Extremely weird graphical bug...

[Philippev]

LONG POST AHEAD sorry :/

Hello everyone,

Since a few weeks, my gaming computer is bugging like never before. And after days and days of testing and fixing stuff I can't find the source of the problem.

It all started with my games freezing on a glitched frame, after a few minutes of playing them, like so :

https://imgur.com/3McAtFf

(at that time I was playing "Control" "Outer World" "We Happy Few" "Dishonored 2" "Tacoma" "Mutant : Year Zero")
Sometimes it was followed by the computer restarting, sometimes it stayed that way till I pressed the restart button myself.

I immediatly thought of a GPU issue, and checked if it wasn't overheating. But everything was fine.
(Important info: I've replaced the original fans from the GPU with a watercooling system a year ago).

My Rig :
Motherboard: ASUS Z97M PLUS
CPU : Intel core i7 - 4790K LGA 1150, 4GHz 8MB
COOLING : Noctua NH-U 125. 120mm U-Type Tower cooler
GPU : GTX 1070 Ti MSI Aero - watercooled with NZXT Kraken G12 + CorsairHydro Series H55
RAM : Ballistix Crucial DDR3 1600, 2x8GB
PSU : COOLER MASTER G550M
SSD : Samsung 850 Pro 256GB (system)
Samsung 840 PRO 512GB (game folders)
HDD : WD Black 3.5" 1To 7200RPM
SCREEN : DELL 21:9 3440x1440
OS : Windows 10 - All drivers up to date.

I tried playing with less demanding games, like Minecraft (bedrock edition) and Gary's Mod, and everything was fine.
Sudently, during a few days, every games started working fine again, for no reason. But it didn't last, the glitch came back as soon as the first frame was displayed this time.
I was still able to play Minecraft, and use Blender perfectly, even the first 30 minutes of "TitanFall 2" ran smoothly, and then glitched out.

The weirdest thing is, that yesterday I took my computer to a friend's place who has a really fast bandwith, to reset windows and reinstall everything. At the begining, and after a clean install, it was still glitching, and at some point, every games started to work well again.
"Witcher 3" "Dishonored 2" "Mutant : Year Zero"... perfect
Then I came back home, and those same games started glitching after less than a minute playing them.
With sometimes only a return to desktop, instead of a full restart from the computer.
And even weirder : Games like "GTA V" "CRYSIS 3" or "BIOSHOCK INFINITE", are still running perfectly fine for hours, on high settings.
(other games that are glitching : "Batman Arkham Knight" "Metro Exodus")

I don't understand what is happening.
I thought about a motherboard issue, a PSU issue, a GPU issue, a Directx issue... it's making me crazy.

So far I've tried :
-Clean install of windows
-Reinstalling everything and every games
-Moving the GPU in another PCI slot
-Using only one RAM, then the other
-Bios update
-Tested the GPU with Furmark and 3Dmark : No glitch !
-Tested the SSDs with Samsung Magician
-Installed games on same SSD as Windows
-Uninstalled and reinstalled Nvidia's drivers and Geforce Experience
-Disabled Steam overlay

Thank's for reading!

 

[newtekie1]

GPU is dying. With that type of artifacts, I'd guess it's the VRAM that's dying.

 

[Athlonite]

Yup certainly looks to be dying VRAM from that screen shot although why some games still play fine and others don't could be down to the way the engine is formatting and compressing the textures say DX9 vs DX11 or 12 which is causing the corruption to be seen on screen... anybody seen a memtest type program for vRam

 

[eidairaman1]

Time to upgrade, get away from msi too.

 

[newtekie1]

Yup certainly looks to be dying VRAM from that screen shot although why some games still play fine and others don't could be down to the way the engine is formatting and compressing the textures say DX9 vs DX11 or 12 which is causing the corruption to be seen on screen... anybody seen a memtest type program for vRam

It could also be that some games are using more VRAM than others. So the corruption is more or less likely depending on how much VRAM is in use.

 

[lsevald]

Looks like the vram is struggling, have you tried to under clock your vram? I see you got custom cooling, are all the ram chips cooled properly?

 

[hat]

Test with a spare card, if you have one (or can get one) just to be sure. If you can't get a spare card, or you get a spare card and there are no problems, RMA the card if possible. If not, it may be possible to get some life back into the card by baking it (yes, in an oven). This is a last ditch effort. If this fails, you're screwed, and you need to buy another card.

 

[Bones]

Agreed - It does appear the card is dying with all the artifacting seen in that screenie.

 

[Waldorf]

@Athlonite Philippev

occt can scan vram.
OCC tool

@eidairaman1
not one msi i had in the past 5y underperformed as much as the gigabytes did.
all gaming x i used for my/other builds most the time performed even better than others with "more" power/phases etc (asus/evga),
and fans are quieter as they tend to be the largest on the market.
not one issue with any, even the used ones. cant say that for the GBs i had.

 

[DeathtoGnomes]

Then I came back home, and those same games started glitching after less than a minute playing them.

2 things here, 1. the PSU might be having issue, remove the gpu and clean the socket, check the power to the plugs from the PSU. 2. check your wall outlet power with reliable good tester.

If thats all good, look into a new card.

 

[Waldorf]

@DeathtoGnomes
How could it be the slot or plugs, as nothing changed and it was working before?
The problems on an outlet that affect a pc (besides no power), cant be found with a simple tester.
The chance that it is the psu and not the vram/gpu is small.

 

[Voluman]

Is that Kraken bracket cools vrm, memories of the card too?
Try out different settings, with various directx, different settings with higher/smaller vram usage. Any overclocking is present? Reset to default.

What kind of display connection and which port on your card? I may check display cable too, but it seems rather hardware based error.
Is there a specific time/date of this error? You can check it with an earlier driver version too.

 

[DeathtoGnomes]

@DeathtoGnomes
How could it be the slot or plugs, as nothing changed and it was working before?
The problems on an outlet that affect a pc (besides no power), cant be found with a simple tester.
The chance that it is the psu and not the vram/gpu is small.

its part of the elimination process. just because someone says its this or this or that, doesnt mean you should just throw money at a problem until its fixed. Due diligence is needed to determine what actually needs to be fixed.

 

[P4-630]

get away from msi too.

Why? They make good products imo.

 

[INSTG8R]

VRAM Artifacts not much can be done to fix it.

 

[cucker tarlson]

get away from msi too.

what a ludicrous statement,best cards around atm are gaming and trio

are you still mad about this

MSI Says not all Radeon RX 5600 XT GPUs Can Hit 14 Gbps Memory Speed

Just before the launch of AMD's latest Radeon RX 5600 XT graphics cards, AMD released a vBIOS update that boosts the GPU clock speed by a few percent margins and increases the GDDR6 memory speed from 12 Gbps to 14 Gbps, which is almost a 17% performance increase in memory bandwidth. However...

www.techpowerup.com

oh well.....

Why? They make good products imo.

they're fine,spiteful user uses tpu to push his agenda again.

@eidairaman1
not one msi i had in the past 5y underperformed as much as the gigabytes did.
all gaming x i used for my/other builds most the time performed even better than others with "more" power/phases etc (asus/evga),
and fans are quieter as they tend to be the largest on the market.
not one issue with any, even the used ones. cant say that for the GBs i had.

I agree,one of the best like I said,and no major screw-ups in the recent history too.
msi is the one to go with in my books.along with palit jetstream/gaming rock cards which are majorly overlooked in favor of brand names like evga and asus

evga,asus,gigabyte - no thanks.

@Philippev if you know someone who does vram replacement then try.

 

[Waldorf]

@DeathtoGnomes
Sure, and i like to do it Spocks way too, except it worked fine and nothing was changed, so the outlet obviously still provides power and even if something is wrong, its not like the wiring changes itself over night, so no need for a (consumer grade) tester.

Same for psu. If its broken, it will cut off pc, freeze it, crash or similar, but not produce artifacts, especially that the time the gpu can be used until it shows problems is now shorter, so its gpu related.

 

[Philippev]

Hello everybody !

Many of you where right, it was a VRAM issue !

I disassembled my GPU, and realised/remebered that 5 VRAM chips weren't cooled properly, because they were located under the watercolling block, and I didn't had enough room to fit the heat sink on them, like I did on the others.
So I ended up using smaller heat sinks, stacked together to cover most of the surface on each VRAM chips left uncooled.

And now, EVERYTHING IS WORKING PERFECTLY !!

So thanks a lot for all of your messages !

 

[lsevald]

Nice If it were me, I would also make sure there's some airflow over those heat sinks. Easy to forget when water cooling, the CPU/GPU will get great cooling, while the rest suffers a bit.