GPU or PSU problem?

[jayzackzz]

I'm sorry I think this issue has been asked a thousand times, but I can't get a definitive answer from all previous cases (sorry i'm dumb, really...).

I built a PC 2 weeks ago with an i5-11400, with a Cougar GS-X 650W PSU. I was waiting for my 3060 to arrive so all the while I was using the integrated graphics. (Did numerous CPU stress tests to try improve thermals, never had any system instabilities.)

Today, my 3060 arrived, and everything went well, the installation, drivers, until I started to game. When launching cyberpunk 2077, the whole pc crashed (was messing around with settings, crashed before I had a chance to actually play the game). Black screen, fans winding up (not sure which fan though) for a few seconds, turning off, then rebooting. I tried a few more times and sometimes it doesn't reboot automatically. I tried to run CPU + GPU stress tests on Aida64, everything was stable even after one hour (cpu temp around 64c, gpu temp at 71c, ambient temp 32c). I tried running Minecraft with ray tracing, which ran nicely for like 5 minutes before the screen going black again, fans wind up, stops and reboot. Same problem when running 3dMark Demo Time Spy Extreme. (not crashing immediately, but like halfway through the benchmark)

 

Tried uninstalling drivers with ddu in safe mode, reinstalled drivers, tried enabling and disabling resizable bar, windows 11, windows 10, reinstalled and reset the whole system, nothing worked. I don't have any other spare GPUs or PSUs to try, so I can't try to rule out one by one.

 

Experienced pc builders, is it more likely to be a PSU or GPU problem?

[emosun]

check whatever motherboard you are using doesnt have a bios update related to system stability

[davidsondakoda]

Kind of hard to say honestly but I would agree with emosun on trying a BIOS update. Possibly also check for a chipset driver update and try reseating the card, make sure all the power cables are connected properly.

[jayzackzz]

5 minutes ago, emosun said:

check whatever motherboard you are using doesnt have a bios update related to system stability

nope, the latest BIOS update released 2 days ago only states "Support WIN 11 installed or upgraded" so I didn't update since I'm on 10. Did not mention anything related to system stability.

[jayzackzz]

Just now, davidsondakoda said:

Kind of hard to say honestly but I would agree with emosun on trying a BIOS update. Possibly also check for a chipset driver update and try reseating the card, make sure all the power cables are connected properly.

I tried reseating the card, and tried another power cable, same results... 

[emosun]

3 minutes ago, jayzackzz said:

nope, the latest BIOS update released 2 days ago only states "Support WIN 11 installed or upgraded" so I didn't update since I'm on 10. Did not mention anything related to system stability.

did you want to share what the boards make and model was?

[davidsondakoda]

2 minutes ago, jayzackzz said:

nope, the latest BIOS update released 2 days ago only states "Support WIN 11 installed or upgraded" so I didn't update since I'm on 10. Did not mention anything related to system stability.

Have you updated the BIOS since getting the board? Sometimes they ship with older BIOS and you could be multiple versions behind, can compare the BIOS version with the available ones on the site to make sure.

[jayzackzz]

3 minutes ago, emosun said:

did you want to share what the boards make and model was?

I’m using the Asus ROG Strix H470i mini itx motherboard, BIOS version 2401

[emosun]

well without other hardware to test it'll be difficult to nail down. you've basically exhausted software fixes. the cheapest easiest componenet to swap is the psu so run out to a bestbuy and snag one and see if it has any effect on it's stability.

[MadGoatHaz]

This is happening a lot with these new gpu's. They are very "spikey" in their power demand and showing that many psu's are not up to the task. Diag:

 

Install msi afterburner

Lower GPU power limit to 60%

Play a game known to crash

If it works, your psu can't handle the power spikes the GPU is demanding. 

 

Best of luck, let us know what you find out.

 

[jayzackzz]

1 hour ago, MadGoatHaz said:

This is happening a lot with these new gpu's. They are very "spikey" in their power demand and showing that many psu's are not up to the task. Diag:

 

Install msi afterburner

Lower GPU power limit to 60%

Play a game known to crash

If it works, your psu can't handle the power spikes the GPU is demanding. 

 

Best of luck, let us know what you find out.

 

oh damnn i tried 3dmark time spy which I never managed to finish before this and finally finished the test lolol erm so....psu problem confirmed?

[jayzackzz]

1 hour ago, MadGoatHaz said:

This is happening a lot with these new gpu's. They are very "spikey" in their power demand and showing that many psu's are not up to the task. Diag:

 

Install msi afterburner

Lower GPU power limit to 60%

Play a game known to crash

If it works, your psu can't handle the power spikes the GPU is demanding. 

 

Best of luck, let us know what you find out.

 

but shouldn't 650W be enough for a 3060? so should i rma my psu or get a new one? 

[MadGoatHaz]

The suggested wattage for the 3000 series cards does not take into account the short power spike demands these cards ask for. 

 

I would return the PSU if you can and look into the 800 watt range to give you enough head room for the card and PSU degradation over time. 

 

Best of luck

[--SID--]

650w is more than enough, even a good 550w can handle a 3060. Problem is that it's not a good psu and has an obsolete platform with double forward on the primary. Get a 650w like be quiet! Straight Power 11, Corsair CX-F, Corsair RMx, Enermax Revolution D.F.,  Fractal Design ION+, MSI A-GF, Thermaltake GF1, XPG Core Reactor

[MadGoatHaz]

I suppose I should clarify why I would suggest 750-800 watt PSU. A 3060 can pull peaks of 18 amps. As a rule of thumb I like to add 15% for PSU degradation ensuring the PSU can power the build for many years. This makes 20-21 amps on the GPU rail requirement for the PSU. Looking more forward towards upgrade possibilities, I then like to consider at least the next power envelope higher. Perhaps a 3070 level power requirement in a future upgrade. That would be 22 amps +15% = ~25 amps. Call this 24 amps and this is why I suggest a 800watt level PSU. 

 

Using the be quiet straight power as an example, you can see the rail amperage on the spec sheets here. To --SID--'s point, yes a 650 "quality" PSU can handle a 3060 (I would never consider a 550w PSU in this situation). However, for how long (degradation) and will you want to upgrade to a higher power card in the next couple years?  

 

Anyway, I've rambled enough. 

Take care

 

 

[--SID--]

56 minutes ago, MadGoatHaz said:

I suppose I should clarify why I would suggest 750-800 watt PSU. A 3060 can pull peaks of 18 amps. As a rule of thumb I like to add 15% for PSU degradation ensuring the PSU can power the build for many years. This makes 20-21 amps on the GPU rail requirement for the PSU. Looking more forward towards upgrade possibilities, I then like to consider at least the next power envelope higher. Perhaps a 3070 level power requirement in a future upgrade. That would be 22 amps +15% = ~25 amps. Call this 24 amps and this is why I suggest a 800watt level PSU. 

 

Using the be quiet straight power as an example, you can see the rail amperage on the spec sheets here. To --SID--'s point, yes a 650 "quality" PSU can handle a 3060 (I would never consider a 550w PSU in this situation). However, for how long (degradation) and will you want to upgrade to a higher power card in the next couple years?  

 

Anyway, I've rambled enough. 

Take care

 

 

A lot of BS. Straight Power 11 has 2 rails for pci-e. RTX 3070 can be connected with 2 different cables to 2 different rails. Each pci-e rail of the 650w version can deliver 22A that makes 44A for the GPU.

 

PSU degradation is in ripple and voltage regulation, not in wattage.

[zoidzoid]

It doesn't mean psu because you dropped power limit.

It can also be an unstable gpu which is now stable with the lower power limit.

 

Is it a single fan 3060 like Palit Storm? Maybe suffering from overheating if in itx case with high ambient temperature of 32c. Not just the gpu core temp matters but maybe other parts overheating.

 

This test will help

Run card with 100% power limit but also force 100% fan speed. Take side panel/roof/whatever off of the case.

Play game/benchmark that previously crashed system and see if stable now.

[jayzackzz]

1 hour ago, zoidzoid said:

It doesn't mean psu because you dropped power limit.

It can also be an unstable gpu which is now stable with the lower power limit.

 

Is it a single fan 3060 like Palit Storm? Maybe suffering from overheating if in itx case with high ambient temperature of 32c. Not just the gpu core temp matters but maybe other parts overheating.

 

This test will help

Run card with 100% power limit but also force 100% fan speed. Take side panel/roof/whatever off of the case.

Play game/benchmark that previously crashed system and see if stable now.

tried 100% fan speed, still crashes and reboot halfway through the test 

[zoidzoid]

20 minutes ago, jayzackzz said:

tried 100% fan speed, still crashes and reboot halfway through the test 

Still maybe a gpu issue. If power supply, you would expect the crash rather instantly and not to pass aida64.
Try 100% power limit but reduce clock speed and memory speed by 200mhz each.
Is your room temp really 32? 32c room temp with closed itx case + mini card can definitely be problematic. Or you get that from a reading on your PC?

[jayzackzz]

1 minute ago, zoidzoid said:

Still maybe a gpu issue. If power supply you would expect the crash rather instantly and not to pass aida64.
Try 100% power limit but reduce clock speed and memory speed by 200mhz each.
Is your room temp really 32? Or you get that from a reading on your PC?

still crashes with -200 gpu aand memory clock🥲 yea i’m in malaysia and it’s 30+ during daytime all year long without air conditioning

[zoidzoid]

26 minutes ago, jayzackzz said:

still crashes with -200 gpu aand memory clock🥲 yea i’m in malaysia and it’s 30+ during daytime all year long without air conditioning

You should try a new power supply but honestly I think it sounds more likely the to be the gpu. If you don't want too much hassle (and just convince yourself the problem is only from high temperature) You could try reducing power limit to ~80% and then overclock to achieve same performance as if at 100% power limit. Otherwise replace gpu with another model, careful with getting the same model (if its single fan) as I think temperature issue is likely and might have same issue in your setup with another of the same card.

[MadGoatHaz]

Nope, I would put money on the PSU. I've already helped many customers with the same issue. I myself had to replace my Corsair 1000watt when I upgraded to a 3080ti for the same reason. In my case the nearly 7 year old PSU Just didn't have the legs to run the race anymore. 

[Guest]

10 hours ago, MadGoatHaz said:

This is happening a lot with these new gpu's. They are very "spikey" in their power demand and showing that many psu's are not up to the task.

 

Yeah.  But in those cases, the PSU shuts down.  @jayzackzz PC is not shutting down.

 

I don't think it's PSU related.

 

[zoidzoid]

19 minutes ago, MadGoatHaz said:

Nope, I would put money on the PSU. I've already helped many customers with the same issue. I myself had to replace my Corsair 1000watt when I upgraded to a 3080ti for the same reason. I'm my case the nearly 7 year old PSU Just didn't have the legs to run the race anymore. 

It really can be alot of things. The psu should be more than enough for a 3060 even with its power spikes. It doesn't crash during aida64 whilst cpu+gpu are pulling max power but games will put a different kind of stress on eg memory chips which maybe running very hot in itx build, high room temp, singlefan? card. These cards don't have much surface area on the heatsink so the heat can build up.

 

Just make sure all basics are properly covered before buying/replacing parts as to not waste time. Check every connector is pushed in properly at psu side as well. Check no cables are are kinked etc. Psu has 3 ports for cpu/pci-e, try the vacant one.

[MadGoatHaz]

27 minutes ago, jonnyGURU said:

Yeah.  But in those cases, the PSU shuts down.  @jayzackzz PC is not shutting down.

 

I don't think it's PSU related.

 

His PSU had 2 rails of 12v. OCP will work separately on each rail. 

 

Regardless of all this talk, I hope OP gets fixed up soon