What else can I do to diagnose an artifacting card?

[boppingXQWB]

Greetings! First time posting although I've been a longtime reader of TPU.

tl;dr: I had sold my XTX but the buyer had issues with it so I've refunded them. I can't seem to replicate the issue in order to RMA the card.

I had recently sold my 7900 XTX in anticipation of 9070 XT for better raytracing perf, but the buyer had issues with the card running COD Black Ops 6 multiplayer so I refunded them and got the card back.

Buyer's videos showing the artifacting are attached below:

http://imgur.com/BYTPOMD

http://imgur.com/b9zlbu2

We had many back-and-forths trying to diagnose/fix the issue, but in the end we were not able to get BO6 from running properly. The followings were what we've tried:

• Reinstalling the driver
• Reinstalling the game
• Trying different versions of the game: Xbox Game Pass & Battle.net
• Installing the card in different machines (buyer tried installing the card in a friend's* machine)

*this same friend also owns an XTX, so the buyer tried that in his machine and had no issue, while the XTX I had sold was artifacting in this 2nd machine as well. The buyer said the card would start artifacting in around 30 mintutes in his PC. Card initially seemed fine in his friend's PC but apparently it had issues around 2 hours in.

Well, OK, seems like a VRAM problem. Simple right? Well, after getting the card back, I ran a number of tests on my own, but I just can't seem to get it to misbehave. Which would be fine if I wanted to keep the card, but I want to sell it, and knowing the card was artifacting in someone else's computer (well two computers, buyer + the friend's), I can't sell the card again in good conscience. Instead, I'm hoping to find out exactly when and how the card artifacts so I can RMA it.

Tests I've done so far:

• Memtest Vulkan (4+ hrs)
• OCCT VRAM stability test (1hr)
• OCCT GPU stability tests (1hr each: 3D standard, 3D adaptive steady, 3D adaptive variable, 3D adaptive switch)
• Memtest Vulkan + various OCCT GPU stability tests (1hr each)
• COD Black Ops 6 multiplayer (3 hrs)
• COD Black Ops 6 single player (3 hrs)

Thought I wasn't having any issues because I was running the card in an open setup, but even putting a cardboard box over the setup to cook the card to 95 hotspot and ~99 VRAM didn't cause any issues in Memtest Vulkan + OCCT GPU stability test (~40 minutes)

I don't want to send the card in for RMA and have Sapphire send it back without doing anything about it. What am I missing here?

Additional info/thoughts, will add more later if I think of anything:

• Tried both BIOSes on the card
• Buyer only had issues with BO6 and no other games

 

[Epaminombas]

If it is under warranty, send it to the manufacturer.

But before that...

Did you change the Vcore? Did you use MSI Afterburner or any similar software that changes the card's settings?

What motherboard do you have? Does it have an updated BIOS?

What Windows? Did you download the original version from Microsoft?

Is the CPU overclocked?

Is there a missing driver in the device manager?

Does the GPU firmware have timing for other types of memory?

Does Windows Event Viewer show anything?

 

[Carlyle2020hs]

Vram has its fingers in everything - so bad Vram would screw with everything, yes?

What i see are clear as day 2D Names and a clear map with clear details that are not affected at all.

Combined with a steady framrate ... except in the second video the fps goes up quiet a bit before the 3D Part gets screwy and the fps drops and stays low.

Without the added info i would have bet that the issue is software related. Like 2 packs of Oreos kinda heavy bet.

So if it´s not you it´s the setup?

Turn off your case intake fans and try again?

 

[boppingXQWB]

If it is under warranty, send it to the manufacturer.

But before that...

Did you change the Vcore? Did you use MSI Afterburner or any similar software that changes the card's settings?

What motherboard do you have? Does it have an updated BIOS?

What Windows? Did you download the original version from Microsoft?

Is the CPU overclocked?

Is there a missing driver in the device manager?

Does the GPU firmware have timing for other types of memory?

Does Windows Event Viewer show anything?

IDK, I don't have access to the buyer's computer. GPU was ran stock. Buyer also had no issues with another XTX as stated above.

So if it´s not you it´s the setup?

Turn of your case intake fans and try again?

It cooould be, but both the buyer and his friend had issues with the card and neither of them had issues with the friend's XTX, so it's not the setup. I'm guessing the intake fan suggestion is for me. I've already cooked the system with a cardboard box over it.

 

[Epaminombas]

then it's his problem.

He needs to update the BIOS

Install Windows 11 24H2 from 0. Format completely to clear cache and drivers of old GPUs.

 

[boppingXQWB]

I've moved it inside a small case (22L, Sama IM01) as my daily GPU for the last 5 days, and did over 5 hours of Memtest Vulkan and still nothing. Any other ideas, folks?

Getting tired of trying to get the card to misbehave and make a record of it. Has anyone had experience with Sapphire's RMA? Are they thorough? I don't want to be charged (on return shiping or service or w/e) after they don't manage to find anything wrong.

Maybe I should just sell it again with a disclaimer

 

[eidairaman1]

I've moved it inside a small case (22L, Sama IM01) as my daily GPU for the last 5 days, and did over 5 hours of Memtest Vulkan and still nothing. Any other ideas, folks?

Getting tired of trying to get the card to misbehave and make a record of it. Has anyone had experience with Sapphire's RMA? Are they thorough? I don't want to be charged (on return shiping or service or w/e) after they don't manage to find anything wrong.

Maybe I should just sell it again with a disclaimer

Probably a problem with their monitor or cable, if you cant reproduce it try a different monitor/cable, put a gaming load on it.

It could be a particular game even.

 

[boppingXQWB]

Thanks for the ideas. I've tried the same game (COD BO6) for three approximately-3-hour sessions without issues, though that was on the 2nd open-case computer, not after I've put it inside the case in my main computer. I suppose I should try running the game again now that it's in a case.

As for the buyer's monitor/cable potentially being the culprit, that wouldn't selectively affect different elements in the game, so that's probably not it.

One last possible culprit that i just thought of is, while I didn't dwell on it at the time of installing the card in my case, I noticed that due to the PCIe bracket being triple-slot and having more resistance/catching on the case-side PCIe brackets, the card was not fully inserted into the PCIe slot. I corrected this after noticing before installing the PCIe bracket screws, but now that I think about it, I wonder if a non-fully-inserted video card could cause an issue that the buyer was observing. While it worries me that poor contact on the PCIe slot's power-lane gold fingers maaaay cause damage to the card, it's 75W max so the heat generated shouldn't be too bad for a brief testing session, so perhaps I should give that a go. Who knows, maybe the PC won't even POST.

 

[rusty caterpillar]

Thanks for the ideas. I've tried the same game (COD BO6) for three approximately-3-hour sessions without issues, though that was on the 2nd open-case computer, not after I've put it inside the case in my main computer. I suppose I should try running the game again now that it's in a case.

As for the buyer's monitor/cable potentially being the culprit, that wouldn't selectively affect different elements in the game, so that's probably not it.

One last possible culprit that i just thought of is, while I didn't dwell on it at the time of installing the card in my case, I noticed that due to the PCIe bracket being triple-slot and having more resistance/catching on the case-side PCIe brackets, the card was not fully inserted into the PCIe slot. I corrected this after noticing before installing the PCIe bracket screws, but now that I think about it, I wonder if a non-fully-inserted video card could cause an issue that the buyer was observing. While it worries me that poor contact on the PCIe slot's power-lane gold fingers maaaay cause damage to the card, it's 75W max so the heat generated shouldn't be too bad for a brief testing session, so perhaps I should give that a go. Who knows, maybe the PC won't even POST.

The tails of the bracket is slightly bent? Inspecting the fingers is a good idea.
I believe buyer case is more crammed than yours and less ventilated maybe causing high temps, VRAM or some power delivery components to misbehave and causes artifacts. VRAM over 95 C can be unstable.

Sapphire RMA is out of the question. Don't sell it with a disclaimer, I would take the risk. If is working fine in your case and only in his not....
Test at least 5 games though at full quality. Try all the games in closed case ONLY. You don't expect buyers to run your card in an open bench or open case do you?

Also next time show us averages of temps not current.
Hope it helps

 

[boppingXQWB]

Oh, no no. They are not bent. I was just saying that the card having a three-slot bracket may have made it a little more difficult to insert all the way into the PCIe slot, and perhaps that's what happened with the buyer (causing artifacting from unstable PCIe connection or something). Just guesses, since I can't get the card to artifact in games or error out on Memtest Vulkan.

Card has been installed in a case for over a week now as my daily GPU. Tried a number of games so far, but I'll keep testing.

 

[tpa-pr]

Bit of an out there suggestion but it turns out I was having more GPU issues than I should've because of my mainboard's poor ability to auto-negotiate PCI-E speeds. After setting them manually in the BIOS games that used to crash before no longer crash (the system even POSTs faster). So it could be specific to their mainboard?

 

[FreedomEclipse]

My experience tells me that thats a GPU core issue. The core clock is too high for it to run stable. It doesnt look like memory instability issues but other folks may have their own opinion on the matter.

Try and downclock the core by like 200mhz and see what happens. Otherwise I would be looking at sending the card back for a refund or replacement ideally.

 

[arbiter]

Looking at the video i don't think that is a card is bad problem, given the game HUD isn't effected seems bad drivers or corrupted textures of the game.

 

[Todestrieb]

There was a thread of me trying to figure out something on a defective PowerColor HellHound 7900XTX...

Trying to sort out instabilities after my GPU upgrade.

I can see the NEW PSU!! argument, and maybe 4080 will be kinder to my PSUs, but it is what it is... Update: Flipped the PCIe link state switch to No Power savings to stress more on the PSU. Situations that otherwise will instant reboot are now fine. It is ridiculously hot here for December (I...

www.techpowerup.com

Looking at the videos, I'm pretty sure it is indeed a defective card.
I think at this point sending the card to reseller / manufacturer for warranty is required to fix whatever minor issue is in your card.

If you have other things to try, be sure that it loads up the GPU heavily. Light games like WWE2K series or some indie games are much less likely to crash.
Also, the usual benchmarks / stress tests like the 3DMark stuff are very unlikely to crash. I think I have crashed once on Port Royal, but I can't crash it for a second time.
IIRC, OCCT also didn't crash my card. Not sure about memtest Vulcan; I don't think it was public / well known enough when I was dealing with the defect.

Here's a list of all things I have semi-consistently crashed back when my card was defective.
Your card / config might end up safe in these games / applications.
All games / applications I have tried were at 4K, as ultra as possible settings, frame limit disabled whenever applicable. GPU driver was always on default settings.

Spoiler: list

- Forza Horizon 5 / Forza Motorsport 8: environment texture disappears, making the game unplayable because the road / world becomes invisible. At most 1hr, usually a few minutes into the game before such artifacts appear in my case. Also may cause a driver timeout crash. FH5 is tested to be bad after a heavy downclock, and bad at 1440p on my brother's similarly configured rig.
- The Crew Motorfest: Driver timeout crash, with the game executable throws a "GPU lost" error message. Normally occurs within 1hr.
- F1 Manager 2023: Cannot finish a race without a driver timeout crash. If you decide to try this one, stay at the screen where the "race" is fully rendered. That is, 2x speed at most.
- folding@home: Not a game. Loads up the GPU by compute load. Depends on RNG, some WUs are completely fine. Some will throw the whole PC into a driver timeout crash loop, or worse, reboot loop. Don't worry, folding@home will sort itself out by discarding the "bad" WU after several crashes.

Good luck on finding something to crash your card and sending the card back.

 

[AusWolf]

• Buyer only had issues with BO6 and no other games

This says it all to me. There's nothing wrong with your card.

What's wrong is the buyer's game or OS installation. Maybe some dependency files (DirecX or MS Visual C++) are missing or corrupted.

Looking at the videos, I'm pretty sure it is indeed a defective card.

It can't be if the problem only presents itself in one particular game in one particular system.

 

[Todestrieb]

This says it all to me. There's nothing wrong with your card.

What's wrong is the buyer's game or OS installation. Maybe some dependency files (DirecX or MS Visual C++) are missing or corrupted.


It can't be if the problem only presents itself in one particular game in one particular system.

Good point.
The buyer might also have skipped some important stuff like DDU. Or he/she wasn't testing enough things (the one game most played being bad is enough reason to ask for refund imo). More testing is probably needed.

I'm still confident that the card is bad, but let's see what happens next...

 

[arbiter]

If card was bad you wouldn't see elements circled in red bad those would be also corrupted like rest of the screen. There is a chance i am wrong but when i had card failing it wasn't limited like this it corrupted everything showing up. "Buyer only had issues with BO6 and no other games" also leads me to believe its something with that game only.

 

[AusWolf]

Good point.
The buyer might also have skipped some important stuff like DDU. Or he/she wasn't testing enough things (the one game most played being bad is enough reason to ask for refund imo). More testing is probably needed.

I've also had issues with DDU deleting files it shouldn't and corrupting my OS in the process. It's not the be-all-end-all of GPU driver updates.

I actually prefer uninstalling drivers from the settings menu / control panel, and using DDU only if I'm getting errors.

I'm still confident that the card is bad, but let's see what happens next...

And I'm 98% confident that it's not.

If it was, we'd see the same error with the card in OP's system.

 

[bug]

If card was bad you wouldn't see elements circled in red bad those would be also corrupted like rest of the screen. There is a chance i am wrong but when i had card failing it wasn't limited like this it corrupted everything showing up. "Buyer only had issues with BO6 and no other games" also leads me to believe its something with that game only.

View attachment 390832

Agreed. The card is fine. The game install was probably corrupted somehow.

 

[Todestrieb]

If card was bad you wouldn't see elements circled in red bad those would be also corrupted like rest of the screen. There is a chance i am wrong but when i had card failing it wasn't limited like this it corrupted everything showing up. "Buyer only had issues with BO6 and no other games" also leads me to believe its something with that game only.

View attachment 390832

In an attempt to disprove the "UI intact = good card" point... These are very different type of artifacts. I don't know enough about rendering processes, maybe I'm completely missing the point here. I didn't know exactly what in my card was wrong, but these are all gone after I sent the card to repair.
The first one is on Forza Horizon 5, others are on Forza Motorsport 8.

EDIT: In my case, most other games usually crashes without showing artifacts. These two often show artifacts before outright crash.

---

After re-reading the OP thoroughly, I still think that there's a slim chance all games that the buyer tested are safe other than COD:BO6 (what games exactly the buyer tested are still unknown), but I know I'm on shaky grounds now.
When I think about this again, I would expect something to actually crash on OP's card, be it on OP or the unlucky buyer.
EDIT2: And the buyer's friend, which throws some contradicting results.

*this same friend also owns an XTX, so the buyer tried that in his machine and had no issue, while the XTX I had sold was artifacting in this 2nd machine as well. The buyer said the card would start artifacting in around 30 mintutes in his PC. Card initially seemed fine in his friend's PC but apparently it had issues around 2 hours in.

 

[boppingXQWB]

Thank you for taking a close look at my post. Unfortunately I'm not into racing games, so I don't own any to test, but F@H is a good idea.
Actually, I did use to run F@H before selling the card and I don't think I ever had issues with it: it seems like the only way to tell if F@H is having troubles is through driver crashes as there are no logs AFAIK, and I haven't noticed F@H and driver issues, but I did usually run it when I'm AFK.

Now, I actually spun it up and have been running it for the last couple of hours (while I'm actually on the computer) and I have not noticed a driver crash nor something indicative of issues.

What is interesting, however, is that GPU temp & hot spot temp difference has gone through the roof: over 30C delta! I did not observe this huge delta when I had the GPU outside of the case, in fan-up position: as you can see in the OP, the difference there is just 9C. I also don't remember hot spot delta being that large prior to selling. I wonder if I bumped the heatsink when pulling the card out of my case just before selling it.

I haven't tried COD BO6 after putting it in the case, as my main computer does not have enough storage in a single partition for that monstrosity, but maybe* it will now start artifacting.
*Unlikely since I've already been using it in the case for over a week without a variety of loads, buuut the buyer also did only have issue with BO6.

Artifacting or not, this seems to be a good excuse for an RMA repaste while it's still under warranty, and if the card does not artifact with on BO6 in the case, well I'll just attach the buyer's video to the RMA case and see if they do anything about it aside from the repaste.

 

[kilo]

Might be worth your time to try every video port on the back of the card.

 

[maxfly]

Yeaaah, buyers remorse is what I'm seeing.

I'm thinking dude was yankin your chain. A bad card caused by memory(which causes 99% of artifacting issues in my experience. When the core fries, the card simply stops working) or core doesn't come and go like he's trying to say it is. 30 minutes into a game? Nope. A bad card starts artifacting the moment it is stressed, like immediately.

My guess. He looked at his bank account and didn't like what he saw. So, he fabricated a clever way to get out of the purchase.

I don't think there's anything wrong with that card. Particularly being that you've had such a difficult time recreating the issue. Nah, just a piss poor buyer IMO.

 

[OolongGoat]

That’s just Call of Duty being Call of Duty. I had similar issues with a 2080 Ti, 2080 Super, 3080 Ti. Going back to Black Ops 4. The new Call of Duty sometimes does weird stuff on my 4090.

It’s the only game that does it.

 

[Toothless]

Download DDU and amddrivercleanup, newest (or your preference) GPU driver.

Disconnect from Internet, reboot into safe mode.

Run amddrivercleanup, do not reboot.

Run DDU. Reboot.

Offline install driver. Reconnect.

Best way I've seen to flush drivers besides nuking the entire drive. I've had what I thought were bad cards come back with this.