Computer Shutting Down Randomly During GPU Usage

[Nick Conrad]

My brother built this PC a few years back as a gaming computer, but roughly two years ago it started having issues where it would randomly shut down during any sort of GPU usage. It turned out that one RAM stick was corrupt so we replaced them all but the issue continued. We also switched the GPU out for a brand new one, but that didn't work.

CPU: i5-3570k (3.4GHz, no overclock)
CPU Cooler: Kuhler h2o 620
Motherboard: ASUS Sabertooth z77
Original GPU: MSI Twin Frozr
Current GPU: AMD Radeon R9 380 (updated drivers)
Original RAM: 4x G.SKILL RIPJAWS X
Current RAM: 2x Micron Crucial
PSU: Corsair HX850
OS: Windows 64bit

Given what I've read on forums, the shut-downs look like a PSU issue, but all PSU voltages are within +0.2v of what they should be. I don't believe it's an overheating issue because all the components stay below 70c (except during an Intel Burn Test, the CPU hit 97c then but didn't shut down). The ambient CPU temp is roughly 35c due to climate, and the GPU loiters around 40c, idle. One interesting thing to note is that whenever I attempt to run a Furmark GPU benchmark, my PC shuts down instantly, within 1 second of opening the OpenGL. Any ideas?

Worth noting that on shutdown, I get no BSOD and the computer restarts automatically, but gives me an 'Error Recovery' screen on startup.

 

[Antonis117]

The fact that the voltage of the PSU is within the 200mV marging doesn't mean that when your card is going to draw a higher current it will remain within this margin. I don't know if you understand what I mean.

 

[Antonis117]

It is very likely that there is something wrong with the PSU and you are experiencing a voltage drop when the GPU is drawing more power. This is normal and that is what it happens when we draw higher current from a power supply. The voltage tends to drop BUT the PSU is designed to keep that voltage stable and supply the required current to the load. So what I think is that you should check your PSU as it may be failing to work properly. If there is any other PSU you can try, do it.

 

[Nick Conrad]

It is very likely that there is something wrong with the PSU and you are experiencing a voltage drop when the GPU is drawing more power. This is normal and that is what it happens when we draw higher current from a power supply. The voltage tends to drop BUT the PSU is designed to keep that voltage stable and supply the required current to the load. So what I think is that you should check your PSU as it may be failing to work properly. If there is any other PSU you can try, do it.

I'll see if I can get my hands on another one, but in the meantime is there any way I can test the PSU output to see if it drops while my GPU is in use? As soon as anything is demanded from the GPU, the whole systems shuts down, so I'd have a hard time measuring the PSU output in that half-second of GPU use.

 

[Antonis117]

Yes it is difficult to test it, you need test instruments to monitor the PSU output which is difficult to have. You can test the PSU by connecting it to another PC if possible and test it there.

 

[Nick Conrad]

Yes it is difficult to test it, you need test instruments to monitor the PSU output which is difficult to have. You can test the PSU by connecting it to another PC if possible and test it there.

Don't have another PC, unfortunately. On the plus side, I updated the motherboard BIOS and it ran Furmark for about a second and a half without crashing, which I'll count as a success.

 

[Nick Conrad]

It is very likely that there is something wrong with the PSU and you are experiencing a voltage drop when the GPU is drawing more power.

I ran a program that tends to cause crashes while looking at PSU voltages and GPU temperatures. The most the PSU would drop would be less than 0.1v before the crash, while one of the GPU temperature sensors reached 80c (but only one fan was running). This could be a cause of crashes over a longer period of time, but doesn't explain the instant crashes from Furmark.

 

[Antonis117]

Yes but since you are measuring the voltages from the system itself, how can you see if the voltage drops at a point that causes the shutdown while the system itself shuts down? You cant. In order to see that you need test equipment such as multimeters and oscilloscopes, equipment that measures the voltage outside the system itself. The point is that if possible find an other PSU (if you can borrow from a friend for example or if you have any spare) and replace the one you have. If the PC stops shutting down then its the PSU. Simple. If with the new PSU it keeps doing it then it is something else and it needs a deeper look.

 

[Nick Conrad]

If the PC stops shutting down then its the PSU. Simple. If with the new PSU it keeps doing it then it is something else and it needs a deeper look.

Sure, I'll see if I can borrow one from somewhere. Any idea how I can figure out the minimum wattage required from a PSU? Thanks for your help!

 

[Antonis117]

You mean the minimum wattage required by your system? The R9 380 in specs has a minimum of 500W. So any decent PSU from 600W and upwards is fine (take in account and the rest of the system). That doesn't mean necessarily that it will draw 500W, but it is more of a guide and a worst case scenario.

 

[Nick Conrad]

So I've borrowed a 550W PSU from a friend and hooked everything up but the computer won't boot. All the components' LEDs would be lit and the fans would be running but the display would remain black (connected and with power). Thoughts?

 

[Antonis117]

Check that you plugged everything correctly and try again. Also do you get any beeps from the motherboard?

 

[Nick Conrad]

Check that you plugged everything correctly and try again. Also do you get any beeps from the motherboard?

Yep I've checked everything is going where it should be. There are a few non-connected PSU cables but that was the same case with the Corsair. All the lights turn on and the fans start spinning, but no beeping from the motherboard or any screen display.

 

[Antonis117]

Try connecting the monitor cable to the motherboard directly leaving the gpu out.

 

[Nick Conrad]

Try connecting the monitor cable to the motherboard directly leaving the gpu out.

Yeah, that's not changing anything. Still black screen and components lighting up. I'm going to try to hook up the old PSU and see if that works still.

 

[Antonis117]

Try connecting the monitor cable to the motherboard directly leaving the gpu out.

Yeah, that's not changing anything. Still black screen and components lighting up. I'm going to try to hook up the old PSU and see if that works still.

Did you take the gpu out of the case and tried with the monitor cable conected to the motherboard and still didnt work? Check that the 24pin power connector and the 12V 8pin for cpu is in place.

 

[Nick Conrad]

Did you take the gpu out of the case and tried with the monitor cable conected to the motherboard and still didnt work? Check that the 24pin power connector and the 12V 8pin for cpu is in place.

Yep, nothing. I reconnected the old PSU and got an interesting error message:
BOOTMGR is missing
Press Ctrl+Alt+Del to restart
Initial research tells me this is because it can't find the HD that MSW7 is on. Curious.

EDIT: Turns out it was a broken SATA III cable to the SSD that hosts MSW7, instead of the HDD that holds everything else. Which means I can't run any tests until I get a new SATA cable. :/

 

[Nick Conrad]

I put the old GPU back in just to see how it would run while waiting for the SATA cable. I believe the GPU I put in is a GeForce GTX660 Twin Frozr, which shouldn't need as much power as the R9. Furmark couldn't run as the GPU is pre-OpenGL 2.0, so I couldn't test it there. The PC still crashes. Any ideas?

 

[Antonis117]

It crashed using your PSU? Also can you change the GPU to the second pcie slot and try again?

 

[Nick Conrad]

It crashed using your PSU? Also can you change the GPU to the second pcie slot and try again?

Yep it was my PSU. It seems to be a problem with anti-aliasing, whenever I raise that bar the PC crashes. Is this at all related to PSU wattage consumption? If it is, then I think we've got a culprit.

 

[Antonis117]

Well Anti-Aliasing can be very demanding to GPUs (especially SSAA). However I don't think it is just that. In order to eliminate the possibility of a faulty PCIe I am going to ask you to change the GPU to the second one and try again.

 

[Nick Conrad]

I tried the old GPU in both motherboard slots but if anything, it crashes faster. I tried the newer GPU in both PCIe slots but still no success.

 

[Antonis117]

Ok so what we have now is that it is not a GPU issue. Also the PCIe slots are working. Now you can check power connectors to make sure there is nothing loose. If the problem remains then I would advise you to remove the motherboard from the case with all the components, put them on a bench carefully and try again. If it insists then I think that we have a faulty PSU.

 

[Nick Conrad]

If the problem remains then I would advise you to remove the motherboard from the case with all the components, put them on a bench carefully and try again.

Do you mean pull everything out and hook it up then try and run it? So the only thing that would be changing would be the case?

If it helps at all, the PC seems to run longer before crashing if I leave it for a while after it crashes, which seems a bit weird.

 

[Antonis117]

If the problem remains then I would advise you to remove the motherboard from the case with all the components, put them on a bench carefully and try again.

Do you mean pull everything out and hook it up then try and run it? So the only thing that would be changing would be the case?

If it helps at all, the PC seems to run longer before crashing if I leave it for a while after it crashes, which seems a bit weird.

Yes remove everything, take the cpu with care out, the ram and then install everything back in again with care. Make sure everything is plugged in and try again. If the problem remains it is probably the psu or the motherboard. But since you tried in different pcie it is hard to say the motherboard is faulty. Also even if it is a little difficult to do, you could take your psu out and try it on an other computer if you have to see if it works. These are all bouble-checking instructions to make sure the fault is in the psu and not somewhere else tricking us.