[SOLVED] What are the typical symptoms of old / failing graphics cards ?

[killingtime]

Hi,

I have a number of Dell first gen i5 base units with NVIDIA GeForce GTS 240 Graphics Cards. Running on Win7x64. Original Dell Mfr spec.

These computers and NVIDIA cards are about 12 years old. They work but ... Some of the graphics cards are 'misbehaving'.

1. One day at power up I got 6 bleeps from the MB (No video BIOS) and a blank screen. Changed the NVIDIA card out and the system was fine. Dead video card.
2. At random intervals, the fan on the graphics card spins at full power, then the screen goes blank. Reboot, it's OK. Computer sat idle. Thing is, the card is cold to touch
3. Sometimes 2) happens with decreasing time intervals and then the card dies completely.
4. If I run a GPU stresser on the card, I can get some cards to blank out with a fan spinning at full power. Computer unresponsive.

All of the heat sinks on these graphics cards have been cleaned - original thermal paste. GPU temp during stress test is about 66 degC - so the paste is fine. Outside of stress testing, there's nothing taxing on the card, just a web browser, youtube and Excel, but some still go dark and noisy.

Q) Do graphics cards die with age these days? Is this behavior normal for an old card?
I always though PC boards just 'died' completely when they went.

I now have a pile of dead GeForce GTS 240, but still plenty left.

Thanks.

 

[Lutfij]

Do graphics cards die with age these days?
Yes, if they're built with subpar componentry or abused in any way, they die.

Is this behavior normal for an old card?
Yes, they could also be caused by a faulty or failing PSU.

 

[killingtime]

Thanks Lutfij.

When I first got these computers they were filthy. CPU and GPU heat sinks 90% blocked with fluff, so I'm guessing they'd cooked at bit. Assumed if I cleaned them and they ran for 30 minutes under stress they'd be fine, but that's not looking correct. Still a gamble.

Would blocking a GPU cooler count as 'abuse'?

 

[Lutfij]

If I artificially prevent someone from breathing, does that count as abuse? Certainly. If you prevent a GPU from exhausting the heat it's meant to be dumping(due to a clogged heatsink) that is the same scenario. Make and model of the PSU used to power the GPU?

 

[KyaraM]

Those GPUs aren't supported by Nvidia anymore and there are no recent drivers available. It's possible there is an issue there, or a driver corruption. It is also possible that over time sensors started malfunctioning, which might cause especially question 2 if the card and OS both believe the false reading. Or there might be overheating in a part of the GPU not monitored by whatever software you use, but the GPU itself, eg the VRAM. You would need an infrared thermometer to find that one out. Or it could be one of a million other things going wrong there... who knows with cards as old as that.

 

[killingtime]

If I artificially prevent someone from breathing, does that count as abuse? Certainly. If you prevent a GPU from exhausting the heat it's meant to be dumping(due to a clogged heatsink) that is the same scenario. Make and model of the PSU used to power the GPU?

Base units came (free) from a software company. No gaming. Cards were used for dual screen functionality. Would have thought not much heat in that use case, but looks like I'm wrong.
Liteon PS6351-2 350W I tested the voltage rails under load and they're all within a few % of stated.

If all this is correct, then any second hand graphics card cannot be treated as 'good' until it's tested under a GPU stresser for 12-24 hrs. Which means buying new or used from platforms with no-hassle returns. I suppose that's true of any used PC component though. Never seen intel or amd cpus fail like this. They either work or don't. Guess graphics cards are different.

...Or it could be one of a million other things going wrong there... who knows with cards as old as that.

Yes. Old computers. Odd that's it's only the video card going in all the base units though. I have 7 of these. Testing all of then now with a GPU stresser.

---EDIT----
Ran FurMark to stress the GPU over a couple of hours and every single GeForce GTS 240 failed in each base unit. And I mean they worked for 2 hours before, and now they crap out within 10 minutes playing back a simple HD video. Put different graphics cards in and each computer runs fine.