7900xtx with weird artifacts

N_L

n00b
Joined
Mar 6, 2010
Messages
37
Here's a rundown of everything I've tried before and recently, in no particular order. If anyone has any idea what else to try let me know. I plan to keep trying new games until I find enough where it happens consistently and then I can send it back because all the stress and benchmarks are fine. If I send it now as is, they'll just run stress test, benchmark or game and say it works fine and send it back, and I'll be without gpu for weeks or months.



Tried integrated GPU and older GPU card I have (nvidia 1070) no artifacts

Tried HDMI 2.1 from PS5

Tried two different dpi cables and 2 HDMI cables

Tried 3 different monitors(or well 2 plus tv)

Installed fresh win11 without anything and tried there

Freshly installed various drivers

Underclocked

Overclocked

Tried all the stress tests and benchmarks I could find

I even recorded in-game footage and it shows these artifacts in the video

Tried different resolutions in-game and in Windows settings

Tried different refresh rates

Fixed card sagging

Expo (for RAM) is disabled in bios

I don't have another PCIe port to test, or I do but gpu can't fit there

Removed drivers through DDU between testings

No remaining PCIE cables from the PSU but no place to plug it in in PSU even if I had it

I switched cables around a bit inside between things

It might have worked in the end when I tried running it without any drivers but the game was so choppy and lagging I couldn't test more times and CPU was maxing so maybe it was running off it.

Installed the driver only from March 2023, for a moment I thought it worked but after a restart same old, not to say the restart made it worse it's just that it's not consistent.

I reinstalled the newest drivers

Overlock and underclock didn't help from either afterburner or amd software, same with some gaming profile t hat does things.

Can't run memory clock under 2500



As for the games, I haven't really been playing much but I did notice it originally in Honkai Star Rail, which is a mobile game and I thought oh it's flashy and unpolished maybe it's just the game and it wasn't much and at random places.

Later, ESO - older mmo it would show up a lot and in all different colours and places.

Wanted to play two mirror but noticed it right away in the menu, noticed it in the game too but it's a bit more discreet there.

Started playing kingdom come deliverence and it only shows in certain cutscenees and just in a black spot above the cutscene though I think it went down a few times as well.

I noticed it in discord as well but this might just be down to hardware acceleration there.

Warframe I just ran once to test and I could see it popping up here and there when paying attention to it.



Oh and this is how the card is powered, it's not dusty or dirty the camera just went into night mode and messed up the colors.

8wUWCDs.jpg

View: https://i.imgur.com/8wUWCDs.jpg


View: https://i.imgur.com/i5vaq1t.mp4




View: https://streamable.com/jpb5ya - mostly purple here


View: https://streamable.com/aqmsdd - here it's running on old GPU (nvidia 1070) and it's fine, also the pet went through pole so it's clipping not artifacting haha


View: https://streamable.com/wkqbar - here I was recording it around discord but saw it behind in game too (honkai)


View: https://streamable.com/ct1qmz - cutscene from kingdom come, mostly on top and top left but it goes down too


View: https://streamable.com/93ox03 - ^


View: https://streamable.com/yw2bfn - it showing in eso


View: https://streamable.com/s2lsi8 - ^ same

config CPU:AMD Ryzen 7 7800X3D MOBO:ASUS ROG STRIX X670E-F Memory: Kingston DIMM DDR5 32GB 6000MT/s Fury beast GPU: XFX AMD Video Card RX-7900XTX Speedster MERC310 BLACK 24GB GD PSU NZXT C1200 Gold 1200W SSD Kingston 2TB M.2 NVMe SFYRD/2000G SSD FURY Renegade as main SSD Samsung 980 PRO 2TB as seconda WD 10TB purple for media stora Monitor: Samsung Odyssey G7 27" K+M: Logitech G502X and SteelSeries Apex 7
 
Last edited:
looks like memory errors. is it supposed to be 2500? i thought it was 2000.... nope, 2500 is normal. even raising the power limit does nothing?
 
Do you always have Discord open and running, while playing a game?

Any other apps open?
 
looks like memory errors. is it supposed to be 2500? i thought it was 2000.... nope, 2500 is normal. even raising the power limit does nothing?
Nothing :|
Do you always have Discord open and running, while playing a game?

Any other apps open?
Not really, just had it in that situation when I noticed it.as for other apps just software for keyboard, mouse, gpu and think thats about it
 
Nothing :|

Not really, just had it in that situation when I noticed it.as for other apps just software for keyboard, mouse, gpu and think thats about it
tried an older driver?
 
Yes, I tried one from march and this has been happening on various versions ever since june of last year. Now im trying to get new cable (16 pin to 2x8 pin) and see if that does something
 
My 2 cents, don't use a power supply that daisy chains the connectors. Get one that has a single connector at each end of the cable.

What is the make, model and rating of your power supply?

Also... I don't see the xfx anti-sag bracket in the picture. Is it installed?
 
The support bracket is easy to overlook in the packaging of the gpu. I had to reexamine the foam packing to locate it when unboxing my xfx card.
 
My 2 cents, don't use a power supply that daisy chains the connectors. Get one that has a single connector at each end of the cable.

What is the make, model and rating of your power supply?

Also... I don't see the xfx anti-sag bracket in the picture. Is it installed?
Daisy chaining the 3rd 8 pin is totally safe as it will draw nowhere near what the first two are drawing.
 
I had something like that on one of the 7900XTX I had, I ended up having to repaste it because the factory paste only covered like 80% of the die.
 
My 2 cents, don't use a power supply that daisy chains the connectors. Get one that has a single connector at each end of the cable.

What is the make, model and rating of your power supply?

Also... I don't see the xfx anti-sag bracket in the picture. Is it installed?
Agreed. I have seen enough posts around the internet about RDNA 3 cards improving when using all 3 cables separate.

N_L To me, the artifacts look like 1 of 2 things:

1. A VRAM issue. It could be the VRAM is bad. Or, it could be its not getting enough power and using 3 separate cables may fix it.
LigTasm mentioned similar issues being fixed by a repaste. XFX does allow you to remove the heatsink, without voiding the warranty (the only brand which allows that). But, I would probably just RMA it if 3 power cables doesn't fix it.

2. If you have any app windows open behind your game, it could be a glitch of those apps. You can try running your games with zero extra apps running, turn off overlays for Steam or anything else.
 
Either VRAM or PSU, possibly but unlikely both.

I had an 850w that couldn't handle the XTX. No (noticable) artifacts, just crashes to desktop, usually with a driver time out. A new 1kw Silverstone fixed that, but I think it's insane that was necessary.
 
My 2 cents, don't use a power supply that daisy chains the connectors. Get one that has a single connector at each end of the cable.

What is the make, model and rating of your power supply?

Also... I don't see the xfx anti-sag bracket in the picture. Is it installed?
I did not consider that when picking the PSU, this one was the best choice out of the bunch back then that was available sigh it only has 4 PCIe/cpu slots and one 12-pin. Even if I get 3rd cable all 4 are already used, I can disconnect 1 CPU to test but still not ideal to run like that, and yeah I'd switch psu if I found this was an issue but unless I confirm it it's a pricy guess. Plus I feel like when I underclocked it and ran games that aren't performance hogs it wouldn't matter right? It's not like the 3rd slot is drawing full power and a lot of these cards come with just the two 8 pins from what I saw.
The support bracket is easy to overlook in the packaging of the gpu. I had to reexamine the foam packing to locate it when unboxing my xfx card.
Well, I'm embarrassed to say I just found out that the big thing from the box is the support bracket... I am using a generic support tool that's placed under it on the far right end, works just as well I guess. Actually you can just see it to the right of the 3rd cable.
Daisy chaining the 3rd 8 pin is totally safe as it will draw nowhere near what the first two are drawing.
This was my thinking as well/
I had something like that on one of the 7900XTX I had, I ended up having to repaste it because the factory paste only covered like 80% of the die.
Can't open it while under warranty plus the temps are fine, great even.
Agreed. I have seen enough posts around the internet about RDNA 3 cards improving when using all 3 cables separate.

N_L To me, the artifacts look like 1 of 2 things:

1. A VRAM issue. It could be the VRAM is bad. Or, it could be its not getting enough power and using 3 separate cables may fix it.
LigTasm mentioned similar issues being fixed by a repaste. XFX does allow you to remove the heatsink, without voiding the warranty (the only brand which allows that). But, I would probably just RMA it if 3 power cables doesn't fix it.

2. If you have any app windows open behind your game, it could be a glitch of those apps. You can try running your games with zero extra apps running, turn off overlays for Steam or anything else.
No apps, and while xfx might allow that in the states here the shop that sold it would void my warranty, then the distributor of the card in the country as well wouldn't accept it and who knows where xfx is in this chain, nowhere I can reach them anyway and I still have a year and a half of warranty.
Tried a undervolt?
yes, same
Either VRAM or PSU, possibly but unlikely both.

I had an 850w that couldn't handle the XTX. No (noticable) artifacts, just crashes to desktop, usually with a driver time out. A new 1kw Silverstone fixed that, but I think it's insane that was necessary.
I've had drivers time out recently in helldivers 2 game but that's the game's known issue, I also noticed Discord would crash too before I disabled hardware acceleration, also haven't seen these artifacts in there since then.. My psu is 1200w and it should be handling this gpu just fine even with daisy chain I think but can't verify unless I get 3rd cable somehow. cablemods don't make them for this psu, I asked if the one from 1000 works and also emailed nzxt support but all of these will take a while and in the meantime I'll continue testing what I can.
 
Last edited:
Fingers crossed, running the game kinda gimped atm with all sort of workarounds but yeah that's unrelated.
Did a few tests, overclocked, underclocked, voltage up, voltage down in CP2077, and monitored my temp, here are the pics, the game ran fine, fully maxed settings and the pic was while in "rage" mode which is and adrenaline oc, ~45-50 fps. didn't notice anything out of the ordinary though the game is flashy and maybe if I played it for few hours something might have popped up.
First pic is while running haven for a while and 2nd cp2077.
nJhuFn7.png

XwwAjit.png

View: https://i.imgur.com/nJhuFn7.png

View: https://i.imgur.com/XwwAjit.png
 
Back
Top