Is Heaven a reliable/worthwhile stability test any more?

[Nine Iron]

Getting annoyed with Unigine Heaven at the moment - I get a clock I think is stable and Heaven runs for two hours straight, but I start the rig up a few days later with the very same settings, and get a CTD within the first minute!

Reading up on this suggests that Heaven is particularly "crashy" as benchmarks go, and people are ending up with rigs that are stable with pretty much every other test. I know that benchmarks don't cover every eventuality, but am I safe relying on Valley for a first approximation for stability? Heaven's unpredictability is kind of defeating the point of using it as a benchmark at the moment!

 

[Excalabur50]

Tom uses OCCT for all his testing on cpu's and valley for gpu's if im correct

 

[Kaapstad]

Valley is awful for testing @1080p as it bottlenecks cards.

Heaven 4 is very temperature sensitive which means if things are getting too warm it will crash if you are pushing an overclock. This actually makes Heaven quite useful in sniffing out dodgy overclocks.

As to testing for stability you should use a wide variety of software as there is no single solution.

 

[Nine Iron]

Valley is awful for testing @1080p as it bottlenecks cards.

Heaven 4 is very temperature sensitive which means if things are getting too warm it will crash if you are pushing an overclock. This actually makes Heaven quite useful in sniffing out dodgy overclocks.

As to testing for stability you should use a wide variety of software as there is no single solution.

There are no temperature issues, I assure you - the cooling mods I've added keeps the card under 50C at full load.

Would turning the settings down in Valley make a difference? I've never had GPU-Z running with that one so I'm not sure of the usages.

 

[WYP]

Valley is awful for testing @1080p as it bottlenecks cards.

Heaven 4 is very temperature sensitive which means if things are getting too warm it will crash if you are pushing an overclock. This actually makes Heaven quite useful in sniffing out dodgy overclocks.

As to testing for stability you should use a wide variety of software as there is no single solution.

Exactly this, multiple tests need to be used to validate overclocks, especially on GPUs. Tom uses a range of tests to validate overclocks, using 3DMARK, Unigine and a variety of games. If we publish overclocked GPU results it passed our whole benchmark suite.

 

[Kushiro]

Just test using some basic ones and then with the games/programs that you would like to utilize it for. It may be stable in one bench but crash in another one or some other games/programs.