Weird graphics card glitches

[Pho]

I'm having some issues with my GPU. It's a 780 GTX so getting a bit long in the tooth but I don't want to upgrade until the nVidia 3XXX range comes out.



Notice all the green/blue/teal/aqua or whatever they are pixels and how in the colour pallet at the top right there's a green curve that shouldn't be there.

If I adjust the current hue the green line moves smoothly - https://photos.app.goo.gl/ejvY541Ngy2estiE8

If I take a screenshot and send it to someone it looks perfectly fine to them. If I take a screenshot and open the image up in Photoshop I see the odd pixels still. But what's really weird is that if I move the image around or resize the image, the weird pixels move with the image, and if I use a colour picker I can see from the RGB colour code that that pixel really is the colour that I'm seeing on my monitor. It's like there's some filter being applied to my graphics that's physically embedding this in the image.

Here's what it looks like in a game:
https://photos.app.goo.gl/regZLaQqiCFKG5zF8

(oh and if you play COD, yes that is the secret bunker Mud Drauber unlock )

If I reinstall the nVidia drivers it seems to go away but may randomly come back again a day or two later.

Motherboard: MSI B450 Gaming Carbon AC
CPU: Ryzen 3700X
OS: Windows 10 2004 update

I do randomly get hard resets every so often, this has only happened since upgrading to the above bits and pretty much always when gaming or playing video. No crash dump, just a reboot.

I'm really confused, especially with the fact it's not just corrupting static parts of the screen.

Cheers

 

[Pho]

what psu do you have, and your ram too?, just noticed the gpu

you mention windows 10 update 2004, is that an insider build as i don't have that only version 1909 on my end (build 18363.836), regarding the gpu it could be on its last legs, try disabling AA in game's and run at medium settings to see if the artifacting goes away.

I did play with overclocking the GPU a couple of months back, both its core and memory and voltage. Wonder if that didn't help? I did end up undoing it though and the temperatures were never over 70c or something so shouldn't have been stressing it too much.

I'm not 100% sure on the wattage but I think it's an OCZ gamexstream 700w. I was using this on my old machine, which was a overclocked 2500K and would have been far more power hungry - I only swapped the motherboard/CPU/RAM and added an NVME. The RAM is Thermaltake Toughram RGB running at 1.35v/3600Mhz as specified.

2004 is the April 2020 update that's being rolled out now. You can install it before it hits your Windows update by clicking update on here.

Thinking about it, I think the greenness problem did start after installing this version. However the crashes have been happening for a while, but I think this version has made it more stable from that point of view.

My motherboard has an 8+4 pin CPU connector, I do only have the 8 pin connected as I didn't have a connector for the 4 in but from what I read the 4 pin is only needed if you're doing extreme overclocking which I'm not and lots are running it without.

 

[Pho]

thats a very old psu, found a review from 2006 online, have you got any monitoring programs installed like hwinfo64 or hwmoniter?, or when you restart head into the bios and check the voltages the psu is supplying, the programs i mentioned just will do the same thing. you need to put the pc under load and have one of the programs above running to show what the psu is outputting, if you get stuck post a screen shot here and we'll take a look and make sure everything's good

the extra 4pin eps next to the 8pin on the board is only needed if your overclocking the cpu hard, tbh if you were the 8 pin cable is more than enough anyway, so no issues there.

i suspect your psu maybe the problem, if your old system was playing up before your upgrade could be further evidence that the psu is on the way out, may have already damaged the gpu too

Oh I just checked, I think that was my previous one . The one I have now is an OCZ ZS 650w, but even that looks like it was released in 2011. Didn't realise it was that old! My old system was rock solid.

I have a couple of other PSUs hanging around so I might give one of those a try too.

I'll give HWInfo a go and see what it reports thanks for the suggestion. Hopefully(!) it'll crash again and we can figure it out.

 

[Pho]

have you got any monitoring programs installed like hwinfo64 or hwmoniter?, or when you restart head into the bios and check the voltages the psu is supplying, the programs i mentioned just will do the same thing. you need to put the pc under load and have one of the programs above running to show what the psu is outputting, if you get stuck post a screen shot here and we'll take a look and make sure everything's good

Managed to capture a log. I I guess it lost a few seconds before the crash due to it not being written to disk.. if you have any ideas that would be great. I'm a bit out of touch as to what voltages are good/bad in PCs now

https://www.dropbox.com/s/aa95k25630zb7t1/log.zip?dl=1

I thought the olde rule was that if you took a screenshot, 'mailed it and everything was fine, but it was broken on your own screen, then it meant it was the monitor, or the cable. Though, that wouldn't explain why it moves unless it's only the rendering part of the card that's broken, not the generation. Though, it fixing if you reinstall drivers suggests that it is either a software issue, or a fault that the drivers re-calibrating corrects, which could also be the monitor, or the cable.

The hard resets may or may not be related.

I guess a compatibility issue between the motherboard and the GPU is possible, since one is much older than the other. Have you tried different drivers for the graphics card and the chipset? Though that still wouldn't explain the resets.

Yep exactly. It's almost as if it's in a green screen mode or something. Utterly confused .

I'm running the latest nVidia drivers but it might be a good idea to roll it back a couple actually and see, good suggestion.

I'd be tempted to try using a different monitor cable, or using a different type of port hdmi/display port/VGA etc. Then try connecting to a TV or other monitor all see if it goes away. But looks like it could be the card.

Ah good suggestion, I have a second monitor so I'll try moving it onto that when it happens next.

 

[Pho]

.

That's awesome thanks for looking into it for me I forgot to say that was whilst playing a game so it was under heavy load.

It sounds like it's a bit of a mystery then.

I'm thinking I might try sticking some more volts through the card and see if that makes it more stable?

 

[Pho]

i doubt more volts will help, most likely it may damage the card, have you tried a different screen or cable as other have suggested above first?

you could try a nvidia driver below v400.00, could be the new rtx features are mucking around with the card, head to this page

https://www.nvidia.com/Download/Find.aspx

and pick a couple of gtx 780 drivers from the list and have a play, there should be version's available from 382.33 upto 391.35 for your card


EDIT not sure if the link i provided above works properly anymore so below is another link to most of the driver releases for NVIDIA stuff

https://www.guru3d.com/files-categories/videocards-nvidia-geforce-vista-|-7.html

pick drives below v400.00 and see if the glitches go away

Slow reply as I've been doing more testing to try and get to the bottom of it.

I ran the built-in memory test and that seemed to report nothing was wrong - I was hoping it was something simple like this! I updated my chipset drivers yesterday and haven't tried a game yet to see it it crashes so hopefully this might help. If it doesn't ill try and see if I can roll the GPU drivers back even further like you suggest.

I thought I had cracked it though: on the occasions where it doesn't fully crash the machine (just the driver) or when I reboot and have the green pixels there still I decided to try the cable like @Tetras suggested. Pulling the DVI cable out the monitor end and putting it back in had no effect. Pulling the cable out the GPU and putting it back in seemed to fix it, strange. However it did start playing up again a day or two later - I'd played hours of games in the meantime, then as soon as it started playing up it crashed several times in a matter of minutes. I'd only been playing for a few minutes too so it wasn't temperature related.

Also another odd thing I've found - I've a second monitor connected so I tried moving a window over to that and the pixels problem didn't seem to exist there. I'm pretty sure I did but I can't remember if I tried switching the cables and ports around between the two monitors to see if the problem magically shifts onto the other one from now on - next time it does it I'll definitely do that.

Maybe this thing is just possessed?

 

[Pho]

They're on my list of things to replace, two Dell Ultrasharps from ~2007. Didn't realise they were that old! No idea how old the cables are, I probably have some spare ones lying around somewhere though.

I'll wait and see if it crashes again and if it does I'll swap out to another monitor with HDMI on instead. Then if it crashes again I'll try a different PSU for the sake of it.

The annoying part is I can't easily trigger it so testing is a pain.

 

[Pho]

Windows 2004 has known issues. I installed it the other day and quickly reverted back to 1909 after it caused multiple BSOD's.

If you haven't done so already, I would recommend going back to that.

I generally found it to be much snappier on my PC but had some odd slowdowns on my laptop, I think it must be driver related as the task manager wasn't showing much going off.

I downgraded back to 1909 and played a day or two of games and everything seemed fine then last night it crashed again . It does seem to be fine for a day or two then go through a spait of crashing.


One thing I have noticed is that if I have multiple video sources playing that seems to aggravate it a lot more. E.g., playing a game fullscreen whilst leaving YouTube or something playing hidden away in the background. Maybe it's causing too much strain on the GPU and it's giving up, but it definitely seems to be a trigger.

 

[Pho]

that could be down to either the gpu or the psu, given that both are quite old now its tricky to know which one is at fault, but if you have a friend who could test your gpu to make sure its fine then i'd say the psu if bad, if the gpu still gives out weird glitches once tested in another pc then the gpu is bad.

either under high load the psu cant keep up or the gpu has a memory or core problem, if the crashing keeps happening i'd suspect the psu, because if a gpu fails thats it, you would get a black screen all the time or horrible artifacting sat in windows at idle

So a bit of an update, I swapped out to a spare PSU I had (750W OCZ something or other) and so far, touch wood, it hasn't crashed once.

I wonder if it was overheating or something? It didn't feel hot out the fan exhaust or anything, but it may explain why it could run fine for ages then suddenly start going on a crashing spree.

I'm not entirely sure it's fixed but let's hope so for now

 

[Pho]

Hopefully!

I've put that PSU back into my old 2500K build which is now acting as my media PC. It's just using the onboard graphics from the motherboard now so should be much less strenuous on the PSU. Perhaps it was just feeling homesick.

Now I'm wondering how likely it is it could go pop though :/

 

[Pho]

I do actually have another spare Corsair 450W PSU as it happens so I might end up doing that then. Though I've not had any issues with the old PSU in the media PC so far.

Another thought - possibly for anyone who comes across this in the future - I did have the extra 4 pin CPU connector available on the 750W PSU that I plugged into the new PC, I know it shouldn't have been required but maybe the board is just sensitive and needed it.

Still no crashes.. cheers for your help again!

 

[Pho]

Hmm so as a bit of an update for anyone who comes across this.. the crashing came back but I think I might have a solution .

I tried flashing to the latest beta MSI bios. I had been getting errors in the event log about "Display driver nvlddmkm stopped responding and has successfully recovered.". When I Googled for a changelog on the BIOS the first thing that came up was:

I have on an MSI B450-A Pro Max with an R5 3600 and 2070 Super. It completely solves the nvlddmkm driver crashes when cards switch between idle/active state (PCI-E lane speed changes).

It also completely solves the blackscreen crashes on a 5700XT in the very same situation.
https://www.reddit.com/r/Amd/comments/g750y5/updated_amd_agesa_comboam4pi_1005/

I do think with my old PSU it was crashing the entire system when this was happening. The new PSU seemed to be able to cope with this crash better, the screen would flicker, but I could carry on.

However, this still didn't solve my crashing whilst gaming issue. So I thought I'd try reinstalling Windows. That didn't fix it. I tried running FurMark for 20min, that didn't crash or have any artefacts that I could see, which I found odd given I thought the GPU was bad. I tried running both the Windows memory test and memtestx86+, neither reported any errors. I tried running OCCT's stress test - that triggered a crash after a few minutes. I tried running the test again, it triggered another crash. Hmm, interesting!

When OCCT was causing crashes I was noticing in my eventlog the following. I'd never noticed this before when crashing during gaming etc.

A fatal hardware error has occurred.

Reported by component: Processor Core
Error Source: Machine Check Exception
Error Type: Bus/Interconnect Error
Processor APIC ID: 0

The details view of this entry contains further information.

After a bit of Googling some people suggested it may be memory that causes these issues. I remembered a friend was having issues with his motherboard crashing when using the XMP memory profile, but he has an old gen Ryzen, and slower memory, and I'd read before I even bought this set-up these had been fixed in Ryzen 5 etc.

But I thought I'd give it a go anyway so I disabled XMP falling back to the default 2666MHz at 1.2v rather than 3600MHz at 1.35v. I ran OCCT for 25min or so and it seemed seemed to remain stable. I played some games for a few hours with friends last night, and it didn't crash. I have read some people were able to get back to near XMP speeds by manually setting the frequencies/timings etc but I haven't tried that yet.

So maybe it was memory all along?