Strix 970 Overclock Stability Testing

[RazrLeaf]

Is it normal for an overclock to be stable in games, but not in benchmarks?

 

I can set the boost clock on my Strix 970 to 1500MHz (actually boosts to 1538) when I'm playing Tomb Raider, but the maximum that I've managed to run Firestrike at is 1455 (1493 actual).

 

And what's the most effective GPU overclock stability test? I'm looking for one that is free, hopefully takes less than 5 minutes to find instability if there will be any, and allows me to browse Chrome while I test.  I've been using Firestrike and ROG Realbench as stability tests, but they both take a tad too long for my tastes, and Realbench can't be set to just stress the GPU.  I've tried Furmark, but for some reason it doesn't seem to push the GPU as hard as either of the previous 2 (the actual clock speed and GPU usage don't seem to go as high).

[IrshaadH]

Honestly, for me, the best GPU stability test is skyrim with 2k textures, climates of tamriel, flora overhaul and real vision enb full

[Najuno]

Is it normal for an overclock to be stable in games, but not in benchmarks?

 

I can set the boost clock on my Strix 970 to 1500MHz (actually boosts to 1538) when I'm playing Tomb Raider, but the maximum that I've managed to run Firestrike at is 1455 (1493 actual).

 

And what's the most effective GPU overclock stability test? I'm looking for one that is free, hopefully takes less than 5 minutes to find instability if there will be any, and allows me to browse Chrome while I test.  I've been using Firestrike and ROG Realbench as stability tests, but they both take a tad too long for my tastes, and Realbench can't be set to just stress the GPU.  I've tried Furmark, but for some reason it doesn't seem to push the GPU as hard as either of the previous 2 (the actual clock speed and GPU usage don't seem to go as high).

If a GPU crashes at any testing your OC isn't stable at all.

Next thing for dialing in a GPU the right way it isn't as easy as testing for 5 minutes. This should be done taking your time and step by step or you will always be under what's possible for the GPU or crashing all the time. It isn't wise to go to the very limits of the GPU all the time and stress it all the time too much, because it will shorten the life span of the graphics card and if you plan to keep it longer than a year or even two this could get you in trouble.

The best stability tests for GPUs are Unigine Heaven, Unigine Valley and ingame benchmarks because you can run them in loops for some hours. 3D Mark is for me personally the best 3D application based benchmark available because it is very accurate and comes with a database of millions of people, but it isn't in terms of looping a GPU.

[RazrLeaf]

I just find it strange that the 1500MHz overclock will let me play for 16 hours on a weekend without issue, but the Nvidia driver crashes about 2 minutes into Firestrike.

 

Can you run Unigine Heaven/Valley and still do other things (youtube, email, excel, etc.)?  My primary issue with the tests I've been using is they either don't seem to push the GPU hard enough (Furmark) or prevent me from doing other things (Firestrike/Realbench).

[Timmy-P]

Surprisingly, the exact opposite is common, where one can run a loop of, say, Heaven, then crash in a game of choice.

Though if it works for you in games, then you should be fine.

[TheMcSame]

Surprisingly, the exact opposite is common

This. However, from my experience this doesn't apply to firestrike (doesn't mean it can't happen though!).

 

 

OP: The game may not be pushing the card as much as a benchmark is. You could risk it but it may not be stable in every single game. 

 

Personally, if I find that an OC isn't stable in firestrike then I don't run with that OC.