Display crashing but card temps ok

[Pete.]

My pc has an annoying habit of crashing the display driver when the room temperature gets moderately warm and the machine has been on some time. It's a modest system (Dell optiplex 755) that is rock stable except for this one issue. The machine runs 24/7 but on warmer days it will sometimes suddenly give a fault where the display corrupts. The control panel reports that the driver has recovered from a fault, but it will keep doing it over and again. Most times I can see enough to use the start button to shut the PC down otherwise I have the power button set to shut down properly. The only thing to do then is leave it off for at least 30mins to let it all cool down thoroughly, then it will be ok. On cooler days there's no problem at all.

Hardware monitor reports reasonable temps. The GPU is passively cooled with heat-pipes and a large sink. The highest temperature I've seen the GPU reported is 54 degrees but it's generally around 51-52. Searches suggest that they are good up into the high 60's low 70's at least.

I'd really like it if I didn't have to rip out this card. I use it with dual DVI monitors and the system is very quiet with the large passive cooler. I suspect I might have an overheating memory module or something. Is there something I can do to turn down the card performance, perhaps by reducing the memory clock speed etc, to reduce internal temps?

Some relevant info:

[RhynoCZ]

Does it have any auxiliary power connectors? Are they securely plugged in? Is the PSU giving you enough power to the graphics card? Damaged PCI-E slot? Dust in the PCI-E slot? Drop in power would also cause the GPU to report fail. Also, do you have an other GPU to test on that MB? The GPU in question could be faulty.
____________________
'87 Honda XBR 500, '96 Kawasaki ZX7R P1, '90 Honda CB-1, '88 Kawasaki GPz550, MZ 150 ETZ
'95 Mercedes-Benz w202 C200 CGI, '98 Mercedes-Benz w210 E200 Kompressor

[CaNsA]

[Loui5D]

What PSU are you using?

[Pete.]

[RhynoCZ]

Did it work alright before? I guess the PSU or even the MB (It could have some custom MB) is not giving the GPU enough power to work.
____________________
'87 Honda XBR 500, '96 Kawasaki ZX7R P1, '90 Honda CB-1, '88 Kawasaki GPz550, MZ 150 ETZ
'95 Mercedes-Benz w202 C200 CGI, '98 Mercedes-Benz w210 E200 Kompressor

[Pete.]

[chris-red]

It's been known for the solder to crack and when the card heats up it can open up and cause issues. It's what caused the xBox 360 RROD.

One fix is to bake the card in the oven, the idea being it melts the solder enough to fill the cracks.

Obviously it is a last resort.

You can likely replace it with something better and just as quiet for ~£30.
____________________
Well, you know what they say. If you want to save the world, you have to push a few old ladies down the stairs.
Skudd:- Perhaps she just thinks you are a window licker and is being nice just in case she becomes another Jill Dando.
WANTED:- Fujinon (Fuji) M42 (Screw on) lenses, let me know if you have anything.

[orac]

unless dell have improved they are generally a bit crap on the hardware side of things.

PSU is a prime suspect, they are generally speced really close to whaat the machine needs, any changed and the thing craps out. The motherboard is most likely cheap so the regulators that supply the PCIe maybe over heating and browning out - hence the 30 mins to cool off.

What happens if you run it with the side off? can you actively cool the voltage regulators.

I don't think the card needs a reflow, there would normally be some other sign of issues. tearing and the likes are prime examples of such things.

Honestly, if someone comes to me asking to fix a dell, I tell them to go somewhere else as I wont touch the sub standard cheap as cheap can be hardware.
____________________
Current rides - 2016 Triumph Street Triple Rx, 1994 Suzuki Bandit 400 VM, TGB 204 Classic 125cc
"with nothing left to lose, there is everything to gain. It's not the size of the dog in the fight, it's the size of the fight in the dog"

[Pete.]

Follow-up:
Everything's been rock steady now for three days, running at the reduced frequencies (which are the standard frequencies for that class GPU). Core temperature is 2 degrees lower than before, though I think that it was more likely the RAM that was getting over-worked.

I used nvidia system tools to adjust the card frequencies.
____________________
a.k.a 'Geri'

132.9mph off and walked away. Gear is good, gear is good, gear is very very good