What do people use to test OC

[sandals100]

Hi guys, I'm just playing around with an OC on my GPU and have began stress test the OC. Some tests work and others fail so I am asking you guys what do you use to test your overclocks?

 

NO OFFENSE IS INTENDED IF I DO NOT REPLY TO YOUR COMMENT, REALLY TIRED SO JUST A TAD BIT TOO MUCH EFFORT TO COMMENT ON EVERY ONE. I WILL JUST LIKE IT TO ALERT YOU OF MY APPRECIATION.

[Fate]

Unigine Heaven and games.

[activated1]

FurMark/MSI Kombustor for stress testing, and you can run Unigine Tropics/Heaven/Vally for 3D

[simtransporter]

I like to use FurMark/Kombustor or as I like to call it 'the hairy doughnut' test.

[kaneandtaker]

Unigine Heaven and Valley.

Crysis 1 in-game benchmark. Loop it.

Metro 2033 in-game benchmark 99 loops.

[Caimath]

Kombustor for initial stress testing and then a combination of whatever games I play at the time. Right now I use BF4 64 player maps and Crysis 3 single player.

 

I used to use Unigine Valley, but I get the annoying white artifacting bug with my 780 which annoyed me to no end, so I dropped that.

[sandals100]

Unigine Heaven and Valley.

Crysis 1 in-game benchmark. Loop it.

Metro 2033 in-game benchmark 99 loops.

Dumb question but how do I access the benchmark in Metro 2033?

[Briggsy]

Unigine Valley and Heaven are great back and forth. In my own experience, Heaven seems to find problems faster for Nvidia cards, while Valley seems to finds problems faster for AMD cards.

 

usually after successfully running through both of those a couple times, I run Resident Evil 6 Benchmark (free download) to see whether there was actually any gaming improvement, as its score is dependent on cpu, ram and gpu combined. I think it favours Nvidia, so best not to compare brands on that one.

 

The most vigorous oc test imo is the "new" tomb raider. If the game will run for an hour or two of uninterrupted gameplay without crashing (settings cranked), you're likely 24/7 stable (24/7 clocks, not full load 24/7 lol).

 

Tools like Furmark, Kombuster and OCCT are unecessary imo. best to stay away from them, although kombuster is a decent performance benchmark.

[simtransporter]

Unigine Valley and Heaven are great back and forth. In my own experience, Heaven seems to find problems faster for Nvidia cards, while Valley seems to finds problems faster for AMD cards.

 

usually after successfully running through both of those a couple times, I run Resident Evil 6 Benchmark (free download) to see whether there was actually any gaming improvement, as its score is dependent on cpu, ram and gpu combined. I think it favours Nvidia, so best not to compare brands on that one.

 

The most vigorous oc test imo is the "new" tomb raider. If the game will run for an hour or two of uninterrupted gameplay without crashing (settings cranked), you're likely 24/7 stable (24/7 clocks, not full load 24/7 lol).

 

Tools like Furmark, Kombuster and OCCT are unecessary imo. best to stay away from them, although kombuster is a decent performance benchmark.

I disagree, there's a difference between benchmarking software and stress testing software. Stress testing is used to test stability, whilst benchmarking software is used to measure performance.

Sure you can use benchmarking software to test for stability too, but wouldn't it be better to use a more stressful test for that? I would assume that if it's stable during stress testing (which is supposed to be 'a worst case scenario' type of load), it should also be stable in benchmarks and games too.

 

[AugmentedFury]

I use Heaven, 3DMark, and of course playing games.

[Briggsy]

I disagree, there's a difference between a benchmarking software and stress testing software. Stress testing is used to test stability, whilst benchmarking software are used to measure performance.

Sure you can use benchmarking software to test for stability too, but wouldn't it be better to use a more stressful test for that? I would assume that if it's stable during stress testing (which is supposed to be 'a worst case scenario' type of load), it should also be stable in benchmarks and games too.

 

 

unfortunately you can run furmark/occt for hours, only to load up a game later on and have the display driver crash in five minutes. Same goes with CPUs too. You can prime95 for half a day with no errors, but bluescreen after opening up a web browser.

 

I've experienced these type of false-positive scenarios so many times in the past that I just don't bother with synthetic tests any more.

 

I find non-synthetic tests like Unigine and certain games as well tend to abuse your gpu in ways a synthetic test never will, without requiring nearly as much time.

[activated1]

unfortunately you can run furmark/occt for hours, only to load up a game later on and have the display driver crash in five minutes. Same goes with CPUs too. You can prime95 for half a day with no errors, but bluescreen after opening up a web browser. I've experience these type of scenarios so many times in the past that I just don't bother with synthetic tests any more.

 

I find non-synthetic stress tests like Unigine and certain games as well tend to abuse your gpu in ways a synthetic test never will, without requiring nearly as much time.

 

Never happed to me in my life and I have had and OCed a fair amount of CPUs, not to mention the CPUs of friends and relatives.

[sandals100]

Never happed to me in my life and I have had and OCed a fair amount of CPUs, not to mention the CPUs of friends and relatives.

You may prolly be more educated on the topic however I must agree with @Briggsy I have had my CPU OC set to a certain level then stress tested with Aida/XTU for hours with no problem. The second I start rendering with sony vegas with a google chrome open I get  a BSOD.

[activated1]

Rendering is something different it loads the CPU like a stress test would. I do a fair bit of video encoding and if my overclock is Prime/LinX stable (12 hours) I never had problems. I had cases where it would pass 2-3 hours of stress testing but crash when encoding, but after 12 it worked like a charm. In the C2D days I used Orthos, I would leave it on blend stress, intensity 8 for 24h if it passed that I had no troubles

 

As I said in my first post in this topic, for GPU stability I like to run FurMark for like 30min, and then leave some 3D looping for a while be it Unigine or 3D Mark

 

And yeah you can also use ATI Tool for stress testing, it`s a bit old school but if you overdid it with the overclock it will show you  :rolleyes:

[sandals100]

Rendering is something different it loads the CPU like a stress test would. I do a fair bit of video encoding and if my overclock is Prime/LinX stable (12 hours) I never had problems. I had cases where it would pass 2-3 hours of stress testing but crash when encoding, but after 12 it worked like a charm. In the C2D days I used Orthos, I would leave it on blend stress, intensity 8 for 24h if it passed that I had no troubles

 

As I said in my first post in this topic, for GPU stability I like to run FurMark for like 30min, and then leave some 3D looping for a while be it Unigine or 3D Mark

 

And yeah you can also use ATI Tool for stress testing, it`s a bit old school but if you overdid it with the overclock it will show you  :rolleyes:

Downloaded ATI Tool - Last update Dec 2006

[activated1]

Yep w1zzard stopped development some time ago...

[SlaughterSmurf]

I run Unigine heaven for 15-30 minutes depending how I am feeling, if it doesn't artifact or crash I turn it up by 15 mhz.

[7850OC]

Prime95 for the CPU

 

Valley for the GPU

[Thony]

Unigine Heaven, 3D Mark Fire Strike and I like to check the scores on Cinebench R15.

Furmark is most demanding so if you want to see how your GPU can handle the 100% load, you can run that (it fried some older GPUs just fyi).

[kaneandtaker]

I don't need my GPU to be rock-solid stable unless you are into mining or video editing stuffs. I'm only interested in my GPUs being game-stable.