Good stability test.

S

sblantipodi

2[H]4U
Joined
Aug 29, 2010
Messages
3,765
Hi,
I know that the best stability tests are games but is there some good synthetics benchmark that is good for stability testing?

Furmark? Evga OC Scanner?

Unigine heaven and valley seems to "light" for my GTX980 Ti SLI.

Thanks
 
Furmark is useless for OC stability testing since the card will always throttle. I normally recommend Heaven, but since you think it's too light, just play a 4 hour session of BF4 in a 64 player map and see if your card can hold its mustard all the way through.
 
n=1 said:
Furmark is useless for OC stability testing since the card will always throttle. I normally recommend Heaven, but since you think it's too light, just play a 4 hour session of BF4 in a 64 player map and see if your card can hold its mustard all the way through.
Click to expand...

I would like something "automatic" I don't want to play to test the stability.

why do you think that heaven is ok?
 
heavan pushes pretty much everything in your graphics card outside of compute (I think unigine has another stress test application too which is purely outdoor scenes which I think is valley), so its a good graphics test for stability. If you want compute too, I would suggest a game like BF4 or any game that uses compute heavily.

Furmark isn't really good as a stress test as n=1 stated it only stresses parts of the GPU, and really works like a power virus more than a stress tester.
 
razor1 said:
heavan pushes pretty much everything in your graphics card outside of compute (I think unigine has another stress test application too which is purely outdoor scenes which I think is valley), so its a good graphics test for stability. If you want compute too, I would suggest a game like BF4 or any game that uses compute heavily.

Furmark isn't really good as a stress test as n=1 stated it only stresses parts of the GPU, and really works like a power virus more than a stress tester.
Click to expand...

I have no time to play this days, I want a synthetic that does it for me.
 
hmm not familiar with any compute stress tests outside of games, maybe an in game benchmark, but again don't know of any that would loop so they wouldn't be that good for stress testing.
 
You can try Fire Strike Extreme/Ultra. In my experience they're much more demanding compared to Heaven/Valley.
 
None of the artificial testing programs or canned benchmarks are very good for stability testing if your desire it to determine actual in-game stability of a card. Best way to determine stability is to go play games for a while... that's how we do our evaluations....

Ninja edit - and by games, please note it is plural. There are many cases where a particular overclock is stable in 4 games but will crash in the 5th game we try....
 
David_Schroth said:
None of the artificial testing programs or canned benchmarks are very good for stability testing if your desire it to determine actual in-game stability of a card. Best way to determine stability is to go play games for a while... that's how we do our evaluations....

Ninja edit - and by games, please note it is plural. There are many cases where a particular overclock is stable in 4 games but will crash in the 5th game we try....
Click to expand...


I agree, different games stress the gpu differently and that would be the ideal way to do it.
 
I've found Crysis 3 to catch instability that sneaks past most other games/benchmarks. Though recently Call of Duty Ghosts found one that got past a 30 minute Crysis 3 run... hmm.
 
trandoanhung1991 said:
You can try Fire Strike Extreme/Ultra. In my experience they're much more demanding compared to Heaven/Valley.
Click to expand...

If anything, my own experience is that Fire Strike runs are far too short to catch any subtle instability. For example I can easily bench my 970s @ 1600/8000 on all 3 Fire Strike runs and they'd finish without crashing, then I run Heaven and it crashes about 10 minutes in.

Back to the OP, I recommended BF4 because it was the only game I had that forced me to dial back my OC for full stability, especially on huge multiplayer maps with lots of action going on. I said "4 hour session" because I've had overclocks that appeared to be stable, only to crash about 3 hours in.

If you want something automatic, Heaven is as good as it gets, but I can tell you right now just because something is stable in Heaven for 5 hours doesn't mean it won't crash in some games. There's no Prime95 equivalent for GPUs is the bottom line.
 
Tomb Raider in-game benchmark has worked pretty well for me. You'll get some black artifacts with unstable GPU clocks.
 
n=1 said:
If anything, my own experience is that Fire Strike runs are far too short to catch any subtle instability. For example I can easily bench my 970s @ 1600/8000 on all 3 Fire Strike runs and they'd finish without crashing, then I run Heaven and it crashes about 10 minutes in.

Back to the OP, I recommended BF4 because it was the only game I had that forced me to dial back my OC for full stability, especially on huge multiplayer maps with lots of action going on. I said "4 hour session" because I've had overclocks that appeared to be stable, only to crash about 3 hours in.

If you want something automatic, Heaven is as good as it gets, but I can tell you right now just because something is stable in Heaven for 5 hours doesn't mean it won't crash in some games. There's no Prime95 equivalent for GPUs is the bottom line.
Click to expand...

With the paid versions, you're allowed to loop each individual tests. The first 2 graphics tests are pretty good when looped. Usually I BSOD/TDR if it's unstable.

The combined test is also good when looped since it's pretty intensive on both your CPU/GPU, and will help catch errors as well.
 
I dunno, I have the paid version and never really found looping Fire Strike to be that effective, probably because of the constant idle/load cycling, unlike Heaven where it pegs the GPU continuously. In my own experience Heaven seems more representative of how game stable your OC is, whereas Fire Strike is good for a quickly figuring out where the absolute limits of your cards are.

That being said, I've had OCs crash instantly in 3DMark11 Xtreme where it would be stable in Fire Strike. So 3DMark11 Xtreme is always the first test I run to see if I've pushed my cards too far.

There's also the problem of quantifying stability. I usually game for 3+ hours at a time, so I need my OC to at least be 3 hour stable. I've had OCs that were borderline stable, worked perfectly fine for some games, whereas in others it would crash after an hour. Setting different profiles is one way to go about it, but this is why I don't really waste too much time with synthetics on the GPU, because there really is no Prime95 equivalent for GPUs.
 
like said above, I find Fire Strike Ultra to be sufficient for finding the best max OC, but its not enough on its own. After i find my max OC i'll let heaven loop for 30 mins to an hour. If it doesn't crash, it is pretty much stable.
 
David_Schroth said:
None of the artificial testing programs or canned benchmarks are very good for stability testing if your desire it to determine actual in-game stability of a card. Best way to determine stability is to go play games for a while... that's how we do our evaluations....

Ninja edit - and by games, please note it is plural. There are many cases where a particular overclock is stable in 4 games but will crash in the 5th game we try....
Click to expand...

Mmm... It will be very difficult to find a stable configuration in this way :(

odditory said:
Just curious why even have a 980 Ti then? :)
Click to expand...

Because I hope that in the next weeks I will be able to play, in the weekend at least.

doug_7506 said:
like said above, I find Fire Strike Ultra to be sufficient for finding the best max OC, but its not enough on its own. After i find my max OC i'll let heaven loop for 30 mins to an hour. If it doesn't crash, it is pretty much stable.
Click to expand...

I can pass hours of heaven but luxmark included in realbench crashes after some minutes.
I can play witcher 4k without problems.
If I run luxmark standalone no problem at all.
Is this a case of instability or realbench is just a mess?
 
doug_7506 said:
like said above, I find Fire Strike Ultra to be sufficient for finding the best max OC, but its not enough on its own. After i find my max OC i'll let heaven loop for 30 mins to an hour. If it doesn't crash, it is pretty much stable.
Click to expand...

This.

After that, the only way to know for sure is give it a week or two of intermediate AND varied gaming etc.
 
sblantipodi said:
Mmm... It will be very difficult to find a stable configuration in this way :(



Because I hope that in the next weeks I will be able to play, in the weekend at least.



I can pass hours of heaven but luxmark included in realbench crashes after some minutes.
I can play witcher 4k without problems.
If I run luxmark standalone no problem at all.
Is this a case of instability or realbench is just a mess?
Click to expand...

I find RealBench to be finicky when it comes to GPU testing. I think it might be caused by overclocking or OSD software, but I can't be arsed to check. It caused GPU driver BSODs for me even when GPU is at stock on multiple occasions. I only really use it to check for CPU stability, since GPU-wise it tests using compute, which causes your GPU to drop into P02, which locks the memory speed and you won't be able to test that.

So what I usually do is loop Heaven over night, then Valley, then FireStrike graphics 1 + 2. After all 3 passes, I just go on and play my games. If it TDR/CTD, as long as it's a game that's well known to be stable, I dial back the OC by 5-10MHz.
 
Personally I don´t find it worth my time trying to squeeze everything out of my graphics card (50mhz makes little difference outside benchmark values). You can easily find an area where the stability is pretty much absolute by backing down from the crash area sufficiently and not being greedy. Heaven is fine for this, although I just use couple games with internal benchmarks that I can cycle and look at the temps and such.

Alternative is to waste possibly hours to find optimized setting for some game and only later find out that you need to readjust as new, more demanding game starts causing crashes. Or that your chip/card/memory has slightly degraded over time and is now crashing. **** that. :)
 
Zorlag said:
Personally I don´t find it worth my time trying to squeeze everything out of my graphics card (50mhz makes little difference outside benchmark values). You can easily find an area where the stability is pretty much absolute by backing down from the crash area sufficiently and not being greedy.
Alternative is to waste possibly hours to find optimized setting for some game and only later find out that you need to readjust as new, more demanding game. **** that. :)
Click to expand...

Yea, after i noticed how much little difference their was...i stopped chasing that extra high OC. Waste of electricity, additional heat, and degradation.

I do use firestrike to compare my previously owned cards to my current cards.
 
You must log in or register to reply here.
Back
Top