Graphics card stops when I run certain games and doesn't work unless I underclock it

Shuierel

Posts: 13   +0
So I have been having this problem for several weeks and I don't know what to do I tried everything.When I start a game like God of War or Warzone 2.0 it works for a short period, it freezes then the screen turns black cause by the graphic card stopping and then instantly booting back.I even went to my local PC store for a checkup but they were unable to reproduce the problem but when I got it back home it worked normally for a couple of hours then the problem occured again.Btw games only work if I underclock my GPU max frequency to 50%.The graphic card is an AMD Radeon RX 6650 XT.If anyone could I'd be very gratefull because I'm kinda getting drepessed by this and I just want it to end.I will also attach a video for a better understanding and I had to put the link to it in a txt file because its too big too upload here.Thank you.
 

Attachments

  • image_2022-12-30_034004000.png
    image_2022-12-30_034004000.png
    93.5 KB · Views: 7
  • img video link.txt
    27 bytes · Views: 2
This sounds like a power supply issue, but it could also be signs of motherboard troubles. Can you add the full details about your PC - I.e. CPU, RAM, motherboard, PSU, cooler, etc?
 
I'm kinda getting drepessed by this and I just want it to end.I will also attach a video for a better understanding and I had to put the link to it in a txt file because its too big too upload here.Thank you.
As @neeyik said, your issue is most likely with the PSU. Under clocking dramatically reduces the power consumption of of the card. Thus, no crash.

The classic example of this is where you have a card like the GTX-1650, with a reference draw of 75 watts and can get enough power from the Pce-e lane. . However, as soon as the GPU is over clocked, power draw exceeds 75 watts, and a 6 pin connector is required, since draw becomes greater than the Pci-e can supply.

The relationship between the PSU's quality, its maximum rated output, and its ability to supply at or near maximum constantly, is another question altogether. Nor are all 500 watt PSUs, "created equal", with respect to weathering transient power spikes above their rated output. Your 6900 has a fairly high power power draw, likely in comparison with the card it replaced. So, we're getting closer to the PSU's "rated output".

You wanted video? Here you go....\



This may, or may not, be the issue, but it's the first thing you have to rule out
 
Last edited:
Your 6900 has a fairly high power draw, likely in comparison with the card it replaced.
Your eyes are getting old - the OP has a 6650 XT. ;) That has a TDP of 176W, so it's not too high. However, you're absolutely correct about power spikes; TechPowerUp's analysis of an slightly overclocked version of that card showed some fairly significant spikes when running Furmark:

power-consumption.png


Now God of War probably isn't going to spiking as much as that (the above Gaming result was taken using Cyberpunk 2077) but without any details about the rest of the OP's system, it's not possible to be sure that it's just power that's the problem.
 
I'm sorry I didn't write about the rest of my sysyem.Here it is:
CPU : AMD Ryzen 5 5600X
PSU : CX650M 650W Bronze
Motherboard : GIGABYTE B550M
RAM : Corsair VENGEANCE LPX 2x8 GB 3200MHz
Cooler : ARCTIC AC Freezer 34 eSports DUO White
Its weird that I've never had this problem before it only started a month ago and I have this PC since May 2022.Anyway I'll wait for your response and if it truly is the PSU the problem I would really apreciate if you have any recommendations.Thanks for your time.
 
PSU : CX650M 650W Bronze
That's a perfectly decent PSU for your PC, so it's not like it couldn't cope with the power spikes. It could be faulty, of course, but I missed something in your original post -- the fact that you're getting an AMD Bug Reporting window pop open, when GoW crashed. What driver version are you using with your card?
 
Could you try a quick test? Instead of reducing the clock frequency to 50%, could you see what happens when you set the power limit to 50% instead? You can do this by enabling the Power Tuning option and setting the slider to the required limit.
 
Sorry for taking so long to respond.I just tried that and its doing the same thing screen goes black then the error pops up.
 
Well, it's almost certainly not a power thing then, although if you wanted to be absolutely sure, try setting the power limit to its lowest value and seeing if you still get the error.

Some other tests worth trying:

> If you are using a DisplayPort cable to connect the monitor to the graphics card, either try the cable in a different port or, and this would be a more useful test, try a different DP cable
> If you have another 8-pin power cable from the PSU, you could try connecting that one instead
> Download Furmark (link), go to Settings and enable logging, then run the standard GPU Stress Test. It will save the results of the logging in the installation folder (by default, this is C:\Program Files (x86)\Geeks3D\Benchmarks\FurMark) as a CSV file. Assuming that Furmark causes the same error, it should log the GPU temp just before the point that the program crashes.
> If you know anyone that has a PC with a PSU similar to yours, you could always test in their machine to see if the same problems occur.
 
Okay well I will try the Furmark test and I'll post the results, as for the displayport test I don't know if I should do it if have HDMI cable and I don't have another 8-pin power cable but the one that I have, has another set of 8-pin connectors and I tried using them but its the same.Anyway thanks for the advice and I have one other question sometimes my mouse cursor look weird I don't know how to explain but I'll attach a photo.
 

Attachments

  • IMG_7749.jpg
    IMG_7749.jpg
    877 KB · Views: 1
Update : I installed Furmark and I did the test but nothing happened it didn't crash or anything it just went on for several minutes until I closed it.
 
Possible GPU overheat?
Poor airflow in the case?
Does the Radeon software measure GPU heat?
Possible bad thermal pad on the GPU?

Since this ran for 2 hours "normally", heat becomes a suspect.
 
Well I don't think the GPU overheats because it never reaches 70 degrees , and the airflow is good I got 6 other smaller fans 3 for pumping air in and 3 for taking it out and about the thermal pad on the GPU I can't verify or change it because its still in warranty I wouldn't want to lose it.And yes the software measures GPU heat.Btw I actually found another 8-pin power cable and changed it and even changed the PCle slot on the PSU but it still crashes.
 
Update : I installed Furmark and I did the test but nothing happened it didn't crash or anything it just went on for several minutes until I closed it.
Here are the results.
 

Attachments

  • image_2022-12-31_163404156.png
    image_2022-12-31_163404156.png
    21 KB · Views: 6
Okay well I will try the Furmark test and I'll post the results, as for the displayport test I don't know if I should do it if have HDMI cable and I don't have another 8-pin power cable but the one that I have, has another set of 8-pin connectors and I tried using them but its the same.Anyway thanks for the advice and I have one other question sometimes my mouse cursor look weird I don't know how to explain but I'll attach a photo.
Sorry, I should have asked first if you were using DP or HDMI. It's highly unlikely to be anything to do with the HDMI cable.

Are those white dots behind the cursor some kind of an active trail, I.e. do they seem to follow the cursor about the screen, like an animation, or is it a permanently fixed set of dots?

Graphics cards displaying odd dots, funny colors, etc is often a sign of problems with the memory chips on the card.

Update : I installed Furmark and I did the test but nothing happened it didn't crash or anything it just went on for several minutes until I closed it.
Okay, so it's definitely not a power issue, as the Furmark test will hit near-maximum power consumption almost instantly.
 
The dots are permanent and they are not always there, I don’t know what seems to causes them to appear but I’ve noticed these dots are more frequent when a game is opened.

If it is not a power problem the only option remains is the GPU and if you said that dots indicate towards memory chip problem.What might have caused this and is the problem a physical one like literally in GPU or is it software one?
 
The dots are permanent and they are not always there, I don’t know what seems to causes them to appear but I’ve noticed these dots are more frequent when a game is opened.
Well, that makes sense, since games are the most demanding programs you likely use on the machine. Just out out of curiosity, are the dots more prevalent on more demanding games?
What might have caused this and is the problem a physical one like literally in GPU or is it software one?
Which software are we talking? If it's specific to GoW, it's a possibility, albeit a remote one. It worked before, unless you've modded it, it should (tentatively), still be working. An errant Windows update is a possibility, assuming this problem started at, or about, the time of one. Software would be easy to rule out, but only if you, (or a friend), have a card of comparable performance "laying around", to substitute it with. Hopefully it would be an AMD card, as you are much less likely to introduce issues from needing a driver swap to Nvidia for the test.

Let's assume the card is faulty. And let's assume you haven't overclocked or over-volted beyond it's tolerance. If you have, I'd keep that to myself, live with the guilt, and start an RMA with Radeon anyway. You did say you got the machine in May, 2022. Again assuming the card was new, you're well within the warranty period.

A bad memory chip sounds as reasonable as anything else we've considered. I might explain why the card works with reduced clock speed. A few balky cells might not be triggering at speed.

That's all I got.

And a very Happy New Year to both of you. :) (y) (Y)
 
Last edited:
It is not specific to GoW , it also does the same to Warzone 2.0, Satisfactory, it is weird that other games like Elden Ring or Nfs Heat work perfectly without underclocking.
I have also been thinking to send it to RMA since I started having this issue but I didn't do it because I'm afraid they wont be able to reproduce the problem I have, as I was saying at the the beginning of the post:
I even went to my local PC store for a checkup but they were unable to reproduce the problem but when I got it back home it worked normally for a couple of hours then the problem occured again.
but its very weird that it worked a few hours normally and I can't understand how and no I didn't overclock or over-volted :).About the dots, yes I do think so they are more prevalent on more demanding games.
 
OK. I've had problems with goofy artifacts on the cursor, and the cursor not tracking and freezing, both due to lack of (system) memory, and bad system RAM. These were far differently configured systems than yours, so I'm hesitant to draw a direct comparison. However, I've never experienced shutdowns from it

Lack of RAM certainly isn't your problem. It wouldn't hurt to run Memtest86 or your rig, if only to exclude the possibility that the system RAM is causing the problem. If it passes, it might at least give you the confidence that the card is the issue.

How long did the store have the PC? If you can summon up the willpower, leave the system sit for an equivalent amount of time, then try to duplicate the 2 hours "ran perfect effect". If you've left the system off for longer than the store had it, I don't know what to say. It was an act of god, maybe? :confused:
 
I've did the test and it passed so its not the RAM and the store had the pc for maybe 12 hours something like that more or less but I remember a time when I forgot to underclock it before opened warzone and like before it worked for some time but then it crashed and I had to underclock it once again. I was thinking about static electricity in my PC and maybe with time it dissipates and it works until it builds up agian, actually I don't even know if that makes sense.
 
The dots around the cursor are almost certainly an indicator of trouble with the VRAM -- the memory controllers for the graphics card's RAM are built into the GPU, and part of them run at the GPU clock rate. If there's an option to decrease the VRAM clock rate too, it might be worth experimenting with lowering that value to see if (a) the cursor dots disappear and (b) games such as GoW are more stable.

If nothing affects the cursor dots, then it's probably time to look at making a warranty claim on the card.
 
Yes there is VRAM tuning but it doesn't allow me to go below 100%, I'll send a picture.
 

Attachments

  • image_2023-01-01_190321027.png
    image_2023-01-01_190321027.png
    20.1 KB · Views: 1
  • image_2023-01-01_190331328.png
    image_2023-01-01_190331328.png
    11.2 KB · Views: 1
Back