I may or may not have a bad 2080Ti FE. How do I check?

T

Thunderdolt

Gawd
Joined
Oct 23, 2018
Messages
1,015
Ever since moving to 2080Tis, my GPU-focused apps haven't been as stable as they were on my old Titan Xps. Maybe it's just a driver thing, maybe it's just all in my head.... but maybe one or more of the cards is flaky.

What's a good way to find out which card, if any, is the culprit? Just play games a bunch and connect my monitor to a different card every other day? Is there a more reliable check?
 
Furmark might get it to shit the bed. Also run any freely downloadable benchmark like Heaven and beat the shit out of it at max settings. If you experience any artifacts you know its bad. If the screen goes nuts and whites out on you thats another indicator. If your video signal dies amd the PC is fine thats typically another (assuming youre already rocking high end cables from the card to the video monitor)

HWinfo will also tell you stuff like if the card is lighting on fire under load (furmark will tell tou this during the testing as well) . Gpuid isnt bad to have om there either. I have had 2 bad cards. Maybe, possibly, who the fuck knows... My third card is wonky by its still alive. I dont consider it 100% stable but it hasnt ahit the bed yet.
 
So, run some benchmarks to give it load and look for artifacts? Guessing this means I should be swapping my monitor to different cards and re-running the benchmarks?
 
remove it from your case and set it on a clear, uncluttered, static-free desk or work area. place a $5 bill in front of it and leave the room for a few minutes. come back and check if the money is gone. if it is, give the card a chance by saying "where do my money go? you didnt decide to hold it for me did you?" pause for a few moments. if there is no response then lift up the card. you will find the $5 bill underneath. this is evidence of a bad card.
 
purple_monster said:
remove it from your case and set it on a clear, uncluttered, static-free desk or work area. place a $5 bill in front of it and leave the room for a few minutes. come back and check if the money is gone. if it is, give the card a chance by saying "where do my money go? you didnt decide to hold it for me did you?" pause for a few moments. if there is no response then lift up the card. you will find the $5 bill underneath. this is evidence of a bad card.
Click to expand...

What if the $5 bill is still there when I come back, but the room now smells of the marijuanas?
 
Thunderdolt said:
Ever since moving to 2080Tis, my GPU-focused apps haven't been as stable as they were on my old Titan Xps. Maybe it's just a driver thing, maybe it's just all in my head.... but maybe one or more of the cards is flaky.

What's a good way to find out which card, if any, is the culprit? Just play games a bunch and connect my monitor to a different card every other day? Is there a more reliable check?
Click to expand...
On a serious note, can you be more specific? What apps are crashing? Is there anything that can reliably make it crash?

Is there any chance the problem could be a power supply getting old? Do you have a spare you can test with?

Also, this might be a stupid question, but, do you have these cards overclocked beyond whatever they come with from the factory? I've found that my factory-water-jacket-equipped 1080 Ti was already pushed pretty hard from the factory, and only has another 50ish MHz before it starts to misbehave. I could see 2080 Tis being even worse about this, especially if there's two.
 
RazorWind said:
On a serious note, can you be more specific? What apps are crashing? Is there anything that can reliably make it crash?

Is there any chance the problem could be a power supply getting old? Do you have a spare you can test with?

Also, this might be a stupid question, but, do you have these cards overclocked beyond whatever they come with from the factory? I've found that my factory-water-jacket-equipped 1080 Ti was already pushed pretty hard from the factory, and only has another 50ish MHz before it starts to misbehave. I could see 2080 Tis being even worse about this, especially if there's two.
Click to expand...

I have a lot of stability issues when using apps based on Iray, like Solidworks Visualize and SketchUp's render module. During renders, the screen will go black for a couple of seconds and then come back, at which point I'm informed that the render has failed. This has persisted through a few driver versions and I always do a clean install when updating those, and it is consistent for software versions going back ~2yrs. This happened zero times when I ran the same software on my former dual Titan Xp machine. That's not a super useful reference point of course due to the obvious architecture change.

I don't have a spare PSU to test with, unfortunately. The good news is that the one I'm using is only about 6mos old, so anything age-related shouldn't be hitting it just yet, short of a bad component.

These cards have not been overclocked. I went with water on them because of both physical constraints and thermal density when plunking four of them together on the same ATX board.
 
purple_monster said:
remove it from your case and set it on a clear, uncluttered, static-free desk or work area. place a $5 bill in front of it and leave the room for a few minutes. come back and check if the money is gone. if it is, give the card a chance by saying "where do my money go? you didnt decide to hold it for me did you?" pause for a few moments. if there is no response then lift up the card. you will find the $5 bill underneath. this is evidence of a bad card.
Click to expand...

omg this is great
 
Reading between the lines of your post, can you confirm that:
1. you have four cards in use
2. You're doing non-game things
3. All four are water cooled, presumably in a single gigantic loop?

Are these early Founder's Edition cards? There was a known QC issue with the first couple of batches that normally manifested as a funky pattern on the screen, but that would obviously not be applicable to applications that aren't games.

Is there any way to test each one individually with your rendering software? The fact that the screen goes black suggests the bad card is the one with the display plugged into it, so I'd start with that one.

I might also suggest that you try playing some games on each card, and see if you get any symptoms of misbehavior in a game.
 
Thunderdolt said:
I have a lot of stability issues when using apps based on Iray, like Solidworks Visualize and SketchUp's render module. During renders, the screen will go black for a couple of seconds and then come back, at which point I'm informed that the render has failed. This has persisted through a few driver versions and I always do a clean install when updating those, and it is consistent for software versions going back ~2yrs. This happened zero times when I ran the same software on my former dual Titan Xp machine. That's not a super useful reference point of course due to the obvious architecture change.

I don't have a spare PSU to test with, unfortunately. The good news is that the one I'm using is only about 6mos old, so anything age-related shouldn't be hitting it just yet, short of a bad component.

These cards have not been overclocked. I went with water on them because of both physical constraints and thermal density when plunking four of them together on the same ATX board.
Click to expand...

What are you waiting for, RMA the shit out of that card or return it for another that doesn't drive you ape shit crazy. The issues you describe are 2080Ti related. You can spend the next 6 months unhappy or you can just do something about it that might yield another, better result (hopefully). If your old setup worked... As had with my old 1080Ti prior to getting the 2080Ti ... It's likely the card. A lot of us have seen a shitload of errors on this series of cards and they have (for a long time) a 20% +/- failure rate. Kyle Bennett even confirmed this was true.
 
RazorWind said:
Reading between the lines of your post, can you confirm that:
1. you have four cards in use
2. You're doing non-game things
3. All four are water cooled, presumably in a single gigantic loop?
Click to expand...

1) Correct - there are currently four cards installed in the machine.
2) Correct - I'm doing renders and FEA.
3) All four are water cooled in two standalone loops. You can see the build details here.

RazorWind said:
Are these early Founder's Edition cards? There was a known QC issue with the first couple of batches that normally manifested as a funky pattern on the screen, but that would obviously not be applicable to applications that aren't games.
Click to expand...

Yes, these are early edition pre-order cards. I don't recall off the top of my head if their serials are below the threshold that the internet collectively decided was the line between good and bad.

RazorWind said:
Is there any way to test each one individually with your rendering software? The fact that the screen goes black suggests the bad card is the one with the display plugged into it, so I'd start with that one.
Click to expand...

With Octane, I can pick and choose the cards that get used. For the other packages, I guess I can selectively disable cards using Nvidia control panel.

The screen going black is consistent with what happens during driver updates. With the driver updates, it goes through four black screen cycles (which I presume means one per card) as one of the final steps of the install. That's part of what makes me believe there is a driver crash happening. Of course, both bad drivers and bad hardware can cause driver crashes.


RazorWind said:
I might also suggest that you try playing some games on each card, and see if you get any symptoms of misbehavior in a game.
Click to expand...

This sounds like my best shot. I can play some games and see what happens. Since I'm getting paid for this, does this technically mean that I'm a professional gamer?

Are there any games which are known to be particularly likely to exhibit artifacting on bad cards? Or am I safe just sticking to catching bullets with my face in PUB?
 
purple_monster said:
remove it from your case and set it on a clear, uncluttered, static-free desk or work area. place a $5 bill in front of it and leave the room for a few minutes. come back and check if the money is gone. if it is, give the card a chance by saying "where do my money go? you didnt decide to hold it for me did you?" pause for a few moments. if there is no response then lift up the card. you will find the $5 bill underneath. this is evidence of a bad card.
Click to expand...

This works I can attest.
 
Thunderdolt a good stressful game that brings a card to it's knees is Kingdom Come: Deliverance. It is using the most modern Crytech engine
 
You must log in or register to reply here.
Back
Top