Disable GeForce GTX 580 Power Throttling using GPU-Z

[RavenXXX2]

Lol they are not worthwhile tests, any demanding games is good enough to test stability.

 

[^Andy^]

Lol they are not worthwhile tests, any demanding games is good enough to test stability.

Rubbish - if the gpu cannot work reliably at 100% of it's claimed speed it is NOT stable

 

[jigger]

If your card pulls over a set amount of amps then the Fermi power monitoring will set in. If Nvidia later decide that 250 watts running Crysis 2 is to much then they have the option to throttle the card if the power monitoring system is running.

Looking at the GTX480 its seems Fermi is good for 350 watt and anyone running an overclocked and over volted GTX480/GTX470 is already pulling over the throttling point of a GTX580.

 

[Marvin]

The other issue for using Furmark for stability testing is that it actually uses very little memory. I was running it the other week at 1920x1080, 32xAA, and it was using something like 350MB Video RAM (according to Afterburner). Kind of pointless on the whole...

Fire up Metro at the same resolution, and you've got 900MB usage or something.

Even Black Ops with 8x AA is using 900MB or thereabouts...

 

[Orangey]

All it does is render a furry doughnut, not surprising. Not sure what your point is though, unless you're saying unaccessed ram chips make it an incomplete test.

 

[Marvin]

All it does is render a furry doughnut, not surprising. Not sure what your point is though, unless you're saying unaccessed ram chips make it an incomplete test.

Well, yeah - it's not a very good test of GPU stability if it doesn't really test the RAM.

 

[kylew]

Yeah no wonder they wanted to cheat furmark, so does this mean the 580 has a higher TDP than a 480?

I thought the very fact that they put a limiter on it showed this. If it had a lower TDP than the 480 then they wouldn't have needed it at all.

 

[kylew]

Rubbish - if the gpu cannot work reliably at 100% of its claimed speed it is NOT stable

That's not how it works, furmark puts a GPU under a level of stress that never occurs in games, ever, simply because the way games work.

 

[kylew]

No need to crash 'your' car because the manufacturers have already been legally obliged to do so using one of their own cars.
The point is that it IS a worthwhile test because loading a processor to 100% should never lead to destruction or massive instability of a GPU (doesnt matter if its YOUR processor or one of the manufacturers samples).

If it can't handle 100% load then it is NOT stable at that clock speed.

100% load, and the load produced by furmark aren't the same, I think that's where you're confusing things. 100% load is fair enough, but it's not quite that simple. Think of it like furmark is pushing it beyond 100%, and that's not just on the GPU, it'll be on the power circuitry too, that's not good for anything. I think the difference with a CPU is that there are so little threads in the first place that it doesn't make much difference, but to load all of a GPU's stream processors (all how ever many thousand) is beyond what you'd call "100% load".