What usually causes artifacting?

[bufulo]

That looks like this? The dots are always red over dark areas, blue over light areas, seem to be static unless the image changes. For example, if I grab the window shown below, the speckles will move around the outside of the frame of it. My guess is that it's a memory issue but I have no experience with the details. It sometimes happens immediately after cold boot, but most of the time not until Windows goes into Standby or a power (and maybe clock) state is changed (when the GPU might be asked to throttle back due to heat, for example). Definitely happens every time the GPU gets too warm (maybe above 85 C). It doesn't seem to be related to driver version at all (I've successfully tried DELL-specific drivers, Microsoft Default NVidia, and NVidia most recent from their website, stable version). It's not gotten any worse for over a year of use. I have lots of ideas but I'm curious what those of you with more experience have seen.

 

[droid-I]

Had one nvidia card showing like that, I got it 2nd hand.
I summed it as the card been overheated for long periods. For spares it went.

 

[crlogic]

That looks like failing VRAM to me

 

[Shrek]

For me it was fatigued solder joints on the GPU

Video Card Reflow | Electronics Repair And Technology News Video Card Reflow (jestineyong.com)

 

[droopyRO]

Did you try it on another PC and/or another monitor or TV to rule out a bad display cable ?

 

[droid-I]

^ good point. I might think if it is a HDMI connection, would be just a black screen, but I am just learning.

 

[Mr.Scott]

Bad Vram.

 

[mouacyk]

Had green/red dots on a bad HDMI cable.

 

[bufulo]

Did you try it on another PC and/or another monitor or TV to rule out a bad display cable ?

It's definitely not cabling. I have tried 3 over the past year and the issue always reappears.

 

[delshay]

Could be a number of things. Possible VRAM cooling, but to me it looks like somekind of leakage from filtering capacitors, but I could be wrong..

 

[Calenhad]

First thing that comes to mind is failing VRAM. You could try underclocking it if you are comfortable with doing that. See if that helps.

But it is possible that other components on the GPU is failing.

Have the display been tested with other equipment? This is also one way displays slowly dies.

 

[Deleted member 185158]

If it's not the cable, it's the v-ram.

 

[maxfly]

That looks like failing VRAM to me

Bad Vram.

First thing that comes to mind is failing VRAM.

If it's not the cable, it's the v-ram.

Yeah. Bad GPU memory.

 

[bufulo]

Definitely not the cable or display. I forgot to mention that I also have integrated video (Intel i7-4790S) and an old Radeon video card that both work perfectly.

I suspect memory as well but was hoping for better news. I've tried reducing the memory clock thru vBIOS changes (haven't found any other way to do this) but I'm not sure the change worked - it behaved exactly the same. How low can I go? I believe the memory is Samsung K4G80325FB-HC03. Good thing I took lots of photos the last time I had it apart!

Could increasing the power help at all with diagnosis? I know Quadro cards are undervolted and underclocked to begin with.

I'm trying to get MODS to run to get more information but having lots of trouble due to my inexperience with Linux plus this unique hardware (essentially a notebook motherboard in a desktop - HP EliteDesk 800 G1 USDT). Technically I got MODS to run and generate a log file (not report.txt) and I'm currently trying to debug the test environment.

 

[droid-I]

IMO, if the vRAM is fried, reducing clocks will not help(by luck it might work, but would not count on it).

 

[bufulo]

I don't think anything would be "fried" in the usual sense. This is a Quadro card that is undervolted and underclocked to begin with. It did get a little warm a few times as I was conjuring up adequate cooling last year though. I guess that could mean "fried in the usual sense" but it's not like the card has been abused in any other way. The times it was run hot were very brief and probably didn't even hit 100 C, or if it did, maybe for a matter of seconds before I shut it down. I forget the details from back then.

In any event, hopefully I can get MODS going to provide a little more insight before I try some last resort crap. I've already tried minor heating (below solder melting temps) and the card comes back every single time. That could mean GPU internal connections, but that's just wild guessing. I was really hopeful someone had seen very similar specks and fixed it so I'd know what to consider trying next.

The artifacts seem to return if the card is asked by Windows and/or motherboard BIOS to enter a low power state (Sleep, Hibernate, etc.) every time and occasionally at higher temps (above 80 C maybe). The former had me originally thinking motherboard BIOS and vBIOS miscommunication (DELL video card in HP motherboard) at first but the heat and artifact pattern seems more likely to be memory or possible memory controller (in the GPU I believe?). Anyway, It takes a lot of time to figure out how to try to test these scenarios and get any useful guidance as a result. All the replies here do help. Thanks!

 

[hat]

I've already tried minor heating (below solder melting temps) and the card comes back every single time. That could mean GPU internal connections, but that's just wild guessing. I was really hopeful someone had seen very similar specks and fixed it so I'd know what to consider trying next.

That does suggest solder joints have gone bad somewhere... but sadly the best you can do yourself is bake the card and hope it comes back. There are no guarantees with this method... it might work, it might not work at all, or it might make matters worse, but if you've got nothing left to lose you can try it. It's your best option short of handing it over to someone who has the skills and tools to properly reflow the card, and that ain't cheap or common.

 

[Shrek]

Be sure to use flux and don't bake the whole card but use a heat gun.

 

[maxfly]

When memory goes bad it just goes bad. There's no rhyme nor reason. Temps can be amazing for years and poof one day, artifacts.
I wouldn't bother baking it. You'll only prolong the inevitable. At best, you might get a week out of it before it black screens and dies completely.

 

[Chingy]

Looks like Vram is failing, that’s usually what the dots like that are. If it was loose solder joints like a gpu I used to have then it would be lines.

 

[Logan7]

Have never used MODS but there are a number of Windows-based tests for testing VRAM, like OCCT, memtest_vulkan, or memtest_CL. I would imagine the dmgg.py script in Arch Linux would work with an NVIDIA card but I have never done it (usually MODS is used obviously).

Also, GPU-Z is showing 0 MB of VRAM, is that always the case for that card or does it show sometimes?