Using Furmark to test stability

Soldato
Joined
25 Dec 2008
Posts
5,976
Location
Sheffield/Norwich
Is there any way to test graphics card stability in a similar manner to Prime/IBT? Using Furmark is all very well but it would seem I have to watch it continuously for any artefacts, which quickly gets boring.

The reason I'm asking is that my 9600GT is not yet stable when folding although appears stable after a couple of minutes of Furmark.. so I have the option of crashing WUs until I get a core/shader/memory overclock that works, or finding a GPU stress program that I can run for longer than a few minutes without having to watch it.

Cheers.
 
Cheers, downloading as I type. And yup, it claims to support nvidia.

Edit: Working well, thanks
 
Last edited:
I've noticed the same behaviour, furmark stable at considerably higher clocks than folding stable. I set up gpu folding, copied the directory and put it somewhere safe. Overclocked the gpu, let it fold for a while. If it didn't crash, put the clocks up, delete the work directory and move the original files back in, and repeat.

Avoids sending loads of broken units off to stanford, and seems to be the most thorough test I've found for graphics cards.

I turn the network off as well for good measure, but that's hardly required.
 
:eek: Anyone know what Tjunction is for an 8600GTS? It's kinda warm in my case and it just hit 93ºC, cue rapid open side of case + window!

So far stable following ATItools clock adjustment, and I'm happy that my clocks are now limited either by artefacting in ATItools (therefore not stable) or by temperature, so I won't be upping the clocks any further! But your method looks as though it could work pretty well, although you'd have to be sure you got a tough WU. :)
 
Sort of, but since it spends a lot of time folding anyway (and is known stable in furmark therefore probably stable in games), you just leave it going. If it starts to throw lots of errors then you drop the clock speeds a bit.

If you're flashing the card it's a bit more awkward to change the clocks, so it's probably best to flash to speeds a bit below what you believe to be folding stable and up them a little bit in windows/use coolbits. Nvflash is great though, cards holding an overclock even when changing operating systems can't be bad.
 
Sort of, but since it spends a lot of time folding anyway (and is known stable in furmark therefore probably stable in games), you just leave it going. If it starts to throw lots of errors then you drop the clock speeds a bit.

That's pretty much what I'll be doing methinks. I only used ATITools to check stability for about a minute on my final clocks, obviously if I get any errors folding I'll drop the clocks a bit.

If you're flashing the card it's a bit more awkward to change the clocks, so it's probably best to flash to speeds a bit below what you believe to be folding stable and up them a little bit in windows/use coolbits. Nvflash is great though, cards holding an overclock even when changing operating systems can't be bad.

Nope I can't be bothered with that, sounds too much like hard work :p though it would be nice not to worry about Rivatuner holding its clocks (it only seems to restore the clocks for one card on reboot, I have to launch the other card clocks manually. Ah well)
 
Set up floppy as dos boot disk from within windows, drop nvflash executable onto it, along with a dll which comes with nvflash.

Boot from floppy. Type nvflash --index=0 -b backup1.rom. Reboot

Open nibitor, open backup1.rom from nibitor. Change clocks to whatever values you wish on the first tab, leave the rest alone. Save as overcl1.rom to the floppy disk.

Boot from floppy. Type nvflash --index=0 overcl1.rom. Y. Reboot.

Finished. Index refers to which card, 90% sure it starts from zero. Nvflash by itself lists your cards. -b is backup. The manual page is pretty good if you want to do more exotic things, crossflashing to the bios of another card perhaps.

So much better than software implementations :)
 
I used 3DMark06 to test my card's stability.

For some reason, the "Deep Freeze" test crashes after a few seconds if the clock isn't 100% stable, even though the others run fine.
 
Just to update, had another failure on the 8600GTS. Tried 3DMark06, it did fine including on Deep Freeze. Did a longer scan on ATITools and it picked up something, fiddled around with the clocks a bit, left it scanning for a few hours while I was out and no errors so it's back up and hopefully it's now properly stable :)
 
Back
Top Bottom