Anyone Revived a dying Nvidia Graphics Card?

[killingtime]

Hi,

I have about 7 NVIDIA GeForce GTS 240 Graphics Cards. All supplied by Dell around 2013, so about 9 years old. Still in the base unit. These cards were working fine in dual screen (HD) mode about 4 years ago. They've not been used for gaming, just office work. Left in dry storage since. Came to boot them up and some had dead graphics cards. OK, it happens.

Ran FurMark GPU stresser on the rest and all failed after a couple of hours. Symptoms are a blank screen and the fan goes to full RPM. Reboot and the video comes back, but only for a while and the screen goes dark again. If I replace the graphics card the computer is fine - so it's the graphics card. Unlikely to be the PSU, as they'd all have to be bad in 7 base units. Possible, but unlikely.

This sounds like a thermal problem because I can get the screens to go dark faster if I run a GPU stresser. I've removed the heatsink and re-pasted but the problem persists. Photo of the cards below. The heatsink is clear of dust and blockages, as I've cleaned them - they were blocked when I got them.

Some questions:

1. Is it normal for graphics cards to just die with age? Remember, these were working a few years ago.
2. Has anyone revived cards with these symptoms by re-pasting? There's paste on the GPU and then 'pads' on the surrounding memory.

I tried re-pasting one card but it didn't work. Same failure.

I would have thought there'd be a thermal sensor on the card to throttle the clock back if the IC gets too hot, but this clearly isn't a feature. These things seem designed to run at blazing hot speeds, as the PCB is too hot to touch after 60 minutes running furmark. There doesn't seem to be any way of 'under clocking' these to reduce heat either. Given the heat, this could be other things as well like blown SMT components.

Thanks.

[50ShadesOfDirt]

Can't give you specific answers to your q's ... not there in my own learning cycle.

I also have a few "suspect" gpu cards, and your post led to some research for me. Youtube vid's from these folks show the toolset I need to assemble, and the process to go through for possibly identifying the failed components:
  - Learn Electronics Repair (aussie tech guy?) ... as he works through some GPU repairs
  - Tech Cemetary (similar to LER)

So, if I decide repair is either worth it, or "fun" to try (once?), I'll have to:

1. rule out external issues for gpu card (bios, power, connectors, drivers/firmware, etc.); probably requires a cpu mobo/psu/etc of same era as the card
2. everything else ruled out, and card is problematic/dead ... disassemble, inspect/smell for obvious smoking gun, then re-inspect with microscopy (these folks are using electronic microscope, and I can see this is a critical tool per the repair vids)
3. diag tools, board-level schematics & such (lots of internet research)
4. if failing components found, parts ordering, various levels of component or SMT r&r skills to work on
5. hopefully, card now works, w/ the old generation of (bench-level) equipment around it

So, much homework for me, with no guarantee of success, but a useful learning curve. Anything I missed?

I think I'll limit it to "once or so for the experience), and/or to that $1000 or more gpu board that just escaped its warranty period (I don't have any of these, as I always buy a few gens back, to be able to afford the gpu).

Not sure I can even recycle these things, as the SMD tech is another level or so beyond my "heat gun" & fill the parts bin approach to electronics recycling these days.

Good luck with your gpu repairs ... I'm off to listen to crappy music on (some of) these vids ... thank goodness LER didn't have that ...

[Bud]

This was extensively discussed in the past. The fault is internal to the GPU chip. Totally dead cards can often can be restored by heating the GPU at 120C for several minutes but they fail again anyway after working for some time after. Usually it is hopeless.