TR Forums

I'm trying to determine some overclocking/fan speed settings for my computer and I'm trying to figure out a good way to maximize heat coming out of this machine. I'm running Prime95 to get the CPU up to max temperature, but I don't know what to use to get the video card up to 100% usage.

I could run a game, but ideally I'd like to run something that was a bit more synthetic, so that I could monitor it while watching temperatures. What do you guys use for graphics card stress testing?

You could try Furmark.

For CPUs, linx (or other linpack derivative) is probably the best for max heat (and stability) testing. Use 4-6gb of ram as the problem size and have it run for about an hour.

For GPUs again furmark is popular, but I really can't comment more than that having never used it myelf.

A lot of demos are stressful. Try cdak! Looks good too, and it's only 4k!

I used "Furmark" it shows temperature, supports multi gpu configurations, if I recall correctly shows fan rpm but do not take my word for it. This test runs in continuous loop until interrupted by the user, so leave it running for a least an hour. Also, it stresses the PSU to the maximum.

I'd run Prime95 and GPU F@H.

Prime95 + Furmark is what I consider a properly sadistic load.

So now I have a new problem. I ran Furmark and Prime95 for three hours and had no problems, but then had one of my two monitors go dark during normal use (word processing/web browsing) on multiple occasions. Now I've backed my overclock down and it's working fine. Why would a torture test work fine, but normal use cause a problem? I suppose I could try a longer stress test on the CPU, but I've generally found in previous tests that CPUs that crash do so in the first 15 minutes.

TurtlePerson2 wrote:

So now I have a new problem. I ran Furmark and Prime95 for three hours and had no problems, but then had one of my two monitors go dark during normal use (word processing/web browsing) on multiple occasions. Now I've backed my overclock down and it's working fine. Why would a torture test work fine, but normal use cause a problem? I suppose I could try a longer stress test on the CPU, but I've generally found in previous tests that CPUs that crash do so in the first 15 minutes.

Your overclock could be pushing the idle clocks above what's stable for the idle voltage. I had a similar issue with overclocking my i5 750.


+1 for Furmark btw, that program will roast your gpu and find any load based problems in a very minimal amount of time.

I would also test Heaven benchmark http://unigine.com/products/heaven/ or F@H. I find that sometimes Heaven will cause instability quicker than Furmark. Power usage is fairly similar though.

Run Furmark + LinX for the maximum power consumption.

For CPU/mem subsystem stress testing, I've come to love IntelBurnTest (http://www.xgamingstudio.com/forum/showthread.php?9-RELEASE-IntelBurnTest-v2.52). I mainly use it to test OCing stability, and for new, stock builds. Despite the name, it can pretty much work on any modern platform released in the past couple of years, AMDs included. A quick test can raise temps by at least a good 10° C or more from idle, depending on your system's cooling capability, more if you used heftier test settings.

For graphics, I agree with others here, Furmark (http://www.makeuseof.com/tag/test-graphics-card-stability-furmark-gaming/) can be a good test as well. Just make sure you've got good cooling and a decent PSU capable of sustaining your vidcard's load needs.