cancel
Showing results for 
Search instead for 
Did you mean: 

1080Ti Strix OC 11GB random blackscreen while stresstest or benchmark (stockspeeds)

JargonHolpick
Level 7
Hi,

Recently I bought myself a ASUS 1080Ti Strix OC 11GB and besides what the titles says, I'm pretty happy with the card.
But the blackscreen issue bugs me.
Realbench benchmark, no problem. Realbench stresstest, blackscreen and instability detected.
HWbot Benchmark, blackscreens...
While gaming, no issues. Same for desktop, valley and Heaven test.
Update: 1 time blackscreen with Arctifacts playing the game RocketLeague.

What causes this blackscreen? Windows Logs (system) says:
- System

- Provider

[ Name] Display

- EventID 4101

[ Qualifiers] 0

Level 3

Task 0

Keywords 0x80000000000000

- TimeCreated

[ SystemTime] 2017-07-02T21:21:17.761432000Z

EventRecordID 12512

Channel System

Computer WORKHORSE7700K

Security


- EventData

nvlddmkm


Displaydriver nvlddmkm didn't respond and is repaired (might be wrong translated. Original is in Dutch ;-).

My config:
-Asus z170 ProGaming
-I7 7700K
-1080Ti Strix OC 11Gb
-Kraken X62
-Samsung SSD 500GB evo 840
-2 HDD's (2 & 3 TB green WD's)
-Seasonic X560

About the PSU:
If I take a look at the power consumption of my total system while gaming (Crysis 3 al maxed out 1440P) I see a wattage of: avg 400watt.
While FURMARK ROG Edition: Furry rROG 1440P Fullscreen stresstest: 380-390Watt
Realbench benchmark: Between 130 en 380 watt

I read somewhere that the max power-consumption of the 1080Ti strix OC lies between 280 and 300 Watts.
So, 300watts divided by 12volts makes 25Amps.
The seasonic X560 max load is 46Amps, so I guess I'm fine with the PSU.

By the way, before this 1080Ti I had a ASUS 1070strix mounted. Never failed nor gave me black screens.
Any help or explanations is welcome.


Thanks
4,932 Views
12 REPLIES 12

Marcello_S
Level 7
Hi Jargon,
If it reports error "Displaydriver stopped working and has restored", it could be a defective driver installation.

So I suggest you to perform a complete graphic driver uninstallation using DDU, then a registry cleaning and after that you can proceed with the normal installation by checking "perform a clean installation" when selected custom installation in the nVidia windows wizard.

If the problem persists in the Realbench stresstest, you could perform it with the card downclocked and see if something changes.

Let us know.

Heini
Level 11
From PSU to GPU .... are you using two separate cables or one cable with a splitter?
Also, though numbers may add up I think you're pushing that little PSU too hard.

JargonHolpick
Level 7
@Heini: 2cables, no splitters. It should require a minimum of 42A and 550Watt, for what I found.
I know that the PSU might be the culprit but I guess it's not. AS long as I don't see powerdrain over 520watts (at the wallsocket), I still believe I'm save.
It's a damn good PSU.
But let's say if the PSU is close to its end what would you suggest? I bought Seasonic like 10 years ago because that was the only decent brand that build PSU's that delivered what was on paper and more.

I also know that a 60% power usage is more efficient for a PSU rather than +80%.
So, be my guest and make a suggestion.

@Marcello: I've tried the DDU + regclean and latest drivers and older drivers and increased voltage... pffff
But I'm gonna try downclocking, never done that before.
If the GPU passes when downclocked, is this card a 'dud'?

JargonHolpick
Level 7
Update:
@Marcello: So now I'm a bit disappointed. I've bought a factory overclocked card that costed me more that an average laptop and I can't get it passed the ASUS realbench stresstest?
It did passed the test, only by downclocking 😞
I slide -50 on the memory and the core.
Passed like it was nothing.
Explain, please. There is nothing in the world that could justify this issue, as far as I'm willing to think.
euh euh, the top of the top... "the best performing singlecore card" , and I have to slide the overclock down to get it stable? And best of all is that it's an ASUS stresstest.
Feeling really bad right now.

@Heini: During the stress test I got a max powerusage of 360watt. I guess that the PSU is save, in this case. I no I'm heating things up when numbers are going 500watt and beyond. But not during the ASUS realbench stresstest.

Marcello_S
Level 7
Well if downclocked and everything runs smooth I highly doubt that PSU is the issue, because -50 MHz on both core and memory doesn't generate a delta of 50 W and by the data you posted I think PSU is ok and to those who aren't yet convinced check this video: https://www.youtube.com/watch?v=M30fAvrpj18 where to power up a 7600K and a 1070 SLI configuration has been used a 650 watt power supply with no problem at all.

I guess you have a faulty GPU at this point. Because if you need to underclock a GPU in order to make it working then there's something wrong.

JargonHolpick
Level 7
Hi guys,

I've been emailing ASUS about this issue and was adviced to change my 560watt PSU for a minimum of 600watt psu.
So I swapped my seasonic for a Corsair RM650x, not the best but sure good enough.
Guess what, same story with the realbench stresstest.

To pinpoint the issue I took a brandnew SSD, put a fresh Win10 pro edition on it and Realbench.
This is what came out of my test. (WIn10 pro complete stock + up to date updates, bios default settings except memory speed is set to 3000Mhz for all tests. All runs of Realbench stresstest are 15min 4GB - GPU running stock settings out of the box)

-WIn10 pro + RealBench stresstest == 5 runs all passed
-Win10 pro + Realbench stresstest + newest Nvidia driver == 5 runs all passed
-Win10 pro + Realbench stresstest + newest Nvidia driver + Asus GPU Tweak ver1473_20170627 == 5 runs all passed
Now it's getting interesting:
-Win10 pro + Realbench stresstest + newest Nvidia driver + Asus GPU Tweak ver1473_20170627 + AI Suite == 5 runs 4 passed 1failed

I guessed that AI suite might cause the issue so tried again without installation of AI Suite.
Cleared the SSD and reinstalled a clean (the same) WIndows version and did the same test again:

-WIn10 pro + RealBench stresstest == 5 runs all passed
-Win10 pro + Realbench stresstest + newest Nvidia driver == 5 runs all passed
-Win10 pro + Realbench stresstest + newest Nvidia driver + Asus GPU Tweak ver1473_20170627 == 3 runs all passed 2 failed
tried the test again without any change of software at all:
-Win10 pro + Realbench stresstest + newest Nvidia driver + Asus GPU Tweak ver1473_20170627 == 1 run passed 4 failed

The second to last fail I even got a black screen with arctifacts on it.
System was complete frozen.

So I might say that the issue is very random.

I had that black screen with arctifacts already before playing a game of RocketLeague.
Don't thing it's a heavy on the GPU game at all.

Anyone any ideas what might cause this mystery?

Menthol
Level 14
All your failures appear to be with GPU Tweak running, do you have the OSD active? OSD's can cause the failures

Do you have the Intel on die GPU enabled/disabled? due you have the Intel Graphics drivers installed? RealBench will use all available GPU's during the openCL test

Menthol wrote:
All your failures appear to be with GPU Tweak running, do you have the OSD active? OSD's can cause the failures

Do you have the Intel on die GPU enabled/disabled? due you have the Intel Graphics drivers installed? RealBench will use all available GPU's during the openCL test


This morning I tried the tests again without GPU Tweak and after some positif runs I also received some fails.
So I'm not sure that GPU Tweak is triggering this issue.
Only the 1080Ti is installed. No drivers installed for Intel's GPU . In the bios I've selected use only PCIe graphics card.
OSD are never on. If I monitor my hardware I prefer HwMonitor.

JargonHolpick
Level 7
I also had found some things about TdrDelay.

One time I ran AIDA64 stresstest and before the test started it said that it has to adjust some registrysettings to prevent issues.
I guess It put the registry TdrDelay because my fresh windows install, to run the tests on, doesn't have that registry in it.

I also wander what causes the arctifacts. Stresstesting might push some hardware pretty far but a game like RocketLeague, no, I don't think so.