Hi, I've just bought a GTX 580 but I've noticed Furmark no longer works properly on these cards thanks to the OCP. The most recent version of Furmark has an option to disable it but It doesn't appear to be doing anything, Furmark still runs slow and my external power meter only reads 400w in total. How are others coping with the loss of that warm feeling you after your GPU survives a few hours of Furmark? Thanks.
Why needlessly stress your card when no game will come near the torture of FurMark? Read the GTX580 Lighning review and see what Hilbert has to say about FurMark: Now just go and play some games and forget these silly stress tests
I loop Heaven for a couple of hours. Never fails to give a good indication of stability for me, and I OC many many systems per calendar month. Leave it running for ~30-45 minutes, then watch a couple of loops. You can sometimes see visual artifatcs that have specific causes. For example, blue/red/green flashes (looks like camera flashes ) indicates a problem (clocks too high, voltage too low etc). Set Heaven's options to a decent level, however don't over do it. Some cards will run out of VRAM at high resolutions with large amounts of AA and Tessellation enabled, causing the GPU to wait for data to be swapped in and out of VRAM. This waiting allows the GPU to cool, and reduces the over all load making the test pointless. With a 580 and ~1.5GB of VRAM though, it's not really a concern I also agree with BarryB's post. FurMark is nothing more than a thermal virus. A good way of putting it would be that Furmark is to Unigine Heaven what LinX is to Prime95. The former will create a situation that will never, ever be replicated by real world stress, where as the latter will do a good job of pushing load, power draw and temps to levels that are much more real world. Harsher tests aren't always better either. I've had countless machines OC'd that would pass 20 runs of LinX and fail Prime before the 8 hour point. LinX is quick and dirty. All this is my opinion, not absolute fact. I respect others like to do things differently, that's cool. Life would be boring if we were all the same
I don't agree with that, but as it currently stands I don't use it because either the video cards can't afford it, or they driver-control it and don't work in full power mode when it tries to force them to. Furmark is a stress test and tries to draw the maximum out of the video card, like a stress test should. The fact that the modern video cards are so aggressive with voltages-clock speeds & cooling that they can't afford the worse case scenario, is not Furmarks fault. Furmark seems like the PRIME95 of the video cards. I never heard Intel stating that since PRIME95 causes an unrealistic thermal load to the CPU, using it can cause problems to its CPUs, nor it has ever considered lowering the CPU power whne it detects PRIME95 running. But when they decide the clock-voltage & heatsink, they seem to look into the worse-case scenario that stress-tests like PRIME95 provide and make sure they can handle it.
Prime95 does not cause an unrealistic load. It causes a very realistic load and temperatures, after all its purpose is to find Mersenne primes, not completely load out the CPU, with the aim of utilizing every last transistor and drawing as much power as physically possible. It's a real program with a real function. The load temps produced by Prime are very comparable to high CPU usage scenarios such as video encoding and folding. It won't produce a realistic temp for a CPU under load while gaming because few if any games will peg all CPU cores. There are some legitimate applications that will though. It's Linpak/LinX that's the FurMark equivalent, forcing the card to an unnatural load level that no real world aplication with any practical use would come near to producing. A typical example on my system would be: Prime95, 4 workers (30 minutes) : 68c, 69c, 69c, 67c Handbrake (30 minutes into a 45 minute encode) 67c, 69c, 68c, 66c LinX (after 10 runs, max temps) 77c, 78c, 78c, 77c Temps produced often don't even correspond to a more thorough/accurate assesment of stability either. I've had CPU's that have proved unstable using other tests/real world use after LinX/Linpak has completed successfully. I've also had graphics cards that had run FurMark for 45 minutes and passed that needed their clocks reduced after running a console port for a few minutes. Like I said, there's nothing particularly wrong with Furmark, but it being blocked by OCP is no great loss.
I'd argue that "go away" is not the best answer to my question :grin: Furmark is a perfectly valid testing tool and has done it's job for years, the fact is that if you like to do a bit of modest overclocking there is nothing better than Furmark and Prime95 to get the *whole* system up to peak temperature and power draw. If a PC can stand a few hours with both tools running then by most reasonable standards, your motherboard, power supply, cpu and graphics card are stable.
It's your choice to use FurMark or any other util that puts undue stress on your components, but just because all seems rosy with a few hours of FurMark or whatever doesn't mean your system is rock solid. I've seen sytems take hours of FurMark then crash for no real reason when gaming. I had a GTX260 that blue screened the system with FurMark when it hit 80°C but never missed a beat when gaming, so should I have ditched the card because it couldn't handle FurMark? Hell NO!! Why not try running OCCT, FurMark, Prime95, Heaven and Kombustor all at the same time while playing Metro 2033 @ 2560x1600 with DX11?? That should stress your rig enough to satisfy you!
^^ This, exactly this. I could easily alter a couple of sliders to get my cards FurMark stable at 880/4200. However, I'd get Crysis freezing, Driver not responding issues in Metro 2033 and even be booted out of Bioshock. I'd also get blue flashing artifacts in Dirt2 and S.T.A.L.K.E.R CoP. At best, FurMark is useful to get a very rough ballpark idea of what might be stable enough to run some extended tests on, or even do a few hours of gaming with to test (the best test of them all IMO).