AMD drivers crashing

[CuelmsiV]

Maybe this will be helpful but now on old drivers (version 18) and turned down core and memory clocks as @Visherasaid the game i wasn't able to play because it would crash drivers works now. To be fair i lowered graphic settings to the lowest possible, but that is still something. Maybe we can go from here and figure out what is really happening. At this point i'm starting to belive my GPU may be half dead.

[Vishera]

5 hours ago, CuelmsiV said:

Maybe this will be helpful but now on old drivers (version 18) and turned down core and memory clocks as @Visherasaid the game i wasn't able to play because it would crash drivers works now. To be fair i lowered graphic settings to the lowest possible, but that is still something. Maybe we can go from here and figure out what is really happening. At this point i'm starting to belive my GPU may be half dead.

What is the GPU usage under load?

Power consumption under load?

And what power supply do you have?

[CuelmsiV]

19 minutes ago, Vishera said:

What is the GPU usage under load?

Power consumption under load?

And what power supply do you have?

I can't really tell you what exact GPU usage is under load and power consumption, because every time i lanuch some kind of benchmark or game that puts GPU under stress display goes dark and i have to restart. As to my power supply it is SilentiumPC Vero M1 600W. Though GPU usage spikes pretty often to 80-100% when launching some games and watching streams but it doesn't crash then. The point is that it spikes pretty often, don't know what the reason is.

[Vishera]

2 minutes ago, CuelmsiV said:

I can't really tell you what exact GPU usage is under load and power consumption, because every time i lanuch some kind of benchmark or game that puts GPU under stress display goes dark and i have to restart. As to my power supply it is SilentiumPC Vero M1 600W. Though GPU usage spikes pretty often to 80-100% when launching some games and watching streams but it doesn't crash then. The point is that it spikes pretty often, don't know what the reason is.

Is everything working like it should be with the downclock and the older drivers?

[Mark Kaine]

6 hours ago, CuelmsiV said:

Maybe this will be helpful but now on old drivers (version 18) and turned down core and memory clocks as @Visherasaid the game i wasn't able to play because it would crash drivers works now

 

9 minutes ago, Vishera said:

Is everything working like it should be with the downclock and the older drivers?

Edit:  misunderstood,  looks like yes? 

 

 

Tbh, I think it's either psu or gpu at this point? 

 

The psu sounds quite dodgy (never heard of it lol)

[CuelmsiV]

46 minutes ago, Vishera said:

Is everything working like it should be with the downclock and the older drivers?

Sadly no, the GPUPI still crashes the GPU and display goes dark when reduction size is set to 256. FurMark as well crashes it the second it is turned on. I don't know about all of the games since i lowered settings in all of them in order to play them. GPU works quietly when i'm playing them and it doesn't crash. So it may be something with working under load?

 

EDIT: Should i go even lower on the memory and core clocks?

[CuelmsiV]

31 minutes ago, Mark Kaine said:

 

Edit:  misunderstood,  looks like yes? 

 

 

Tbh, I think it's either psu or gpu at this point? 

 

The psu sounds quite dodgy (never heard of it lol)

I mean that was the first thing i suspected and checked it, the voltage on cables that were connected to GPU and PSU itself. And everything seemed to be working fine. Last time i got this issue it stopped occurring when i pulled the GPU out, switched slots for RAM and put GPU back in. It was working for quite a while after i did that. But now i did that as first thing and that didn't solve it.

[Vishera]

8 minutes ago, CuelmsiV said:

EDIT: Should i go even lower on the memory and core clocks?

Yeah,it's a good idea.

 

Do you have other GPUs to test with?

To be honest it doesn't look like a driver issue,but a hardware fault.

 

[Mark Kaine]

13 minutes ago, CuelmsiV said:

I mean that was the first thing i suspected and checked it, the voltage on cables that were connected to GPU and PSU itself. And everything seemed to be working fine. Last time i got this issue it stopped occurring when i pulled the GPU out, switched slots for RAM and put GPU back in. It was working for quite a while after i did that. But now i did that as first thing and that didn't solve it.

Well, set a power limit of 80 or so...  

and also keep the clocks lowered or lower them even more perhaps. 

 

If it's the psu it should at some point stop crashing... 

 

 

[CuelmsiV]

17 minutes ago, Vishera said:

Yeah,it's a good idea.

 

Do you have other GPUs to test with?

To be honest it doesn't look like a driver issue,but a hardware fault.

 

How much lower should i go? I already went -200 on core and -100 on memory as you said previously? Should i go the same amount again?

 

I have one other gpu, it is a gtx 1080, but it is a weird one without the slots for PSU cables from what i know this is some kind of mini GPU i don't really know what that is, sorry.

Yeah, that's what i was afraid of. 

[CuelmsiV]

16 minutes ago, Mark Kaine said:

Well, set a power limit of 80 or so...

How do i do that in Afterburner?

[Vishera]

33 minutes ago, CuelmsiV said:

How much lower should i go? I already went -200 on core and -100 on memory as you said previously? Should i go the same amount again?

 

Yes,do the same.

[Mark Kaine]

1 hour ago, CuelmsiV said:

How do i do that in Afterburner?

Just set it and click apply 

"Power limit"

 

[CuelmsiV]

8 hours ago, Mark Kaine said:

Just set it and click apply 

"Power limit"

In my Afterburner i can only go +50 or -50 on the power limit

[CuelmsiV]

8 hours ago, Vishera said:

Yes,do the same.

So after underclocking it -400 on core and -200 on memory GPU seems to be somewhat stable now. GPUPI with reduction size set to 256 didn't crash it, FurMark as well didn't crash it. So all is left for is to just test how high i can go backup with clocks until it starts to crash again? How much underclocking will affect the performance of the GPU? I do 3D renders and simulations so i'm wondering how much will it be affected.

[Mark Kaine]

27 minutes ago, CuelmsiV said:

So after underclocking it -400 on core and -200 on memory GPU seems to be somewhat stable now. GPUPI with reduction size set to 256 didn't crash it, FurMark as well didn't crash it. So all is left for is to just test how high i can go backup with clocks until it starts to crash again? How much underclocking will affect the performance of the GPU? I do 3D renders and simulations so i'm wondering how much will it be affected.

That,to me, sounds like indeed a PSU issue... i mean if it was the card artifacting and other weird stuff would be more likely then crashing...and what you effectively did by underclocking is make the card use less power,  and suddenly it's "ok" .... for now. 

 

I think you'll lose a lot performance with this thou, and it probably won't last, whatever is bad now will keep getting worse. I'd definitely try a new psu, that's literally a security concern as well,  this psu could blow up any time (if it's the psu, which, again,  is super likely)

 

"Crash under load" is like 95% chance psu,  cause gpus naturally spike, and that's when psu gives in...

 

It's actually quite common with cheap  / noname psus.  

[Vishera]

1 hour ago, CuelmsiV said:

So all is left for is to just test how high i can go backup with clocks until it starts to crash again?

There is no need to do that anymore,it seems like it's either the GPU itself or the power supply.

1 hour ago, CuelmsiV said:

 How much underclocking will affect the performance of the GPU?

Underclocking heavily impacts performance,for the worse.

You can change the clocks now to higher clcoks,you can use the -200,-100 if you want.

[CuelmsiV]

3 hours ago, Mark Kaine said:

That,to me, sounds like indeed a PSU issue... i mean if it was the card artifacting and other weird stuff would be more likely then crashing...and what you effectively did by underclocking is make the card use less power,  and suddenly it's "ok" .... for now.

The thing is there were some screen artifacts after some crashes. Display went black but then went back on but with some minor artifacts on screen. And these crashes dont turn the PC off it seems like just the GPU is affected, because rest of the PC is still on i can hear fans and all. So it really may be a PSU issue even with artifacts sometimes appearing? If that is the case it is so much easier for me to just buy new PSU than new GPU. 

 

EDIT: Is it possible that power input on GPU can be in some way malfunctioning, because i was checking voltage with voltmeter on PSU cables and PSU itself and it seemed fine and in check with values written on the PSU itself, that is why i crossed PSU from my suspect list when i first started having this issue.

[CuelmsiV]

2 hours ago, Vishera said:

There is no need to do that anymore,it seems like it's either the GPU itself or the power supply.

Underclocking heavily impacts performance,for the worse.

You can change the clocks now to higher clcoks,you can use the -200,-100 if you want.

I will try to turn clocks up a bit more as you are suggesting. I really hope that the PSU is the malfunctioning one here. 

[Vishera]

4 hours ago, CuelmsiV said:

I will try to turn clocks up a bit more as you are suggesting. I really hope that the PSU is the malfunctioning one here. 

Try resetting the underclock settings,then set a -50 power limit.

[CuelmsiV]

15 minutes ago, Vishera said:

Try resetting the underclock settings,then set a -50 power limit.

After doing that (reseting clocks and setting power limit to -50) i run GPUPI and FurMark again and it didn't crash, it took a lot longer to calculate baches in GPUPI and in FurMark it was a bit laggy, but i guess that is the effect of lowering the power limit. So is it safe to assume now that the PSU is the one causing all of this?

 

EDIT: Or is it still possible that the power connectors on GPU are in some way damaged?

[Vishera]

7 minutes ago, CuelmsiV said:

After doing that (reseting clocks and setting power limit to -50) i run GPUPI and FurMark again and it didn't crash, it took a lot longer to calculate baches in GPUPI and in FurMark it was a bit laggy, but i guess that is the effect of lowering the power limit. So is it safe to assume now that the PSU is the one causing all of this?

 

EDIT: Or is it still possible that the power connectors on GPU are in some way damaged?

It does seem like a power delivery issue,it could be in the graphics card itself (a failed capacitor for example) or the power supply.

You did that with stock clock speed,right?

Now try to raise the power limit  until you reach a good balance of performance/stability.

[CuelmsiV]

13 minutes ago, Vishera said:

You did that with stock clock speed,right?

Yes

 

13 minutes ago, Vishera said:

Now try to raise the power limit  until you reach a good balance of performance/stability.

Stock Clocks & Power -25% - works
Stock Clocks & Power -10% - works
Stock Clocks & Power -5% - GPUPI works, when FurMark is lanuched display starts to flash between being black and being on

So after doing tests with -5% power limit, and then redoing -10% and -25% for good measure FurMark was glitching wierdly and display started to flash between being black and being on. But after redoing tests once again after a while with -25% and -10% power limit it was working good. So it seems that -10% is fairly stable

 

18 minutes ago, Vishera said:

It does seem like a power delivery issue,it could be in the graphics card itself (a failed capacitor for example) or the power supply.

I can't really test if the GPU has failed capacitors or something without actually getting a new PSU and trying it out do i? Because i don't really have spare ones and would need to buy a new one to just test it.

[CuelmsiV]

One more question though @Vishera. If it really is the GPU fault like failed capactior or something like you mentioned and right now getting a new decent GPU is nearly impossible, how long can i run on settings like these (i mean lowering power limit to -10% or even more for safety) is it like a safety hazard for the rest of PC, that i don't know, something might blow up after some time? Or am i good to run it with limited power limit and clock set to stock values until i get a new GPU.

 

I will have to buy new PSU to check it if is acutally GPU or PSU so that may take some time.

[Vishera]

50 minutes ago, CuelmsiV said:

One more question though @Vishera. If it really is the GPU fault like failed capactior or something like you mentioned and right now getting a new decent GPU is nearly impossible, how long can i run on settings like these (i mean lowering power limit to -10% or even more for safety) is it like a safety hazard for the rest of PC, that i don't know, something might blow up after some time? Or am i good to run it with limited power limit and clock set to stock values until i get a new GPU.

It's likely that it's not a fire hazard.