System black screen.

[RandomTroll]

Specs:

PSU: Silverstone 850w 80+ gold

GPU: TUF-RTX3080-O10G-GAMING

Board: ROG Strix X570-F Gaming

CPU: R9 5900x

 

My system black screens after a while on high GPU load. It is not immediate. I discovered this one day while I was playing destiny 2 and kept crashing over and over. Went and tested it with superposition benchmark. Same thing happens.

What is weird is that the black screen happens but I can still hear the youtube video I was watching playing in the background. I tried the driver reset keys ctrl+shift+win+b but it is still black screened. Force restart and it goes back to normal.

This only happens on high load. I have limited the GPU power to 90% seems to have resolved the issue but that is a band aid solution.

Things I have tried include reseating the GPU, clean reinstalling the drivers, updating bios and vbios.

 

I would like some advice before I bring the GPU for repair. Maybe some troubleshooting tips I missed or if it is something else entirely. I am not comfortable disassembling the GPU and do not have spare parts to test with.

[bebejapes]

18 minutes ago, RandomTroll said:

Specs:

PSU: Silverstone 800w 80+ gold

GPU: TUF-RTX3080-O10G-GAMING

Board: ROG Strix X570-F Gaming

CPU: R9 5900x

 

My system black screens after a while on high GPU load. It is not immediate. I discovered this one day while I was playing destiny 2 and kept crashing over and over. Went and tested it with superposition benchmark. Same thing happens.

What is weird is that the black screen happens but I can still hear the youtube video I was watching playing in the background. I tried the driver reset keys ctrl+shift+win+b but it is still black screened. Force restart and it goes back to normal.

This only happens on high load. I have limited the GPU power to 90% seems to have resolved the issue but that is a band aid solution.

Things I have tried include reseating the GPU, clean reinstalling the drivers, updating bios and vbios.

 

I would like some advice before I bring the GPU for repair. Maybe some troubleshooting tips I missed or if it is something else entirely. I am not comfortable disassembling the GPU and do not have spare parts to test with.

the psu isnt enough. i contacted the manufacturer. 800w is only ment for the 2080s. i was planning on doing agpu upgrade but the psu wont support 3080

[RandomTroll]

15 minutes ago, bebejapes said:

the psu isnt enough. i contacted the manufacturer. 800w is only ment for the 2080s. i was planning on doing agpu upgrade but the psu wont support 3080

That is just false information. NVIDIA recommends 750w. Assuming NVIDIA bakes in some extra headroom even with the highest end CPU and I added another 100w on top of that doesn't seem likely. Along with the fact I have been using it for years without issue. This is a recent issue.

[bebejapes]

1 minute ago, RandomTroll said:

That is just false information. NVIDIA recommends 750w. Assuming NVIDIA bakes in some extra headroom even with the highest end CPU and I added another 100w on top of that doesn't seem likely. Along with the fact I have been using it for years without issue. This is a recent issue.

I have an email from Silverstone saying otherwise.

[bebejapes]

5 minutes ago, RandomTroll said:

That is just false information. NVIDIA recommends 750w. Assuming NVIDIA bakes in some extra headroom even with the highest end CPU and I added another 100w on top of that doesn't seem likely. Along with the fact I have been using it for years without issue. This is a recent issue.

 

just FYI

[bebejapes]

7 minutes ago, RandomTroll said:

That is just false information. NVIDIA recommends 750w. Assuming NVIDIA bakes in some extra headroom even with the highest end CPU and I added another 100w on top of that doesn't seem likely. Along with the fact I have been using it for years without issue. This is a recent issue.

you probably got away with it for a while but now the psu gave out.

[RandomTroll]

2 minutes ago, bebejapes said:

you probably got away with it for a while but now the psu gave out.

You really think a PSU manufacturer who is incentivised to get you to buy a higher end PSU is a reliable source? NVIDIA recommends 750w. A simple google search you will find people recommending 850w. Not to mention the failure method is where the screen goes back but the system is still running in the background NOT A FULL POWER TRIP. And not to mention the CPU is barely loaded in GPU benchmarks which I used to test which means the system is not at peak draw which I tested on the day I built the PC with no issues.

[bebejapes]

7 minutes ago, RandomTroll said:

You really think a PSU manufacturer who is incentivised to get you to buy a higher end PSU is a reliable source? NVIDIA recommends 750w. A simple google search you will find people recommending 850w. Not to mention the failure method is where the screen goes back but the system is still running in the background NOT A FULL POWER TRIP. And not to mention the CPU is barely loaded in GPU benchmarks which I used to test which means the system is not at peak draw which I tested on the day I built the PC with no issues.

i believe the psu manufacturer would know what they are doing since they designed the psu. i had similar instances where it worked fine right off the bat but later failed in other components not just psu. Yes it would work when you tested it but its not a guarantee.

[bebejapes]

20 minutes ago, RandomTroll said:

You really think a PSU manufacturer who is incentivised to get you to buy a higher end PSU is a reliable source? NVIDIA recommends 750w. A simple google search you will find people recommending 850w. Not to mention the failure method is where the screen goes back but the system is still running in the background NOT A FULL POWER TRIP. And not to mention the CPU is barely loaded in GPU benchmarks which I used to test which means the system is not at peak draw which I tested on the day I built the PC with no issues.

As per the email from silverstone transient spikes from the gpu might be high at any point of time so the psu might not handle it.

[bebejapes]

@filpo this is the thread.

[filpo]

1 hour ago, RandomTroll said:

Specs:

PSU: Silverstone 850w 80+ gold

GPU: TUF-RTX3080-O10G-GAMING

Board: ROG Strix X570-F Gaming

CPU: R9 5900x

 

My system black screens after a while on high GPU load. It is not immediate. I discovered this one day while I was playing destiny 2 and kept crashing over and over. Went and tested it with superposition benchmark. Same thing happens.

What is weird is that the black screen happens but I can still hear the youtube video I was watching playing in the background. I tried the driver reset keys ctrl+shift+win+b but it is still black screened. Force restart and it goes back to normal.

This only happens on high load. I have limited the GPU power to 90% seems to have resolved the issue but that is a band aid solution.

Things I have tried include reseating the GPU, clean reinstalling the drivers, updating bios and vbios.

 

I would like some advice before I bring the GPU for repair. Maybe some troubleshooting tips I missed or if it is something else entirely. I am not comfortable disassembling the GPU and do not have spare parts to test with.

That should be enough to handle a 3080. Even 750 watts is fine

Does it shut down or is it just a black screen? if its just a black screen and the pc is still on it doesn't sound like a psu issue. But if it crashes then shuts down its a psu issue most likely

 

EDIT: 

Btw when I say 'fine' i mean it might not take an OC or two from both cpu and gpu. So when I say 'fine' I mean it should work but might not depending on both silicon lottery and model

Edited May 20, 2023 by filpo

[RandomTroll]

13 minutes ago, bebejapes said:

i believe the psu manufacturer would know what they are doing since they designed the psu. i had similar instances where it worked fine right off the bat but later failed in other components not just psu. Yes it would work when you tested it but its not a guarantee.

Just to double check I went and searched power pc calculator. I went through the top 3 results. Cooler master new egg and be quite. Calculated the PC power draw and all of them say 850w or less works. General advice I see from reddit is 750 is recommended by NVIDIA 850 should be good. And yes 850w does take into account transient spikes as the GPU power is suppose to be 350w assuming since I have an OC version it is 450w+150w from the CPU even though it has a curve optimizer offset of -15 and it is not drawing full power when system crashes. Another 100w from the system that is still 700w total.

[RandomTroll]

1 minute ago, filpo said:

That should be enough to handle a 3080. Even 750 watts is fine

Does it shut down or is it just a black screen? if its just a black screen and the pc is still on it doesn't sound like a psu issue. But if it crashes then shuts down its a psu issue most likely

 

EDIT: 

Btw when I say 'fine' i mean it might not take an OC or two from both cpu and gpu. So when I say 'fine' I mean it should work but might not depending on both silicon lottery and model

As per the original post system does not shut down. I still can hear the YT vid running in the background. CPU is undervolted -15 on the curve optimizer. GPU is an OC version crashes even at 100% power on MSI after burner. Ran it at 112% for years. 90% seems fine.

[filpo]

Just now, RandomTroll said:

As per the original post system does not shut down. I still can hear the YT vid running in the background. CPU is undervolted -15 on the curve optimizer. GPU is an OC version crashes even at 100% power on MSI after burner. Ran it at 112% for years. 90% seems fine.

Oh ok sorry didn't see that

Maybe try DDUing your drivers and using a different Displayport port or going from HDMI to displayport (if you can) 

You could also try reseating your gpu power connectors and the gpu itself in the gpu slot

[RandomTroll]

Just now, filpo said:

Oh ok sorry didn't see that

Maybe try DDUing your drivers and using a different Displayport port or going from HDMI to displayport (if you can) 

You could also try reseating your gpu power connectors and the gpu itself in the gpu slot

As per the original post I also did that already.

[filpo]

1 minute ago, RandomTroll said:

As per the original post I also did that already.

have you tried undervolting your gpu?

[bebejapes]

2 minutes ago, filpo said:

have you tried undervolting your gpu?

he's done it already. but i still think the psu is nearly gone, it no longer can handle the output to the gpu. he could do a heavy undervolt of 50%

[RandomTroll]

Just now, filpo said:

have you tried undervolting your gpu?

No only set the power slider in MSI afterburner to 90. And undervolting doesn't seem like a perma fix. It fixes when I limit the power to 90 but that is not an acceptable solution IMO. I'll just bring the card in for repair if that is the only option since it is still under warrenty.

[RandomTroll]

Just now, bebejapes said:

he's done it already. but i still think the psu is nearly gone, it no longer can handle the output to the gpu. he could do a heavy undervolt of 50%

I have not I only reduced the power to 90%. that is not the same as undervolting.

[bebejapes]

@adm0n please comment

[filpo]

15 minutes ago, RandomTroll said:

I have not I only reduced the power to 90%. that is not the same as undervolting.

You could always try undervolting and it brings gains. I do it on my r5 5600 and my 3060 and I get higher clocks for the same power on both

Linus shows how at 13:50 

 

[adm0n]

19 minutes ago, RandomTroll said:

. CPU is undervolted -15 on the curve optimizer.

Did you test this for stability long enough? The curve optimizer can make your CPU spit out wrong calculations, which can result in almost anything.

 

14 minutes ago, RandomTroll said:

I have not I only reduced the power to 90%. that is not the same as undervolting.

It somewhat is. If you limit the power your GPU can draw, it will have to actually use less power, which it usually does by going to a lower frequency, which usually also corresponds with a voltage drop.

Usually people like to overclock their chips at a lower voltage, which is what normally is referred to as undervolting.

 

But if your card is still under warranty, it would be unfortunate to get this diagnosis wrong. So if you have the option, trying it with another PSU, would be the best. Maybe in a friends PC.

[RandomTroll]

6 minutes ago, adm0n said:

Did you test this for stability long enough? The curve optimizer can make your CPU spit out wrong calculations, which can result in almost anything.

 

It somewhat is. If you limit the power your GPU can draw, it will have to actually use less power, which it usually does by going to a lower frequency, which usually also corresponds with a voltage drop.

Usually people like to overclock their chips at a lower voltage, which is what normally is referred to as undervolting.

 

But if your card is still under warranty, it would be unfortunate to get this diagnosis wrong. So if you have the option, trying it with another PSU, would be the best. Maybe in a friends PC.

Yes I ran prime 95 for a couple hours. The issue popped up recently and I have been running curve optimizer before that. I have no other parts to test with. My friends use laptops.

[adm0n]

8 minutes ago, RandomTroll said:

Yes I ran prime 95 for a couple hours. The issue popped up recently and I have been running curve optimizer before that. I have no other parts to test with. My friends use laptops.

 Did you just ran prime 95? Because the errors it produces don't necessarily crash your system. Core cycler runs prime 95 on each CPU core for a couple of minutes before going to the next one and checks if the results it produces are fine. (https://github.com/sp00n/corecycler). If you run this for lets say at least 3 iterations without problems you are probably fine.

 

10 minutes ago, RandomTroll said:

The issue popped up recently and I have been running curve optimizer before that

It could be that the degradation of your PSU is causing your CPU to run ever so slightly less stable (because the power it gets is less stable). If you are running it at the limit with curve optimizer, that could push it over the brink. But I actually don't know if it works that way.

 

13 minutes ago, RandomTroll said:

I have no other parts to test with. My friends use laptops.

Then you have a couple of options.

1. Heavily undervolt your GPU and see if you are happy with that performance and hope that everything else is fine. (I wouldn't just take off 10% of the power, because than you could run into the problem again in some time)
2. RMA the PSU, if it still has warrenty
3. RMA the GPU and hope that it fixes it
4. Buy a new PSU and hope that it fixes the problem.

I can't really see what other options you have.

[RandomTroll]

11 minutes ago, adm0n said:

 Did you just ran prime 95? Because the errors it produces don't necessarily crash your system. Core cycler runs prime 95 on each CPU core for a couple of minutes before going to the next one and checks if the results it produces are fine. (https://github.com/sp00n/corecycler). If you run this for lets say at least 3 iterations without problems you are probably fine.

 

It could be that the degradation of your PSU is causing your CPU to run ever so slightly less stable (because the power it gets is less stable). If you are running it at the limit with curve optimizer, that could push it over the brink. But I actually don't know if it works that way.

 

Then you have a couple of options.

1. Heavily undervolt your GPU and see if you are happy with that performance and hope that everything else is fine. (I wouldn't just take off 10% of the power, because than you could run into the problem again in some time)
2. RMA the PSU, if it still has warrenty
3. RMA the GPU and hope that it fixes it
4. Buy a new PSU and hope that it fixes the problem.

I can't really see what other options you have.

Yeah I am just gonna try and go and get the GPU repaired. Sucks I have to travel and prob not be able to use my PC.