So this is an odd topic, but I've sunk enough time into the last 3 days troubleshooting my poor graphics card hoping it would be a driver issue, and learned enough on the software side to almost certainly deduce that it must be a physical problem.
So when people talk about a graphics card getting old and weathered, what parts of a GPU actually degrade? and how would you detect this kind of damage?
What about how a card sits within the case itself, are there any specific parts that need to be padded so as not to touch other metals? or is that all for the sake of heat mitigation?
Some articles mentioned 'baking' a GPU to re-solder all of the joints.
Memory chips can degrade over time. Long term overheating can cause the transistors in the GPU chip and peripheral IC's to break down I guess?
Biggest and most likely failure points are any electrolytic capacitors. Is the OP picture related or just a random google picture?
Don't touch it with a soldering iron unless you absolutely have to - honestly it's actually better to heat the whole board in an oven than to prod around with a soldering iron unless you really know exactly what you are looking for. Even then most surface mount in very sketchy to hand solder.
Most likely it's dying electrolytic caps.
>>970141
Thats just a google image, this however is the real thing.
I thought at first maybe some thermal paste had gotten elsewhere on the card and could be harmful, but ive given it several pretty thorough cleanings with alcohol. Still cant get rid of what you see here though.
>>970158
>GTX 580M
it could be any number of things, but since this is a rather new graphics card it's highly unlikely to be solder problems on the BGA
those tantalum capacitors though are prone to bust every now and then
it would be way more helpful to post what exactly is the problem
>is the picture getting artifacts?
>>in 3d load, 2d load or even in idle?
>does the computer crash after some time?
>>does it bsod or does it just lose the screen image and the monitor goes to sleep
>can you run say a linux distribution and execute glxgears and check if the problems persist?
>>or load another windows installation with another game?
>>970161
PS: you are using silver paste on the gpu, some are conductive. clean all that gunk with alcohol, especially around those tinny capacitors on the side of the silicon (use an old tooth brush to lightly brush it off) and apply new paste, you dont need much, on the silicon chip only
>>970161
Oh I left a pretty detailed description of the problems over here:
https://forums.geforce.com/default/topic/927218/gtx-580m-is-not-removable-and-and-cannot-be-ejected-/?offset=11#4847461
But to summarize it's bluescreening with nvlddmkm.sys, has made the desktop freeze, has gone to a white screen, a black screen, has black screened at post. And everything I've tried in the last couple of days is all in that thread.
>>970163
I think it is silver paste you're right, I tried my best with qtips and coffee filters and that was the result. Do you think those small bits around the metal prongs which I assume are the capacitors could be responsible?
Also old toothbrushes may have remnants of toothpaste, would it be safer to just buy a new one?
>>970181
you can buy a new toothbrush if you really want too, $2 wont hurt your wallet that much, although you can use your current tooth brush to clean it and then buy a new one for yourself (the residue of paste on the brush will be next to nil if you rinse it after properly)
just make sure that when you scrub it you do it lightly, with almost no force applied, let the alcohol and the bristles dilute all of the paste, suck it all up with a napkin/cloth, repeat until clean.
you mention on that thread that you replaced the thermal pads on the vram modules, did you forgot to put some padding on the VRMs of the card? they are way more important (its crucial actually) to keep cool than the ram itself since they might be overheating almost instantly and tripping the over-temperature protection
>pic related
>>970259
No i didnt actually, the previos pads that were on it got taken out and lost at some point since i thought they might have paste remnants on them. The one i cut up was from the underside which sat between some black plastic and the bottom of the card. It had possible paste remnants too so i was hesitant to use it.
But i cut it up anyway and had success for a short while. I think i covered 4 of them although very lightly because it was a tiny strip. The one on the right is definitely uncovered.
I called around looking for a place to buy them but they only order in. Do you know a good spot online for canada?
>>970273
sorry, am not a canadian anon, but when in need there is always chinkbay
>>http://www.ebay.ca/itm/100X-10x10x1mm-Silicon-Chip-Thermal-Pad-Heatsink-Conductive-Insulation-Paste-WA-/371483930793
>>http://www.ebay.ca/sch/Thermal-Compounds-Supplies-/46322/i.html?_sop=15
>>970178
Sounds like memory issues as in one or more of the vram pins has come loose. I'd try the oven reflow trick before anything else.
>>971532
baking shit should be the very last resort before throwing it away
>https://www.youtube.com/watch?v=E9aZZxNptp0
don't do it unless you tried everything else, OP, the chance of getting stuff to work by cooking it is very very low (you got better changes replacing all those tantalum caps than cooking it)
>>970137
You might be able to get it running again for a couple weeks by reflowing it with a heating gun, but it's probably time for the dumpster.
>>971594
Took a industrial heat gun and liquid flux to a chipset on a motherboard.
Deployed it and hasn't had any issues. Even fixed the intergrated graphics that never worked properly since buying it.
Just cover up or replace electrolytics.
Oven on the other hand will melt plastic and degrade electrolytic caps, without flux it may worsen a bad joint.
>2016
>i hate my HD6970
>no signal
>fan screaming like a bitch
>i realize that it is running at 100%
>something wrong is not right
>the cycle is repeated a few times
>is_it_too_late_now_to_say_sorry.mp3
>brazil's economy faces the worst recession in its modern history
>all imported electronics cost 7x more here
>motherboard has no video output
>buy a new card is not an option
>paraguay also will not be worth
>yolo
>oven at 245 degrees for 13 minutes
>continued the same, maybe a little bit worse
>the problem only happens on windows login
>idea
>remove the video driver in safe mode
>problem solved
>minefield 1024x768 at 3fps
>windows update installs the new drivers
>the card runs for a while and dies again
>mfw
>>971906
you've cooked your graphics card. its completely dead. try replacing the thermal compound and thermal pads for a temporary fix. but the damage has already been done.
>>971909
How are you so sure that this dead? I mean ... apparently it works yet, but just for a short time, sometimes comes close to 2 minutes before stops. As if I needed to replace some components to make it work, i don't think that the oven has destroyed it completely, I read MANY reports of people leaving by 250 degrees for 15 minutes and up to 20 and then the card worked.
ae mano sou do japao e tenho uma BEEM peba aqui (placa de video) nao da pra jogar game mas da pra assistir uns youtube da vida se tu qer me passa teu email pra noes trocar ideia (de graça
>>971906
br electronics seem so stupid, its like 6 grand for a pc right? No wonder your economy cant grow
>>971912
The heatsonk for the gpu is probably not making good contact. As soon as it boots it pegs its temperature and the fan tries to cool it down, but can't so it just runs at full speed and throttled performance, and still proeuces some fatal errors eventually.
