Help - failing GPU?

Hi guys.

Laptop specs:

Clevo P650HS, Intel 7700HQ, Nvidia GTX 1070 (non maxQ), corsair vengeance 2400MHz 16gb x1, OS installed on 250GB Western Digital Blue M.2 SSD

Basically, since about 1-2 months ago, I’ve been having an increasing frequency of BSOD crashes. This tends to happen with graphically intensive games, which lately have been Final Fantasy 15, Assassins Creed Odyssey and Beat Sabre VR. Initially, I thought this was a driver issue, so I reinstalled the nvidia drivers, and also updated all the other drivers on my pc. This did not help. I’ve also done a clean installation of windows 10. This also was unsuccessful in fixing the problem. I ran the FurMark gpu stress test see if it was a gpu issue, and it seems like the pc always crashes around 10 minutes when the temperature is around 85 degrees Celsius. I would not expect a well working gpu to be crashing at this temperature. Nonetheless, I tried replacing the thermal paste, but that’s done little to improve the thermals.

Throughout all this, the BSODs have been getting more frequent. Initially, it was once every 2 hours but now it can happen even 5 minutes after loading into the game. Interestingly, it doesn't happen in lower graphics games like Dota 2 and The Binding of Isaac, although it actually did crash once in the latter when I had youtube playing in the background and skype on at the same time. And it's also not happened when I'm doing other stuff like browsing or watching videos or working on documents. I tried underclocking the GPU by -300MHz using MSI Afterburner, and I don't know if it's placebo or not, but the computer seemed to last longer in Assassin's Creed Odyssey and also managed to last 32 minutes in FurMark.

The BSODs themselves tend to give a different error each time; I’ve had “IRQL_NOT_GREATER_OR_EQUAL”, “SYSTEM_SERVICE_EXCEPTION”, CRITICAL_SERVICE_FAILED", “APC_INDEX_MISMATCH", "PAGE_FAULT_IN_NONPAGED_AREA", and quite a few others that I did not note down. Using blue screen viewer, the failure usually occurs in ntoskrnl.exe.

Given all these, is there any chance that I can do anything apart from sending it back for repairs? I'm just outside the warranty period, and I suspect that if the GPU is faulty, then the whole mobo will have to be replaced, and factoring in shipping, that's gonna be at least £600 or more, so I'm reluctant to do so until I've exhausted other possibilities especially since the computer seems to otherwise work.

Any suggestions?

 

If you have spare memory and won't risk your warranty opening the system, you can retest with that. Given the common denominators (load and heat), there's a probability you'll have to send it out for service, as it does kind of sound GPU-related.

 

In my opinion those BSODs aren't gpu related. More cpu related. Correct me if i'm wrong.

Gesendet von meinem SM-G965F mit Tapatalk

 

BSODs are usually a little difficult to speak for. CPU fault is certainly a possibility. The inconsistency of the errors has me suspecting CPU or memory.

 

Can you lower the ram speed in Intel xtu? Can you try the ram ins different slot? Does it do it in hybrid and dedicated gpu models?

 

"underclocking" the GPU doesn't do much, since turbo is designed to use as much power as is available, it'll usually just end up running a lower clock at a higher voltage and still be power limited.

Try setting a low frame cap to reduce gpu usage (and therefore temps) and see if that helps stability.

When you repasted, how did the thermal pads look? Were they still oily and flexible, or dry? Maybe contact with a non-core component isn't good enough and something is overheating

Good idea to check ram, I've found even just taking them out and swapping the sticks around has sometimes fixed random bsods

 

In intel XTU my ram was shown to be at 133% clock speed. I've lowered that to 100% but it still crashed after 10 minutes of furmark, although interestingly this time, there was no BSOD. Just completely shut down and then rebooted 10 seconds later.

I've swapped ram slots. No change. Sadly I don't have spare ram. But I've also run 4 passes of memory86 and it's showing no errors


Basically I think it's almost certainly a GPU hardware problem now. I'm gonna try limiting the FPS to 30 and see if I can have some stability because I'd rather not spend so much for repairs if it can still function outside of games

 

you have switchable gpu on your laptop or only 1070?

 

It's gtx 1070 and the integrated intel HD 630, running on hybrid/Optimus mode. So I guess I could potentially disable the Nvidia drivers completely to see if it still crashes when running solely on Intel

 

Random BSOD errors usually point to a memory issue. The fact that each error is blaming ntoskrnl.exe also points to memory - IIRC Windows will "blame" ntoskrnl.exe when it doesn't actually know what the faulting module is, or can't pinpoint the BSOD to a particular process.

Let memtest complete 8 passes. Sometimes bad RAM can pass memtest without errors, though. If you already swapped RAM slots, then that memory stick could very well be the culprit. One thing you can try to do is gently blow compressed air into the RAM seats to clear out any potential dust.

Your top priority should be getting your hands on a different stick of RAM and seeing if the problems persist with only that other memory installed. If you can isolate the BSODs to a single stick of memory, then hopefully you can get a warranty replacement.

 

I would try to give the cpu a little more power (few more mV in xtu)

Gesendet von meinem SM-G965F mit Tapatalk

 

Are you undervolting your CPU? I had an issue where it worked at -.125v for a month, and then it started crashing on me. Raised it to -.110v and now is stable.

 

yes, that's why asked. You should try running on intel HD then you will know if the gpu is the true problem of bsods.

 

So, I did two things. I underclocked my ram from 2400mhz to 1800mhz and I tried setting a 30fps cap for Assassins Creed Odyssey, and that seems to have lowered the average GPU temps from 85 degrees to about 60 degrees. As a result, I can play the game for hours on end without crashing. However, as soon as I set the ram back to 2400mhz, I’m getting BSODs again.

Seems like a ram issue, right? Sadly I don’t have any spare ram lying around.

But I don’t wanna invest money in buying new ram yet, until I’m certain. One reason I’m doubtful about the ram issue is that no errors have showed up on memtest64. And also, following one of the BSODs, when the laptop rebooted, the MSI afterburner program was intermittently unable to detect the GPU (the temperature kept fluctuating between 35 degrees and 0 degrees, and the clock speed kept fluctuating between the idle speed and 0 MHz), suggesting that it may still be a gpu problem? And finally, I disabled the 1070 in device manager and ran FurMark on the intel he 630, and it ran for over 30 minutes (at a measly 15 FPS) without any problems, far longer than what I’ve achieved with the 1070 enabled; would ram usage be less when using the integrated gpu in this setting ?

Any further ideas anyone?

 

It's the ram, the flickering in MSI afterburner is optimus sleeping the gpu and afterburner walking it again as it polls it for info.

 

I had bsods and it was the ram (same notebook), just one stick died.

Try to get a new one

Enviado desde mi SM-G950F mediante Tapatalk

 

I forgot, to make sure use memtest to check the ram.

I tryed windows app for that but it didnt give me any issues and then i try memtest and bam, bad ram stick.


Enviado desde mi SM-G950F mediante Tapatalk

 

Also memtest is not perfect, it cant test every possibility.