Question Help! 3DMark stress test failing with RTX 3090

tabur

Distinguished
Dec 19, 2009
45
1
18,535
Hi all, I've been experiencing crashes in multiple games (RDR2, Horizon Zero Dawn, others) so I figured there has to be something wrong with my system. So I started stress testing different components.

When stress testing my CPU with Prime95, it was largely fine but after a while every window turned black, so I had to restart explorer.exe. Besides that, no crashes or BSODs.

Given the weird black windows, I suspected something could be wrong with my GPU, so I downloaded 3DMark and launched Time Spy Stress Test. After a few runs, the test just suddenly ends/crashes and I get a 0%. I looked at the event logs, and I did get an Error and a Warning.

The error shows the following:

Code:
The description for Event ID 0 from source nvlddmkm cannot be found. Either the component that raises this event is not installed on your local computer or the installation is corrupted. You can install or repair the component on the local computer.

If the event originated on another computer, the display information had to be saved with the event.

The following information was included with the event:

\Device\Video3
Error occurred on GPUID: a00

The message resource is present but the message was not found in the message table

And the warning shows the following:

Code:
Display driver nvlddmkm stopped responding and has successfully recovered.

Here are links to the actual logs and a picture with 3DMark's results:
Logs: https://drive.google.com/file/d/15P6tn-UOph2gQI4l1CCXvq6gJBMAXDi6/view
Stress test result pic: https://drive.google.com/file/d/1OLQYfhfrA-hpKzZwkz7oo1_Jjx9Vn0Vh/view

My systems specs are:
Asus ROG Crosshair VIII Dark Hero
AMD Ryzen 5900X
AORUS GeForce RTX™ 3090 XTREME WATERFORCE 24G
CORSAIR Vengeance LPX 64GB (2 x 32GB) 288-Pin PC RAM DDR4 3200 (PC4 25600) Intel XMP 2.0 Desktop Memory Model CMK64GX4M2E3200C16
Corsair SF Series, SF750, 750 Watt, SFX, 80+ Platinum Certified, Fully Modular Power Supply (CP-9020186-NA)

-> Everything is at stock levels, no overclocking at all. <-

Any ideas about what could be wrong with my system? How can I diagnose what's wrong with my PC?

Thank you very much!
 

Lutfij

Titan
Moderator
AORUS GeForce RTX™ 3090 XTREME WATERFORCE 24G
Corsair SF Series, SF750, 750 Watt, SFX, 80+ Platinum Certified, Fully Modular Power Supply (CP-9020186-NA)

You've got the wrong PSU for your build. You're advised to have a reliably built 1KW or higher PSU to power the entire build due to transient load spikes by the GPU.
 
  • Like
Reactions: cAllen and ötzi
Definitely a power issue.
If the OP needs a SFF-ish PSU (for whatever reason) they should check out the EVGA P6 1000W. It's not exactly SFX dimensions but it's smaller than most standard sized PSUs and has been working flawlessly in my set up (in sig).
 

tabur

Distinguished
Dec 19, 2009
45
1
18,535
Oh wow, I didn't know. Gigabyte's website said the recommended PSU was 750W, which is why I bought this one. The math worked out, but I didn't know about the spike issue. Any way I can test this?

The reason for the PSU form factor is my case: LIAN LI O11D MINI-W White SPCC

I'm happy to change the PSU and/or the case. In fact, one problem with this case is that it's quite wide, so it doesn't fit in most under-desk PC holders. I would like to stick to a smaller form factor though.

Is there a PSU you guys would recommend? Hopefully that fits this case, but if not, I think I'm OK changing cases as well (recommendations are welcome of course)

Thank you very much for the prompt and detailed replies, these are much appreciated :)
 
Oh wow, I didn't know. Gigabyte's website said the recommended PSU was 750W, which is why I bought this one. The math worked out, but I didn't know about the spike issue. Any way I can test this?
With regards to the transient power spikes, it's not a question of if, it's a question of when will you experience them. Unfortunately, they are difficult to test for because they only happen for a split second.

That 750W is entirely inadequate to support those spikes. You'll have to find a better SFX/SFX-L PSU, or, experiement with undervolting your GPU. You may be able to tame it enough so that the spikes aren't as great and are still covered by what the PSU can handle. You will lose performance but, it may not be that much - it's worth a try.
 
  • Like
Reactions: Phaaze88
Sep 10, 2022
113
17
95
Oh wow, I didn't know. Gigabyte's website said the recommended PSU was 750W, which is why I bought this one. The math worked out, but I didn't know about the spike issue. Any way I can test this?

The reason for the PSU form factor is my case: LIAN LI O11D MINI-W White SPCC

I'm happy to change the PSU and/or the case. In fact, one problem with this case is that it's quite wide, so it doesn't fit in most under-desk PC holders. I would like to stick to a smaller form factor though.

Is there a PSU you guys would recommend? Hopefully that fits this case, but if not, I think I'm OK changing cases as well (recommendations are welcome of course)

Thank you very much for the prompt and detailed replies, these are much appreciated :)

Get at 1000w PSU, only a touch more expensive than 850w and you'd have more room for future upgrade, considering you got the money to own 3090 now LOL.

I mean, I have 6800xt, and I am using 850w, people say 750w should be enough, I said, will I have problem running bigger PSU? No. So yeah, get 1000w.

To really really make yourself confident if your PSU is the culprit, I wonder, does NVIDIA PPT can be adjusted? Because on AMD Adrenaline I could increase or decrease the power limit by significant value.

Try to decrease your PPT if you can, to the lowest possible number. If that solve the issue, or at least, if it doesn't happen as frequent, then yeah your PSU is not enough.

Which PSU? depends on where you are, but for example, in where I live, EVGA, Seasonic, and Be Quiet is quiet popular, Corsair too but the RMx
 

tabur

Distinguished
Dec 19, 2009
45
1
18,535
Get at 1000w PSU, only a touch more expensive than 850w and you'd have more room for future upgrade, considering you got the money to own 3090 now LOL.

I mean, I have 6800xt, and I am using 850w, people say 750w should be enough, I said, will I have problem running bigger PSU? No. So yeah, get 1000w.

To really really make yourself confident if your PSU is the culprit, I wonder, does NVIDIA PPT can be adjusted? Because on AMD Adrenaline I could increase or decrease the power limit by significant value.

Try to decrease your PPT if you can, to the lowest possible number. If that solve the issue, or at least, if it doesn't happen as frequent, then yeah your PSU is not enough.

Which PSU? depends on where you are, but for example, in where I live, EVGA, Seasonic, and Be Quiet is quiet popular, Corsair too but the RMx

I'm in the US, any suggestions for an SFX or SFX-L PSU that would work? It seems like there aren't many 1000w PSUs in that form factor... And I don't plan on upgrading this PC ever, so future-proofing isn't necessary, so if you think 850w would suffice, I'd go for that.

Also, what's "NVIDIA PPT"?
 
Sep 10, 2022
113
17
95
I'm in the US, any suggestions for an SFX or SFX-L PSU that would work? It seems like there aren't many 1000w PSUs in that form factor... And I don't plan on upgrading this PC ever, so future-proofing isn't necessary, so if you think 850w would suffice, I'd go for that.

Also, what's "NVIDIA PPT"?

PPT is a power limit. With AMD i Got this setting where I can limit the power by -15 or +15 percent from stock. So if stock comes at 200w, then I could set my gpu not to hit more than 170 or to hit up to 230w. That’s the way to explain that. In cpu both intel and amd we can also adjust this power limit. So I am kinda curious how this will troubleshoot your problem.

I did quick search and confused on how to set power limit for this card it seems like NVIDIA may not allowing users to do it. While it can be useful to test this case.

And as in PSU, you know it better you have the access to information as much as I do so pick what’s best for you. But first, to really sure it’s psu related problem not being able to feed the gpu enough power. I believe it’s best for you to find ways to test it by REDUCING power limit if there is a way at all.
 

tabur

Distinguished
Dec 19, 2009
45
1
18,535
Btw, for anyone looking into this - I got myself a Silverlake 1000W PSU, it did not solve the problems.... 3DMark extreme stress test lasts a bit longer now, but still fails and Horizon New Zero Dawn crashes as much as before.
 

Phaaze88

Titan
Ambassador
What hasn't been covered yet..?

Using any psu cable extentions or riser devices? Remove them if so.
PC is plugged in to a UPS? Try plugging it directly to the wall outlet.

How do the games crash?
A)Black screen to gpu fans revving up like a jet engine.
B)Black screen, can still hear game sounds, but no jet engine gpu.
C)Other: Please describe.
 

tabur

Distinguished
Dec 19, 2009
45
1
18,535
What hasn't been covered yet..?

Using any psu cable extentions or riser devices? Remove them if so.
PC is plugged in to a UPS? Try plugging it directly to the wall outlet.

How do the games crash?
A)Black screen to gpu fans revving up like a jet engine.
B)Black screen, can still hear game sounds, but no jet engine gpu.
C)Other: Please describe.

My computer was indeed connected to a UPS, but I just now connected everything to the wall outlet directly and got the same crash in HZD, so I'm thinking that's not the issue.

Not using any PSU cable extensions. GPU attached directly to the motherboard.

In terms of the crash, it's kinda weird... nothing flashy or anything, and it seems like it's related to parts of the game and video configurations. I can get the same crash in the same place multiple times, but then all of the sudden, the game will run flawlessly for 20 mins straight... and then crashes... Please see below a snapshot of the message I get. So no black screen, can still hear audio, and an in-game error message appears (pic below). If I don't click OK fast enough, I get the typical "Application has stopped responding" windows message. The GPU fans are indeed revved up blowing very hot air, but Open Hardware Monitor indicates that temps never go over 73C and power over 349W. My CPU (Ryzen 5900) does get hot, up to 90C, but I hear this isn't too hot for this package.

Screenshot-2022-10-27-214452.png


Any ideas on how to diagnose this?

Also, I ran 3DMark Time Spy Extreme stress test, and it'll crash before finishing. It'll do 10 loops or so, and then it'll say the test failed and get me a 0%. I doubt these 2 issues are unrelated, but who knows!
 

Phaaze88

Titan
Ambassador
Also, I ran 3DMark Time Spy Extreme stress test, and it'll crash before finishing. It'll do 10 loops or so, and then it'll say the test failed and get me a 0%. I doubt these 2 issues are unrelated, but who knows!
Does the app at least tell you what failed and where?
Userbenchmark gives a hint when a device fails its part of the benchmark, it won't have a score next to it. You should close the browser and other monitoring software before running it. Post the public link to your results.


In terms of the crash, it's kinda weird... nothing flashy or anything, and it seems like it's related to parts of the game and video configurations. I can get the same crash in the same place multiple times, but then all of the sudden, the game will run flawlessly for 20 mins straight...
I've only seen that happen with unstable gpu overclocks. That doesn't mean it's the only plausible cause though.
I'd suggest taking Msi Afterburner - or you can use Gigabyte's software if you want - and reduce the core clock 200-ish mhz and see if the crashing stops.
 

tabur

Distinguished
Dec 19, 2009
45
1
18,535
Does the app at least tell you what failed and where?
Userbenchmark gives a hint when a device fails its part of the benchmark, it won't have a score next to it. You should close the browser and other monitoring software before running it. Post the public link to your results.



I've only seen that happen with unstable gpu overclocks. That doesn't mean it's the only plausible cause though.
I'd suggest taking Msi Afterburner - or you can use Gigabyte's software if you want - and reduce the core clock 200-ish mhz and see if the crashing stops.

Thank you for your help. I'll try this later tonight or over the weekend and report back
 

tabur

Distinguished
Dec 19, 2009
45
1
18,535
Does the app at least tell you what failed and where?
Userbenchmark gives a hint when a device fails its part of the benchmark, it won't have a score next to it. You should close the browser and other monitoring software before running it. Post the public link to your results.



I've only seen that happen with unstable gpu overclocks. That doesn't mean it's the only plausible cause though.
I'd suggest taking Msi Afterburner - or you can use Gigabyte's software if you want - and reduce the core clock 200-ish mhz and see if the crashing stops.

This is the result I got from Userbenchmark: https://www.userbenchmark.com/UserRun/56285265

I can't say I understand it though, do you see something here that's worrisome?
 

tabur

Distinguished
Dec 19, 2009
45
1
18,535
No overclocks or overvolts, just a slight CPU undervolt offset to avoid high temps (I have the crash issues with and without the undervolt). I also did a BIOS reset, same issue. The only thing I haven't tried is lowering the RAM frequency, I use the XMP default profile (I actually don't know how to change JUST the frequency). Do you think that could be related to this issue?

I just find it weird that 3DMark "Time Spy" stress tests, HZD, and RDR2 all fail seemingly due to graphics issues
 
No overclocks or overvolts, just a slight CPU undervolt offset to avoid high temps (I have the crash issues with and without the undervolt). I also did a BIOS reset, same issue. The only thing I haven't tried is lowering the RAM frequency, I use the XMP default profile (I actually don't know how to change JUST the frequency). Do you think that could be related to this issue?

I just find it weird that 3DMark "Time Spy" stress tests, HZD, and RDR2 all fail seemingly due to graphics issues
Let's do separate component tests. Keep BIOS at default but watch temps. If you are undervolting too much you can introduce CPU instability. What CPU cooler are you using?

Bootable Memtest86 USB to test RAM. At least 1 full pass to confirm decent stability.
Prime 95 (inside OS). 15 minute blended test with all AVX off.
Furmark GPU stress at least 15 mins.
 

tabur

Distinguished
Dec 19, 2009
45
1
18,535
Let's do separate component tests. Keep BIOS at default but watch temps. If you are undervolting too much you can introduce CPU instability. What CPU cooler are you using?

Bootable Memtest86 USB to test RAM. At least 1 full pass to confirm decent stability.
Prime 95 (inside OS). 15 minute blended test with all AVX off.
Furmark GPU stress at least 15 mins.

Hi, I ran Memtest86 and it failed due to too many errors. In particular, it seems like "test 6" is the one that fails constantly, the others seem OK (results below). And even though I've got a pair of dual channel Corsair Vengance LPX, Memtest86 only detected SPD settings for 1 of the sticks (pic below too). Any ideas about what could be causing this?

results-screenshot.png



Timings:
mt86-Screen-20221211-183144.png
 
Last edited:

tabur

Distinguished
Dec 19, 2009
45
1
18,535
UPDATE: For some incredibly odd reason, I swapped the sticks' positions and now all Memtest86 tests pass, and even 3DMark's Time Spy extreme stress test with a whooping 97.5% score!! Hooray!!

I tested each stick separately, in each slot, and things worked. Which was weird. Then I tested both at the same time, and things worked too!! I guess some pads weren't making solid contact, or perhaps there was some sort of static charge or spurious current? Not sure, but now things work. I'll test this out with Horizon and report back shortly. Thanks a lot for your help guys, in particular @alceryes , you're the MVP!
 
  • Like
Reactions: alceryes