3090 Black Screens .

[Johnys7]

Hello . Recently after i upgraded to a newer platform (Z690) i have started experiencing a black screen issue. This occurs ONLY if i increase the power limiter and temp limiter through MSI afterburner of the card. If default, it some times completed or not the PR Stress Test. I checked everything i could. After running Memtest i found out that there was an issue with one stick as it wouldnt get recognised from pc and mobo would post cod 55. I then tried newer sticks (Corsair Dominator 5600 DDR5 36-36-36-76) and the issue happened again when i tried the test, but when i was completely default on the 3090 the PR got completed. Previously with the defective kit , it managed to complete it but i had less than 97% stability.Also, when i tried Unigine Superposition i am getting quite a low score for a 3090. Everything else seems to be fine, as i run multiple CPU stress tests (Intel Burn Test, Aida etc) and CPU is fine. I dont think that it is a PSU related issue, as i consider the Dark Power Pro 12 (1500W) quiet an overkill.. Temps are fine. CPU is stock, only thing i done is Editted Power Plan to High Performance.  Also , memory is running at XMP 1. All drivers are updated and BIOS version of the mobo is the latest (1403). I really cant find out what is causing this issue as i thought that replacing the faulty ram would fix it, but its still there. My only guess, would be that it has something to do with GPU?!

 

OS: Windows 11 Home , Build: 22000.652

CPU:  i9 12900K

Mobo: Asus ROG Hero Z690

RAM: Corsair Dominator 5600 32g (2x16) 

GPU: Asus Rog Stric 3090 OC

PSU: Be Quiet Dark Power Pro 12 1500W Titanium

[Lord Bloobus]

39 minutes ago, Johnys7 said:

This occurs ONLY if i increase the power limiter and temp limiter through MSI afterburner of the card.

Don't do this then, most cards are running right on the line between stable and unstable out of the box. Yours might just not have any headroom left. If you need more you'll probably have to up the voltage, which I personally don't recommend.

[mr fobs]

So it only happens when you OC? The ROG Strix 3090 is already Factory OC'd. When I had a SYTIX 3090, if I tried anything other than a very very very very slight increase, it would black screen.

[Johnys7]

14 minutes ago, mr fobs said:

So it only happens when you OC? The ROG Strix 3090 is already Factory OC'd. When I had a SYTIX 3090, if I tried anything other than a very very very very slight increase, it would black screen.

 

15 minutes ago, Lord Bloobus said:

Don't do this then, most cards are running right on the line between stable and unstable out of the box. Yours might just not have any headroom left. If you need more you'll probably have to up the voltage, which I personally don't recommend.

Sometimes can even happen if i just up the fan speed. Its so weird this issue. But what about the scores i posted? Arent they showcasing that there is something going on with GPU considering that i my score is even lower than a 3080 ? 

[Lord Bloobus]

3 minutes ago, Johnys7 said:

 

Sometimes can even happen if i just up the fan speed. Its so weird this issue. But what about the scores i posted? Arent they showcasing that there is something going on with GPU considering that i my score is even lower than a 3080 ? 

I see a handful of 3090's with scores close to or similar to yours. It can also be influenced by other things in the system. It sounds like power delivery is a bit spotty though. Can you get a picture of how the GPU is connected to the PSU? It should be using 3 individual 8 pin connectors, no pigtails.

[Johnys7]

5 minutes ago, Lord Bloobus said:

I see a handful of 3090's with scores close to or similar to yours. It can also be influenced by other things in the system. It sounds like power delivery is a bit spotty though. Can you get a picture of how the GPU is connected to the PSU? It should be using 3 individual 8 pin connectors, no pigtails.

Not quite easy to show this, but they are going straight to the PSU. No splitter used or anything similar. Isnt 1500W enough for my system?

 

[Lord Bloobus]

4 minutes ago, Johnys7 said:

Not quite easy to show this, but they are going straight to the PSU. No splitter used or anything similar. Isnt 1500W enough for my system?

 

Yes, you can maybe try a different BIOS for the card from the Asus support page or reach out to customer support and say your card seems to be underperforming.

[Johnys7]

3 minutes ago, Lord Bloobus said:

Yes, you can maybe try a different BIOS for the card from the Asus support page or reach out to customer support and say your card seems to be underperforming.

Also on this specific GPU, on the top of its power connection, there are 3 LED indiicators which they will light up red when something is going on with the power delivery. For example, faulty cable, not enough power etc. None of them ever lighted up, which means that the GPU is getting the power it needs.

[Falkentyne]

12 hours ago, Johnys7 said:

Also on this specific GPU, on the top of its power connection, there are 3 LED indiicators which they will light up red when something is going on with the power delivery. For example, faulty cable, not enough power etc. None of them ever lighted up, which means that the GPU is getting the power it needs.

That isn't necessarily true at all.

That just shows that there isn't a short or a fault on the line.  It does NOT mean that the card is 'getting the power it needs'.  Far from it.  And we're not talking about power actually.  We're talking about voltage.  Not watts.

 

Run HWinfo64, go to sensors only, *EXPAND* all of the GPU sensor "collapsed" threads/fields and then post a screenshot of the voltage rails.

I'm going to bet more than one of them will be below 11.9v.

 

Basically, you can try:

1.  Use a single dedicated 8 pin PCIE cable (going to one 8 pin connector on the PSU) for each 8 pin on the video card.  Meaning do NOT use a daisy-chained 2x8 pin cable into two ports on the video card, going to 1 8 pin on the PSU end.  So on a 3x8 pin card you would need three separate 8 pin cables.  For FE cards, the adapter itself is usually enough, or a OEM official supplied micro-fit 3.0 cable (Seasonic makes one for seasonic OEM branded PSU's, I believe there is at least one other manufacturer who makes them, but Corsair does NOT make their own PSU's in-house at their own factory, and their cables are supplied by the OEM).

 

Can you explain the black screen please?
Is it just a 'black screen then the system powers off and reboots'?

Or is it a 'black screen followed by GPU fans running at 100% speed and a hard lockup?'

[xg32]

12 hours ago, Johnys7 said:

.

from the info you've given it's most likely too much going through 1 rail, if you are already using 3 different cables and make sure nothing is sharing the same rail as the gpu, Your psu has 6 12v rails, and they are all labeled, make sure to use 3 different numbers for the gpu,

 

it's down to an issue with the gpu power delivery being uneven on the cables, current gen gpus are notorious for having this issue with psus under 60A single rail. 

 

To be clear, the fault is with nvidia, not the psu.

[Johnys7]

4 hours ago, xg32 said:

from the info you've given it's most likely too much going through 1 rail, if you are already using 3 different cables and make sure nothing is sharing the same rail as the gpu, Your psu has 6 12v rails, and they are all labeled, make sure to use 3 different numbers for the gpu,

 

it's down to an issue with the gpu power delivery being uneven on the cables, current gen gpus are notorious for having this issue with psus under 60A single rail. 

 

To be clear, the fault is with nvidia, not the psu.

So what do you actually suggest me to do? I am really so desperate.

 

4 hours ago, Falkentyne said:

That isn't necessarily true at all.

That just shows that there isn't a short or a fault on the line.  It does NOT mean that the card is 'getting the power it needs'.  Far from it.  And we're not talking about power actually.  We're talking about voltage.  Not watts.

 

Run HWinfo64, go to sensors only, *EXPAND* all of the GPU sensor "collapsed" threads/fields and then post a screenshot of the voltage rails.

I'm going to bet more than one of them will be below 11.9v.

 

Basically, you can try:

1.  Use a single dedicated 8 pin PCIE cable (going to one 8 pin connector on the PSU) for each 8 pin on the video card.  Meaning do NOT use a daisy-chained 2x8 pin cable into two ports on the video card, going to 1 8 pin on the PSU end.  So on a 3x8 pin card you would need three separate 8 pin cables.  For FE cards, the adapter itself is usually enough, or a OEM official supplied micro-fit 3.0 cable (Seasonic makes one for seasonic OEM branded PSU's, I believe there is at least one other manufacturer who makes them, but Corsair does NOT make their own PSU's in-house at their own factory, and their cables are supplied by the OEM).

 

Can you explain the black screen please?
Is it just a 'black screen then the system powers off and reboots'?

Or is it a 'black screen followed by GPU fans running at 100% speed and a hard lockup?'

For Example, when i am gaming or stresstesting , everything will go black, pc will still keep runing and my only option is to hard shutdown/reset .

[Johnys7]

4 hours ago, Falkentyne said:

That isn't necessarily true at all.

That just shows that there isn't a short or a fault on the line.  It does NOT mean that the card is 'getting the power it needs'.  Far from it.  And we're not talking about power actually.  We're talking about voltage.  Not watts.

 

Run HWinfo64, go to sensors only, *EXPAND* all of the GPU sensor "collapsed" threads/fields and then post a screenshot of the voltage rails.

I'm going to bet more than one of them will be below 11.9v.

 

Basically, you can try:

1.  Use a single dedicated 8 pin PCIE cable (going to one 8 pin connector on the PSU) for each 8 pin on the video card.  Meaning do NOT use a daisy-chained 2x8 pin cable into two ports on the video card, going to 1 8 pin on the PSU end.  So on a 3x8 pin card you would need three separate 8 pin cables.  For FE cards, the adapter itself is usually enough, or a OEM official supplied micro-fit 3.0 cable (Seasonic makes one for seasonic OEM branded PSU's, I believe there is at least one other manufacturer who makes them, but Corsair does NOT make their own PSU's in-house at their own factory, and their cables are supplied by the OEM).

 

Can you explain the black screen please?
Is it just a 'black screen then the system powers off and reboots'?

Or is it a 'black screen followed by GPU fans running at 100% speed and a hard lockup?'

 

[Falkentyne]

8 hours ago, Johnys7 said:

So what do you actually suggest me to do? I am really so desperate.

 

For Example, when i am gaming or stresstesting , everything will go black, pc will still keep runing and my only option is to hard shutdown/reset .

 

You didn't expand the GPU Rail voltages like I asked you to do...

 

Also make sure you do it while the GPU is at full load, not at idle.

seeing a voltage of 0.739v isn't helpful.

[xg32]

8 hours ago, Johnys7 said:

 

the 3 cables on the gpu, plug them into 12v4 12v5 12v6 on ur psu, for everything else on ur pc use 1 2 and 3

[Johnys7]

3 hours ago, Falkentyne said:

 

You didn't expand the GPU Rail voltages like I asked you to do...

 

Also make sure you do it while the GPU is at full load, not at idle.

seeing a voltage of 0.739v isn't helpful.

I dont see any difference in between the numbers when idle or when under load... This is the Rail Voltages while i was runnine Unigine Superposition.

[Falkentyne]

2 hours ago, Johnys7 said:

I dont see any difference in between the numbers when idle or when under load... This is the Rail Voltages while i was runnine Unigine Superposition.

Thank you.  The voltages are in spec.  That means your PSU cables are not degraded with high resistance and dropping the +12v input rails.  That's good news.

Are you using three independent separate 8 pin cables, one for each 8 pin on the GPU?

[jones177]

I have had MSI Afterburner freeze a GPU bios causing it to underperform. 

This freeze could survive a reboot so I had to switch the computer off and on again.

 

Easy to test use another overclocking tool and see if it happens.  

[Johnys7]

10 hours ago, jones177 said:

I have had MSI Afterburner freeze a GPU bios causing it to underperform. 

This freeze could survive a reboot so I had to switch the computer off and on again.

 

Easy to test use another overclocking tool and see if it happens.  

The same thing happens even with Asus Tweak .....

 

13 hours ago, Falkentyne said:

Thank you.  The voltages are in spec.  That means your PSU cables are not degraded with high resistance and dropping the +12v input rails.  That's good news.

Are you using three independent separate 8 pin cables, one for each 8 pin on the GPU?

Yes....

[Falkentyne]

15 hours ago, Johnys7 said:

The same thing happens even with Asus Tweak .....

 

Yes....

Ok, as I asked earlier, is the black screen a "max GPU fans+100%" crash with a hard lock? Or a normal "black screen then system reboots in a few seconds?"

[Johnys7]

On 5/1/2022 at 4:55 AM, Falkentyne said:

Ok, as I asked earlier, is the black screen a "max GPU fans+100%" crash with a hard lock? Or a normal "black screen then system reboots in a few seconds?"

Its a black screen, everything works like normal and i have to press the On/Off button to restart the PC. Sometimes i get a broken image of the desktop, you know like broken colours etc. Most common is the 1st case though. By the way, i upgrade to a newer mobo (z690 rog Formula) and the problem is still there. Also did what you say on the psu 12v4,5,6 but nothing changed. Tomorrow i am waiting for a 3090Ti . If the problem still exist after that, only thing that could be is the PSU.... 

 

[Falkentyne]

4 hours ago, Johnys7 said:

Its a black screen, everything works like normal and i have to press the On/Off button to restart the PC. Sometimes i get a broken image of the desktop, you know like broken colours etc. Most common is the 1st case though. By the way, i upgrade to a newer mobo (z690 rog Formula) and the problem is still there. Also did what you say on the psu 12v4,5,6 but nothing changed. Tomorrow i am waiting for a 3090Ti . If the problem still exist after that, only thing that could be is the PSU.... 

 

Might be a bad card.

Black screen+100% fan speed is usually the GPU detecting a fatal fault and shutting itself down.  If it's VRM's overheating because of a defective thermal pad, this will usually be shown via "GPU Hotspot temperature" reaching over 100C, since as far as I know, both the GPU Core and the VRM's report to the hotspot together.

 

But usually, this can be caused by an issue with the core itself, or with the power delivery (VRM's).  Let us know if the 3090 Ti solves the problem because if it does, it means the old card was just bad.

[Johnys7]

On 5/2/2022 at 11:11 PM, Falkentyne said:

Might be a bad card.

Black screen+100% fan speed is usually the GPU detecting a fatal fault and shutting itself down.  If it's VRM's overheating because of a defective thermal pad, this will usually be shown via "GPU Hotspot temperature" reaching over 100C, since as far as I know, both the GPU Core and the VRM's report to the hotspot together.

 

But usually, this can be caused by an issue with the core itself, or with the power delivery (VRM's).  Let us know if the 3090 Ti solves the problem because if it does, it means the old card was just bad.

Case Update:
After installing 3090 Ti, no blackscreens and i could complete a 20 loop stress test on Port Royal with maximum Power Target. So, it seems that my Asus Rog Strix 3090 OC is deffective.