Software glitch or faulty GPU hardware?

[fastfishy2]

I have an RTX 2070 Super, Overclocked slightly (+75 core clock, Power limit set to 109%) although the OC used to be higher but OCCT threw a small amount of errors when I tested it so I dropped it down to current settings.

Clock holds at 1995 mhz during Port Royal.

I'm encountering a strange lighting glitch in the 3DMark port royal stress test, where at a certain point every time I get a strange flickering effect on a select few surfaces, as if the shadows are being moved backward suddenly and then forward again.

No other benchmarks do this, RT or otherwise, including running Control on ultra with DLSS disabled.

I also did some testing with OCCT's GPU tester, unable to find any errors there, and I did the VRAM test too which again showed no errors (was using 95% VRAM). I also ran OCCT's gpu test in the background of the Port Royal test to see if it would pick up any errors but again, it did not. 

I guess I'm trying to figure out if this is a potential hardware issue or simply a driver issue? Is this in any way indicative of faulty hardware?

 

I'm going to go and try backdating my drivers now to see if it  fixes the problem - in the meantime I've attached a video file clipped to the exact point on the benchmark so that you can see what is happening.

Thanks for any and all help you can give.

[cr8tor]

What are you doing with your memory clocks?

 

You might run a log during the test and share it, see whats up in the background.

 

Download HWiNFO here - https://www.hwinfo.com/

 

How to create a log:

Open HWiNFO and choose sensors only.
In the bottom right there is an icon that looks like a spreadsheet with a green plus on it.

Click that button to start logging, the green plus will switch to a red X to let you know its saving data. 

It will also ask you for a file name and where to save it. I save them to my desktop, name it whatever you want.

Let it run for a few moments or until the machine locks up, resets, or does whatever it's not supposed to do.

Then (assuming it didn't restart or lockup) click the spreadsheet again so the red X turns back to a green +.

If it did lock or or reset, it will leave the log file where you specified.

You now have a log file, please share it.

 

How to view the log file:

You can view the log using the GenericLogViewer addon here https://www.hwinfo.com/add-ons/

Download, extract, run GenericLogViewer.exe, click the open file button in the top right and point it to the log file you made above.

Use the dropdown in the upper right to choose which category to view

[Mark Kaine]

well generally if you have glitches while running a benchmark it means oc not stable. there isn't really much to think about it... 

 

does the glitch occur with no OC?

yes = well that's strange

no = oc not stable 

[cr8tor]

2 minutes ago, Mark Kaine said:

it means oc not stable

Doesn't mean we cant make it stable.

Maybe he has some temp or voltage issues going on also. 

Never hurts to ask. Doesn't hurt to try. 

[fastfishy2]

9 minutes ago, cr8tor said:

What are you doing with your memory clocks?

 

You might run a log during the test and share it, see whats up in the background.

 

Download HWiNFO here - https://www.hwinfo.com/

 

How to create a log:

Open HWiNFO and choose sensors only.
In the bottom right there is an icon that looks like a spreadsheet with a green plus on it.

Click that button to start logging, the green plus will switch to a red X to let you know its saving data. 

It will also ask you for a file name and where to save it. I save them to my desktop, name it whatever you want.

Let it run for a few moments or until the machine locks up, resets, or does whatever it's not supposed to do.

Then (assuming it didn't restart or lockup) click the spreadsheet again so the red X turns back to a green +.

If it did lock or or reset, it will leave the log file where you specified.

You now have a log file, please share it.

 

How to view the log file:

You can view the log using the GenericLogViewer addon here https://www.hwinfo.com/add-ons/

Download, extract, run GenericLogViewer.exe, click the open file button in the top right and point it to the log file you made above.

Use the dropdown in the upper right to choose which category to view

I'll try that. I'm not doing anything with my memory clocks, they are stock. 

[fastfishy2]

7 minutes ago, Mark Kaine said:

well generally if you have glitches while running a benchmark it means oc not stable. there isn't really much to think about it... 

 

does the glitch occur with no OC?

yes = well that's strange

no = oc not stable 

happens when on stock settings as well

 

[fastfishy2]

9 minutes ago, cr8tor said:

Doesn't mean we cant make it stable.

Maybe he has some temp or voltage issues going on also. 

Never hurts to ask. Doesn't hurt to try. 

Log file attached.

flickr102.CSV

[cr8tor]

8 minutes ago, fastfishy2 said:

Log file attached.

Looking at it now.

Please post your pc specs also.

[fastfishy2]

4 minutes ago, cr8tor said:

Looking at it now.

Please post your pc specs also.

AMD Ryzen 5 3600 @ Stock speed (3.95 GHZ)

16gb (2x8gb) DDR4-3200mhz

Asus PRIME B450-Plus mobo

Gigabyte RTX 2070 Super Gaming OC 3X

Samsung 860 Evo 500gb (OS Drive)

WD Blue 2tb HDD

Samsung 970 Evo PLUS 1tb NVMe 

EVGA SuperNOVA 650 G3 PSU

 

any other info I can provide too.

[cr8tor]

Ok here is what i see.

First, all temps look good.

However, your 12v rail drops and stays a bit low, it does not look bad on its own.
This is a bit concerning though due to that bigger drop. On its own, its not alarming and within spec. But combined with everything. Its suspect.

The 5v rail is barely making 5v's and drops below 5v often.
This is within spec also, but concerning. On its on, not a big deal, combine with problems and other voltage issues its concerning also.

 

GPU input voltage. Drops and just keeps going down.

 

Im betting you can see what im getting at.

If this was my system, i would not hesitate to order a new PSU.

I believe its power irregularities are causing your stuttering.

[fastfishy2]

Just now, cr8tor said:

Ok here is what i see.

First, all temps look good.

However, your 12v rail drops and stays a bit low, it does not look bad on its own.
This is a bit concerning though due to that bigger drop. On its own, its not alarming and within spec. But combined with everything. Its suspect.

The 5v rail is barely making 5v's and drops below 5v often.
This is within spec also, but concerning. On its on, not a big deal, combine with problems and other voltage issues its concerning also.

 

GPU input voltage. Drops and just keeps going down.

 

Im betting you can see what im getting at.

If this was my system, i would not hesitate to order a new PSU.

I believe its power irregularities are causing your stuttering.

oh shoot really?! I'll get to it right away then. It's gonna be a massive pain to replace though.

 

[cr8tor]

24 minutes ago, fastfishy2 said:

oh shoot really?! I'll get to it right away then. It's gonna be a massive pain to replace though.

 

I get that, sucks.

But from the logs i feel confident in the diagnosis. 
Like i said, if it was my system and i saw that, i'd cuss and order a power supply.

[cr8tor]

Also, assuming this does fix it, lemme know. 
I like to know for sure when diagnostics are accurate. 

[fastfishy2]

40 minutes ago, cr8tor said:

Also, assuming this does fix it, lemme know. 
I like to know for sure when diagnostics are accurate. 

If it doesn't fix it the only other thing I might try is getting an Asus 2070 super on discount. Literally at the shop picking up a corsair RM750X RIGHT NOW lol. Undervoltage shouldn't have damaged anything right? Only over voltage does that aye 

[cr8tor]

6 minutes ago, fastfishy2 said:

Undervoltage shouldn't have damaged anything right? Only over voltage does that aye 

Correct. 

I'll be on for a while so let us know.

[fastfishy2]

4 hours ago, cr8tor said:

Correct. 

I'll be on for a while so let us know.

Dang... new PSU fitted and it still does it after the 1st run. I have a log file that covers most of the 2nd run where it started happening again. I think I'll just order a replacement GPU as this one is outside warranty because I replaced the fans, and also it is unlikely my local retailer will refund for one test.

If the new one does it also, then at that point I'll just be happy Port Royal isn't one of my games. I do like the look of them Asus 2070 Supers anyway and they're on big discount over here.

 

EDIT: OR i might just keep this one because if it only shows that ONE glitch at that ONE part of that ONE benchmark and not in anything else then it's unlikely to be a hardware fault, or at least not one that actually matters, because as I said it's not like I play that benchmark as a game, right? Also I've attached another log file, from a complete Port Royal cycle that didn't show the flickering at that one scene (flickr104)

flickr105.CSV

flickr104.CSV

Edited November 26, 2020 by fastfishy2
Edit / update and additional log file