Stress test Tool for RTX 2080 ti

[Carsomyr]

I'm looking for the best "free" tool to torture test and test the stability of my overclock for the rtx 2080 ti
any suggestions?

is valley benchmark or superposition benchmark good for this?

 

[erocker]

I like using Valley bench . Stresses the GPU nicely and comparable to running games.

 

[Chomiq]

What about Furmark?

 

[AsRock]

I like OCCT.

What about Furmark?

Haha that shit should be banned.

 

[Chomiq]

I like OCCT.



Haha that shit should be banned.

Hey, he did want to "torture" test it

 

[Athlonite]

I like OCCT.



Haha that shit should be banned.

isn't the graphics part of OCCT just furmark in disguise

 

[Vayra86]

Furmark is pointless for testing an OC, because it doesn't test an OC. It tests power draw, and only that. Not whether it can actually sustain clocks at certain temp or power states.

Valley is a tad lightweight, even at 4K to really stress test an OC. It will likely fall apart in newer/heavier titles.

3DMark still hits the right buttons; TimeSpy being a DX12 API test and FS Extreme still being DX11 torture mode. If your clocks can sustain 3DMark runs, they can probably last in games.

 

[Zubasa]

Furmark is pointless for testing an OC, because it doesn't test an OC. It tests power draw, and only that. Not whether it can actually sustain clocks at certain temp or power states.

Valley is a tad lightweight, even at 4K to really stress test an OC. It will likely fall apart in newer/heavier titles.

3DMark still hits the right buttons; TimeSpy being a DX12 API test and FS Extreme still being DX11 torture mode. If your clocks can sustain 3DMark runs, they can probably last in games.

Correct, Timespy Extreme with custom setting at 8k is good at testing both the GPU and vRAM. This is especially true for graphics test 2.
Also Port Royal is great for RTX cards since it can use a lot of vRAM and stress the RT cores as well.
I usually loop test those to test if an OC is truly stable. The issue with other tools is they just stress the GPU core and use like 2GB of vRAM at most.

 

[Jetster]

Heaven 4.0 or 3DMark Time Spy

Correct, Timespy Extreme with custom setting at 8k is good at testing both the GPU and vRAM.

Extreme is not free

 

[Zubasa]

Heaven 4.0 or 3DMark Time Spy



Extreme is not free

Sure, but what can you do when the free tools are not all that useful?
The only one that is somewhat stressful which comes to mind is Super Position.

 

[Hockster]

Does it crash in a 4 hour gaming session? If the answer is no then your OC is stable.

That's the only test that matters to me.

 

[Zubasa]

Does it crash in a 4 hour gaming session? If the answer is no then your OC is stable.

That's the only test that matters to me.

I mostly agree, although often with newer games you don't know if it is the OC is bad or it is just another buggy AAA game on launch.

 

[Jetster]

That why you use Heaven 4.0 or Timespy. If the OC is too much it will crash

 

[las]

Does it crash in a 4 hour gaming session? If the answer is no then your OC is stable.

That's the only test that matters to me.

You can be stable in some games, and unstable in others. Same with benchmarks.
You have to test in multiple benchmarks and games to be 100% sure it's stable.

 

[HTC]

You can be stable in some games, and unstable in others. Same with benchmarks.
You have to test in multiple benchmarks and games to be 100% sure it's stable.

Perhaps stressing both CPU and GPU with different programs simultaneously?

Something like prime 95 for the CPU part and Heaven for the GPU part and run them @ the same time: if there's instability, it should crash shortly but it should be ran for @ least 12 hours, just to be sure.

 

[trog100]

That why you use Heaven 4.0 or Timespy. If the OC is too much it will crash

as a quick initial test yes.. but just playing your Favorite games for a few hours is the real test..

trog

 

[AsRock]

isn't the graphics part of OCCT just furmark in disguise

First i heard about it, but were the posts that OCCT has killed peoples pc's ?, Either way some common sense like keeping a eye on your temps and knowing your power draw ( which i do though a fan controller ) never had a issue with it.

I noticed the power draw with my system to be 420w which is about 20-30 watt more than any game has pushed the system. In fact the Unity game engine will push a systems power usage pretty hard on it's own. Were most games will put a load o about 180-320w unity engine has always been higher for me.

 

[las]

Perhaps stressing both CPU and GPU with different programs simultaneously?

Something like prime 95 for the CPU part and Heaven for the GPU part and run them @ the same time: if there's instability, it should crash shortly but it should be ran for @ least 12 hours, just to be sure.

No because if you peak your CPU the GPU won't be utilized fully - I always test for stability individually first, then in different games, it can take days or even weeks to fully test everything.

I have experienced many times that I can loop Heaven and Valley for hours and hours, then crash in a game after 1-2 minutes, or get artifacts.

 

[delshay]

You need to be testing in a loop & the best "free benchmark" is UNIGINE Heaven & Valley. ..Heaven would be my first choice, then Valley. Run any of them in a loop.

EDIT @OP: If you can afford RTX 2080ti, then you can afford a paid version of 3DMARK & UNIGINE Superposition.

 

[phanbuey]

i found RTX quake crashes almost instantly if the OC is unstable

 

[HTC]

No because if you peak your CPU the GPU won't be utilized fully - I always test for stability individually first, then in different games, it can take days or even weeks to fully test everything.

I have experienced many times that I can loop Heaven and Valley for hours and hours, then crash in a game after 1-2 minutes, or get artifacts.

chew* from XtremeSystems forum tests stability this way (except he tests his for 24h instead of 12h). You can check for yourself in his Ryzen 3000 topic here.

Since he's a well known and respected AMD overclocker, i figure he knows what he's doing a lot more then most of us.

 

[infrared]

Not saying anyone's way is right or wrong, but this is my method

I only really bother stability testing the CPU and RAM tbh, and not that heavily, half an hour running a memtest aplication, and an hour tops on the CPU with a few different things.. once I know I have a pretty solid foundation I move to the GPU OC but I don't bother stability testing it really. Unstable CPU/RAM can cause data corruption and give you a real bad time, an unstable Graphics card will usually only crash the graphics driver or at worst require a restart.

The only subtlety I'd bother pointing out is with the graphics memory, you can clock it high enough to begin introducting errors but not enough to crash it, the error correction will fix the errors and cost some performance, so IMO it's best to dial this in using benchmarks like firestrike/timespy/heaven/valley etc so you can see when the performance stops improving or goes backwards. For the GPU itself, just play your games and dial it up a bit at a time until you start getting game/graphics driver crashes or artifacts, and dial it back a 30mhz or so and enjoy. Spending hours at full load isn't necessary imo.

 

[las]

chew* from XtremeSystems forum tests stability this way (except he tests his for 24h instead of 12h). You can check for yourself in his Ryzen 3000 topic here.

Since he's a well known and respected AMD overclocker, i figure he knows what he's doing a lot more then most of us.

If you load up all your CPU cores, your GPU usage won't be peaked, this is something you can easily test

When GPU is not peaked, I would not trust the stability test

 

[John Naylor]

I'm looking for the best "free" tool to torture test and test the stability of my overclock for the rtx 2080 ti
any suggestions?

is valley benchmark or superposition benchmark good for this?

Valley is one of them ... I use each of the 3D mark and Unigine benchmarks as they all test a bit differently and you can have a failure on Valley on e day and Heaven another. The scores are also important as, in the age of Boost 3, the logic that highest core / memory almost never has the highest fps. For thermal testing, .... setting up pump and fan speeds, we use Furmark.

 

[Deleted member 50521]

I found VR to be pretty intensive on GPUs. Some OCworks perfectly for monitor gaming while would crash quickly

i found RTX quake crashes almost instantly if the OC is unstable

Same. Probably because it need to utilize the RT cores as well as CUDA cores.