Best way to stress test overclocked GPU?

[Chris Redfield]

Hello everybody,

 

I have MSI GTX 1080 Gaming + which I tried to overclock today.

Since it's ''+'' version I can't overclock memory at all.

I raised power and temperature limit to max and I overclocked core to +150mhz.

I tried to run OCCT and it found 2 errors in first 10 minutes.

I did a research on stress tests and I conclude that everybody suggest different methods.

Many say OCCT is good,other say it pops up fake errors.

I heard that Furmark is great,but then again,I heard it will fry GPU and it's not realistic.

Some say MSI Kombustor is good,other say it's Furmark in disguise.

 

I tried some games that had benchmark tool built in and I got 3 FPS on 1440p after overclock in games like Far Cry 5,Gears 5,Wolfenstein Young Blood-is that good gain of FPS?

 

Can you tell me what should I do to check that my GPU is 100% stable (I plan to play some games after all stress tests are passed) and how to test it and how much?
 

Thanks

[Fasauceome]

the only stress tests I care about in my personal rig:

do the games I play run faster? are the programs I use the most stable under load?

 

any other synthetic tests would just be for benchmarking/e-peen score sharing. I don't really care what my CB score or 3D mark firestrike score is when I'm just playing overwatch and targeting 144 fps.

[Kupal]

6 minutes ago, Chris Redfield said:

Hello everybody,

 

I have MSI GTX 1080 Gaming + which I tried to overclock today.

Since it's ''+'' version I can't overclock memory at all.

I raised power and temperature limit to max and I overclocked core to +150mhz.

I tried to run OCCT and it found 2 errors in first 10 minutes.

I did a research on stress tests and I conclude that everybody suggest different methods.

Many say OCCT is good,other say it pops up fake errors.

I heard that Furmark is great,but then again,I heard it will fry GPU and it's not realistic.

Some say MSI Kombustor is good,other say it's Furmark in disguise.

 

I tried some games that had benchmark tool built in and I got 3 FPS on 1440p after overclock in games like Far Cry 5,Gears 5,Wolfenstein Young Blood-is that good gain of FPS?

 

Can you tell me what should I do to check that my GPU is 100% stable (I plan to play some games after all stress tests are passed) and how to test it and how much?
 

Thanks

Would you like your GPU last longer?

For 3 FPS Boost, Does it you make your game run faster and shoot faster?

 

[Mateyyy]

I personally like using Unigine Superposition and 3DMark Time Spy and Fire Strike.

They're about as heavy, while not being unnecessarily unrealistic, as you can get. And they also double as benchmarks, which is never a bad thing.

 

That being said, what I have found to be the most sensitive to unstable GPUs is folding. My 2070 Super would be stable at around 2050MHz on the core at 950mV in Superposition, 3DMark and any game I threw at it, but when folding it'd constantly spit out bad work units.

[Chris Redfield]

5 minutes ago, Fasauceome said:

the only stress tests I care about in my personal rig:

do the games I play run faster? are the programs I use the most stable under load?

 

any other synthetic tests would just be for benchmarking/e-peen score sharing. I don't really care what my CB score or 3D mark firestrike score is when I'm just playing overwatch and targeting 144 fps.

I don't care for CB score either,I just want to have higher fps after OC and stable PC.

5 minutes ago, WikiForce said:

MSI Kombuster or Time Spy

Is MSI Kombuster safe?

I read that Kombustor is like Prime95 for GPUs.

3 minutes ago, Kupal said:

Would you like your GPU last longer?

For 3 FPS Boost, Does it you make your game run faster and shoot faster?

 

Of course I would like my GPU to last longer,but I wouldn't mind making it's lifespan a bit shorter if I'll get better performance.

As for 3 FPS boost,I know it's not much but it's still something.

1 minute ago, Mateyyy said:

I personally like using Unigine Superposition and 3DMark Time Spy and Fire Strike.

They're about as heavy, while not being unnecessarily unrealistic, as you can get. And they also double as benchmarks, which is never a bad thing.

 

That being said, what I have found to be the most sensitive to unstable GPUs is folding. My 2070 Super would be stable at around 2050MHz on the core at 950mV in Superposition, 3DMark and any game I threw at it, but when folding it'd constantly spit out bad work units.

How long should I run Superposition,3D Mark Time Spy and Fire Strike to be sure it's stable?

 

[Mark Kaine]

Totally not worth it for 3fps and recommended "stress test" : none (it's of no value at all) 

6 minutes ago, Mateyyy said:

personally like using Unigine Superposition and 3DMark Time Spy and Fire Strike.

They're about as heavy, while not being unnecessarily unrealistic, as you can get. And they also double as benchmarks, which is never a bad thing.

Same, but not for ~3 fps ¯\_(ツ)_/¯

 

[Mateyyy]

1 minute ago, Chris Redfield said:

How long should I run Superposition,3D Mark Time Spy and Fire Strike to be sure it's stable?

I just do a run each of Fire Strike and Time Spy to get my results, then do a few runs of Superposition. 

 

If it's unstable, then you'll generally notice it quite quickly (artifacting/crashing).

[Chris Redfield]

3 minutes ago, Mark Kaine said:

Totally not worth it for 3fps and recommended "stress test" : none (it's of no value at all) 

Same, but not for ~3 fps ¯\_(ツ)_/¯

 

Is there any other way to boost FPS other than overclocking (without lowering settings and resolution in games)?

[Mark Kaine]

1 minute ago, Chris Redfield said:

How long should I run Superposition,3D Mark Time Spy and Fire Strike to be sure it's stable?

Best "test" to see if card is stable is heaven imo, it'll crash fairly quickly and you can even do adjustments while it's running in windowed mode. 

 

But as said above, 3fps isn't worth it, unless you were somehow missing exactly those 3 fps. And you don't need a stress test, overclocking is already stress enough for the card - I'm gonna say IF it's a stable OC tear and wear isn't that bad, but to figure that out you should use rather something lightweight like heaven and not something that's putting unrealistic loads like "kombustor"... 

 

20 minutes ago, Chris Redfield said:

+150mhz.

This is probably way too high. Try 50 for a start, also no memory overclocking? Yeah it's really not worth it imo. 

[Mark Kaine]

7 minutes ago, Chris Redfield said:

Is there any other way to boost FPS other than overclocking (without lowering settings and resolution in games)?

Actually yes... Cooling. A good start is to let everything at *default* settings and have all fans, case, gpu, cpu running at max then do a heaven run so you have a baseline. 

 

Then put +50 MHz and run again. 

 

 

Thats probably about the max you can get, and you can care about  less noisy fans afterwards. 

 

 

PS: Also keep in mind there are diminishing returns when overclocking, it's totally possible (and realistic) a card produces better frames at ~+45 Hz than at~75 Hz, you have to test a lot of different settings, hence something nice and easy like heaven is ideal, you can "fact check" with something like superposition if you found some settings you think are good... 

[Chris Redfield]

1 minute ago, Mark Kaine said:

Actually yes... Cooling. A good start is to let everything at *default* settings and have all fans, case, gpu, cpu running at max then do a heaven run so you have a baseline. 

 

Then put +50 MHz and run again. 

 

 

Thats probably about the max you can get, and you can care about  less noisy fans afterwards. 

I already have my CPU 360 AIO fans at max,and GPU fans on 90% after OC.

To be honest I'm not sure how can I turn on case fans?

I don't care for fan noise too much.

But wouldn't +50mhz gave like less than 1 FPS if +150 gave me 3 FPS?

[dilpickle]

You need to test every game that you play.

 

I was running overclocked and played Jedi Fallen Order from start to finish without any problems. Then I loaded up Tomb Raider and it crashed every time on the benchmark. Turned off overclock and it works fine.

 

This is why the overclocking utilities have multiple profiles because every game runs differently. And this is also why I don't bother with overclocking anymore. I ain't got time for that.

[Mark Kaine]

15 hours ago, Chris Redfield said:

I already have my CPU 360 AIO fans at max,and GPU fans on 90% after OC.

So what's your temps then, *under load*(playing games etc)? 

 

15 hours ago, Chris Redfield said:

To be honest I'm not sure how can I turn on case fans?

You need some fan control software from mb manufacturer, or set in bios (software would be preferable at least until you found a setting you want to keep) 

 

15 hours ago, Chris Redfield said:

But wouldn't +50mhz gave like less than 1 FPS if +150 gave me 3 FPS?

True, but that was just an example what I think you can do... as said I don't think it's worth it personally, though I do think it's worth it to keep temps as low as somehow possible as that means less automatic downclocking and can at least make for a smoother experience. 

[Chris Redfield]

8 hours ago, Mark Kaine said:

So what's your temps then, *under load*(playing games etc)? 

 

You need some fan control software from mb manufacturer, or set in bios (software would be preferable at least until you found a setting you want to keep) 

 

True, but that was just an example what I think you can do... as said I don't think it's worth it personally, though I do think it's worth it to keep temps as low as somehow possible as that means less automatic downclocking and can at least make for a smoother experience. 

GPU temps are max 71C during benchmarks and stress tests.

So if I get it right,if GPU gets too hot I may experience automatic downclocking?

If I always have GPU fan at 100% does it mean I will have better performance?