murissokah]

ok i will leave perfmon open next time but could you estimate what is giving it the throttle down , i mean what part would cause such mess ?

The GPU idling means it has nothing to do, so the easiest and more obvious place to look at is CPU. Check your cpu usage for each individual core and see anything anomalous happens when the hiccup occurs. By anomalous I mean a sudden increase in usage. I don't think this is the most likely cause, but it needs to be ruled out before looking at anything else.

Just to be clear, I'm not suspecting your CPU "can't handle" the load, but sometimes the way software uses CPU resources can cause stalls.

I would then look at the discs, as two PCI-E 4 NVME SSDs can use a lot of bandwidth. You could do the test I mentioned, run a game then saturate both your disks with I/O to see if anything happens.

Keep an eye on memory too, just to be sure. No particular reason to think the issue is there, but a memory hiccup could cause those symptoms too. I'd capture at least these measurements in perfmon:

CPU Usage (% processor time)
Memory (Committed bytes %)
Physical Disk (Current Queue Length)

 

[iStungu]

The GPU idling means it has nothing to do, so the easiest and more obvious place to look at is CPU. Check your cpu usage for each individual core and see anything anomalous happens when the hiccup occurs. By anomalous I mean a sudden increase in usage. I don't think this is the most likely cause, but it needs to be ruled out before looking at anything else.

Just to be clear, I'm not suspecting your CPU "can't handle" the load, but sometimes the way software uses CPU resources can cause stalls.

I would then look at the discs, as two PCI-E 4 NVME SSDs can use a lot of bandwidth. You could do the test I mentioned, run a game then saturate both your disks with I/O to see if anything happens.

Keep an eye on memory too, just to be sure. No particular reason to think the issue is there, but a memory hiccup could cause those symptoms too. I'd capture at least these measurements in perfmon:

CPU Usage (% processor time)
Memory (Committed bytes %)
Physical Disk (Current Queue Length)

Thanks a lot for your support i will monitor as you advised and how may i saturate both nvme ? i don't know what that is

 

[Murissokah]

Thanks a lot for your support i will monitor as you advised and how may i saturate both nvme ? i don't know what that is

I meant your PCI-E SSDs, NVME is just the protocol they use. You could run AS-SSD on both disks while the game is running, this should be enough to saturate them. If it ends too fast choose a larger file size in the test.

 

[iStungu]

i did a test for both drives but couldn't do them at the same time though so here are some results while playing

https://ibb.co/pZy0Cmp

image hosted at ImgBB

Image image hosted in ImgBB

ibb.co

These results are not related to the problem it was a smooth experience

 

[Murissokah]

i did a test for both drives but couldn't do them at the same time though so here are some results while playing

https://ibb.co/pZy0Cmp

image hosted at ImgBB

Image image hosted in ImgBB

ibb.co

These results are not related to the problem it was a smooth experience

When you were running the tests, were you also running a graphics intensive load? The idea is to check if there's contention on the PCI-E bus, the disk test alone is not expected to cause issues. Ideally, run a GPU intensive game, alt+tab, run AS-SSD, alt+tab back and continue playing.

 

[iStungu]

When you were running the tests, were you also running a graphics intensive load? The idea is to check if there's contention on the PCI-E bus, the disk test alone is not expected to cause issues. Ideally, run a GPU intensive game, alt+tab, run AS-SSD, alt+tab back and continue playing.

i was playing SQUAD at Max settings , Yes

 

[iStungu]

after testing and monitoring the results i couldn't find any anomaly in hardware monitored but the problem still exist what could be the issue here guys ?

 

[ayushgarg82]

Hey bro m facing the same issue did u find any solution? Pls tell

 

[iStungu]

Hey bro m facing the same issue did u find any solution? Pls tell

Actually i changed my psu into a 850 watt gold seasonic and changed my thermal paste of my cpu then issue has been solved but i doubt what i did solved the problem i dunno