What’s your go-to GPU stress test?

Parja

[H]F Junkie
Joined
Oct 4, 2002
Messages
12,670
Just curious what everyone’s using these days. Are 3DMark and the Unigine apps still the best all-around options?
 
iv used furmark, and the one built into gpuz actually for quick dirty testing more so on amd cards, since they suffered from that mem lock 0-50mhz bug in drivers,
 
The toughest games I play. I find what is stable and knock 10 to 15MHz off so its fine in very hot weather.
Theres no sense wearing out my gfx card (read: being forced to reduce max overclock because it is no longer stable) with benchmark loads it will never see when in normal use.
If you only game I recommend using your toughest games to test stability.
You can of course knock a little more off the max overclock to be extremely sure.
 
Just curious what everyone’s using these days. Are 3DMark and the Unigine apps still the best all-around options?
I like Unigen's Superposition. It hits all angles (resolution, big textures, high quality shader and lighting effects, physics, interactivity). So, it feels more like an actual game area. And it does it all in a relatively small area. So you can just point the camera in one spot and let the heat rise.

Otherwise, loading into a 64 player battlefield server has always been a great way to tax your system.
 
  • Like
Reactions: Parja
like this
haha like i used to do with Crysis (i know i know) with my GTX 8800 though i melted it near the end, since then gpu tech has improved.
 
Heaven/Superposition for stability and temp checks.
 
Heaven works for me. Any of them really.

I found many games would crash or give a driver error while gaming after using Heaven as the stress test.
The game that had the most issues was Witcher 3 even with vsync on at 1080p 60Hz!
So Witcher 3 became my stress test at that time and havent trusted Heaven since.
 
Just curious what everyone’s using these days. Are 3DMark and the Unigine apps still the best all-around options?
The only benchmark I really care about for actual performance measuring purposes is the games I intend to play. Synthetic benchmarks like FireStrike and TimeSpy are fun to look at, but they're really only useful as a way of measuring who won the "game" of competitive overclocking. Actual games don't scale the same way.

I really like the built-in test in GPU-Z for sanity checking my repairs, though. After I get the board repaired and working, I usually end up cutting new thermal pads, and I sometimes have to trial and error my way to the right thickness. The GPU-Z test is gentle enough to not damage most cards, but intense enough to immediately tell you if you got the heatsink mounting right.

Once I can run the GPU-Z test without overheating, I like Heaven as a means of making sure the card can handle running something resembling a game.
 
I usually use Furmark. Constant synthetic load to make sure frequencies and temps are as high as they can go because the majority of gaming loads won't be as intensive.
 
Heaven is good just to see if a card is performing properly, but it's a fairly light load. Benchmarks can get you close but games are usually the real test.

I am finding COD:MW to be a solid stability test, or maybe it's just sensitive, but I can run clocks quite a bit higher in something like FS than I can that game.
 
The only benchmark I really care about for actual performance measuring purposes is the games I intend to play. Synthetic benchmarks like FireStrike and TimeSpy are fun to look at, but they're really only useful as a way of measuring who won the "game" of competitive overclocking. Actual games don't scale the same way.

Cool story, bro. I asked about stress testing, though, not benchmarks.
 
And in reality, some games make GPU's "cry" much better that any "fake" test you can do....go figure.

Which is a bummer. You would think there would be a good stress test option out there, but when people are still using an 11 year old program (Heaven), it's a pretty sad state of affairs.
 
Honestly, I just throw BOINC at the card with 100% GPU utilization for a few uninterrupted days. If the cooling holds up and there's no instability after 24 hours, that's probably going to heat the GPU up as much as any synthetic test I could download while contributing useful scientific computation in the process.
 
Cool story, bro. I asked about stress testing, though, not benchmarks.
You need to be more specific than "stress testing" then. Which part of the card are you hoping to test, and what type of stress do you have in mind?

Furmark is the closest thing I'm aware of to a test of the power and cooling components, but I wouldn't count on it being a particularly valid test of stability for any purpose other than FurMark, because it's not a very realistic approximation of a game, particularly now that nvidia's GPU architectures include all sorts of extra hardware for accelerating math operations that aren't required for FurMark. If you just want to test that the card isn't going to catch fire, then FurMark is your jam.
 
Uniengine Heaven endless loop until I'm out of stability to get a perfect fail mark and step back from there in larger increments to kinda broadly gain stability without trying to tweak for small gains. I occasionally run futuremark but more for substantiating my gains than stability testing because I have absolutely passed it while unstable.
 
finished my overclock recently. i started with several hours of furmark. once that was stable, i played a couple hours of control to finish it off.
 
The old Furmark is still a bloodbath. Tread with caution.

I had a friend send me some cheap chinese PSU once in the late 00s that was supposedly rated to 750w. Furmark set it on fire and the case fans literally sent pitch black smoke pouring straight up to my ceiling about 2 hours into a stress test one night.

I'll never forget his laugh when I told him about it and he informed me he bought it for seven US dollars straight form the manufacturor.
 
The old Furmark is still a bloodbath. Tread with caution.

I had a friend send me some cheap chinese PSU once in the late 00s that was supposedly rated to 750w. Furmark set it on fire and the case fans literally sent pitch black smoke pouring straight up to my ceiling about 2 hours into a stress test one night.

I'll never forget his laugh when I told him about it and he informed me he bought it for seven US dollars straight form the manufacturor.
So the takeaway is don't buy cheap POS PSUs? ;)
 
I think furmark does a fair jon tbh

Msi kombuster does well too. Its kind of the prime95 for GPUs.

My 2080ti is not stable past 2050mhz even fuly water cooled. Im thinking of taking the card apart and remounting the ek block but I need to get new thermal pads to replace the ones that are on it. Something is getting too hot. Gpu temps rarely surpass 50c so im thinking vrms or something else is getting hot.

My card is the cheaper evga without all the extra sensors. I didnt even realize it when I bought it
 
I think furmark does a fair jon tbh

Msi kombuster does well too. Its kind of the prime95 for GPUs.

My 2080ti is not stable past 2050mhz even fuly water cooled. Im thinking of taking the card apart and remounting the ek block but I need to get new thermal pads to replace the ones that are on it. Something is getting too hot. Gpu temps rarely surpass 50c so im thinking vrms or something else is getting hot.

My card is the cheaper evga without all the extra sensors. I didnt even realize it when I bought it

Yeah I've had cards with the core at like 50c and the vrm >120c. A lot of cards that are under water do nothing or little outside of cooling the gpu with the block.
 
Yeah I've had cards with the core at like 50c and the vrm >120c. A lot of cards that are under water do nothing or little outside of cooling the gpu with the block.

Well the ek is supposed to cool those but i dont know. Ill take the card apart tomorrow or something and find out.
 
The old Furmark is still a bloodbath. Tread with caution.

I had a friend send me some cheap chinese PSU once in the late 00s that was supposedly rated to 750w. Furmark set it on fire and the case fans literally sent pitch black smoke pouring straight up to my ceiling about 2 hours into a stress test one night.

I'll never forget his laugh when I told him about it and he informed me he bought it for seven US dollars straight form the manufacturor.
some old AMD gpus wouldnt have a power limiter (Pre Navi) so running 150% PL on Furmark probably would cause a huge power draw beyond realistic levels (check TPU test when they used furmark vs game power consumption in their reviews or anandtech), But with a higher power limit the power draw would increase even further in gpus like Vega
https://www.anandtech.com/show/11717/the-amd-radeon-rx-vega-64-and-56-review/19https://www.anandtech.com/show/14618/the-amd-radeon-rx-5700-xt-rx-5700-review/15
 
I usually start with Unigine Super Position to see if the card will just crash at the settings I'm trying then if it passes I'll load up MSI Kombustor and hammer it for about an hour to verify stability.
 
XCOM2 seems to generate enormous amounts of heat on the GPU in game, which immediately stops in the main menu despite showing the same graphics as in game. Probably a bug, but it stresses the GPU.
 
some old AMD gpus wouldnt have a power limiter (Pre Navi) so running 150% PL on Furmark probably would cause a huge power draw beyond realistic levels (check TPU test when they used furmark vs game power consumption in their reviews or anandtech), But with a higher power limit the power draw would increase even further in gpus like Vega
https://www.anandtech.com/show/11717/the-amd-radeon-rx-vega-64-and-56-review/19https://www.anandtech.com/show/14618/the-amd-radeon-rx-5700-xt-rx-5700-review/15

I could have sworn I saw a power draw well in excess of 300w on the 5700xt I have, with their stock bios. I'll have to check it out and report, the Ananadtech furmark power draw seems wayy too low compared to what I've seen in person.
 
Back
Top