Best utilities for stress testing in 2020

[Regeneration]

New threads related to unstable overclocks and defective hardware pop up on a regular occasion. It is often asked how to stress test some component. Below you'll find my top picks after weeks of testing - updated for 2020.

GPU
Unigine Heaven
FurMark / MSI Kombuster
*GPUs tend to clock down due to power and thermal limits when running FurMark*
*Lower resolutions put more stress on the VRMs*

Video RAM
Final Fantasy XV Benchmark
*Set to maximum quality and run in a loop*

Memory modules (configuration, timings and overclock)
MemTest64
HCI MemTest (one instance per thread)
*Best run from safe mode without a pagefile*
Linpack Xtreme

Memory modules (physical integrity)
MemTest86

Processor
Linpack Xtreme
Prime95 small FFTs

Integrated Memory Controller
Prime95 custom run of FFTs from 512K to 1024K with 90% of RAM
Linpack Xtreme

Northbridge, bus and PSU
Prime95 custom run of FFTs from 512K to 1024K with 80% of RAM + LuxMark or FurMark running in the background at 720p.

 

[EarthDog]

Thanks for taking the time to write this list. I would love to see reasons behind why some of these were chosen. What makes it the best for each and why you came to that conclusion.

ty.

 

[Regeneration]

RealBench is cute but not demanding enough. One can pass it without a single error for 8 hours and then crash in a game after 2 minutes (real story). RealBench wasn't updated for years now. The only good use for it is to test LLC and power saving features (C states) while overclocking.

AIDA64 is a shareware and most of stress tests aren't good enough. GPU test triggers driver timeouts, CPU test is useless, FPU is just AVX baking, and really doubt if the cache test does anything. Never tried the local disks test. The only useful test is the memory, and there are already free alternatives.

 

[EarthDog]

Sorry about my edit... lol.

Everyone has those stories from each and every stress test is the problem.

I used XXX for XX hours and it was stable. But when I YYY, I had issues. It's best to find what works for you.. even if it isn't the most stressful for a given test.

Thanks for sharing your preferences.

 

[Deleted member 197986]

Cinebench

 

[freeagent]

I quit running P95 many many moons ago. Mainly because I've seen it run for 24 hours and then crash when I stop it, or crash when I open a webpage or open a game lol. Was a waste of time for me. Linpack Xtreme is pretty good, but not the end all be all either, because I've see it pass that then crash during something else. I have to agree with realbench, not very stressful at all, ok if you don't want to listen to your fans though!

Linpack, Cinibench, 3DMark stress loop, SuperPi, wPrime, are some of the tools I use, and not in that order heh..

 

[Vya Domus]

*Lower resolutions put more stress on the VRMs*

Can't see why that would be the case. Technically whatever causes the most power draw will stress the most the VRMs. You need something that keeps the GPU right under it's power limit because otherwise it will just downclock and not stress the VRMs as much.

Compute benchmarks are better for this, when a GPU is running a typical graphics pipeline that incurs some stalls when the shaders aren't doing anything for short periods of time therefore not stressing the VRMs as much. I wrote a basic matrix multiple in OpenCL that puts more load than any other graphics benchmark I've ever tried.

 

[freeagent]

I ran furmark once on my 980 Classi. Normally that GPU pulls about 250w plus whatever the rest of the rig is doing, but with that app the rig was pulling well over 500w. I don't recall the exact number, I really wanted to say 700w, but I don't want to run it again to find out.

 

[Regeneration]

Everyone hated Windows Vista when it first launched. Microsoft improved it over time and in SP2 it was pretty good. But people kept saying "Vista sucks", so Microsoft made some cosmetic changes and released it as Windows 7, still used by many until this very day.

FurMark / MSI Kombuster were improved over the years and now feature several 3D demos in both OpenGL and Vulkan. PhysX demo, Tessellation demo, VRAM burner, and artifacts scanner to detect the tiniest artifact that the human's eye can't catch.

 

[Vayra86]

Can't see why that would be the case. Technically whatever causes the most power draw will stress the most the VRMs. You need something that keeps the GPU right under it's power limit because otherwise it will just downclock and not stress the VRMs as much.

Compute benchmarks are better for this, when a GPU is running a typical graphics pipeline that incurs some stalls when the shaders aren't doing anything for short periods of time therefore not stressing the VRMs as much. I wrote a basic matrix multiple in OpenCL that puts more load than any other graphics benchmark I've ever tried.

Higher FPS = higher temp. If you put a GPU at 99% load it will likely still show somewhat higher temps in the lower resolution runs.

Not really in agreement on the listed apps. Furmark for example... nnnyeh to what end? Stability at sub peak clocks? Youre not seeing the cards end performance or power behavior.

Linpak and long runs of P95... nice if all you care about is how hot things can get, but I question the use of long crunch time when it comes to stability in power state switching and highly varying loads which is what most PCs are used for.

One thing is definitely true, you need several testing apps to get some idea of your stability, your peak temp under sustained load is just one part of that.

Still partial to OCCT because for some reason it will throw you an error extremely quickly if you arent stable. Its a quick method to narrow down on required volts and settings. After that when Im left with a small handful of options, I use some of the other stuff and a long late game play session of TW Warhammer 2 on campaign map to really get a definitive answer. So far so good for this 8700K..

 

[nguyen]

As someone who always monitor the GPU power usage and not the GPU load %, I would say Unigine heaven is a terrible stress test for GPU, it barely stresses the GPU at all.
Now if you want to stress GPU's VRM and thermal solution, use Timespy 4K graphical test 2, GPU would run at its max TDP all the time, generating the most heat.

Note: use Hwinfo64 plugins inside Afterburner to monitor GPU power usage

 

[Lindatje]

The best tests are the programs that are used, for example games.
It has often been heard that people run "stable" on a benchmark program, but as soon as they start gaming (the actual thing they use) it is no longer stable.
I therefore never use any of those programs. I make an overclock and then play a few games for a few days, if there are no problems I know I can overclock a little further.

 

[nguyen]

The best tests are the programs that are used, for example games.
It has often been heard that people run "stable" on a benchmark program, but as soon as they start gaming (the actual thing they use) it is no longer stable.
I therefore never use any of those programs. I make an overclock and then play a few games for a few days, if there are no problems I know I can overclock a little further.

Still depends on which game you play though, you might be stable on one just to crash on another, or the crash could happen randomly after extended gaming sessions. I find the quickest stability test are OCCT, Memtest64 and Timespy 4k Graphic Test 2 stability test.

 

[freeagent]

I’m probably the only one who liked vista

 

[Splinterdog]

Still partial to OCCT because for some reason it will throw you an error extremely quickly if you arent stable. Its a quick method to narrow down on required volts and settings. After that when Im left with a small handful of options, I use some of the other stuff and a long late game play session of TW Warhammer 2 on campaign map to really get a definitive answer. So far so good for this 8700K..

I've just been trying OCCT on a customer PC and yes, it's pretty good with an excellent UI. However, it's no longer as free as the website suggests since the author has limited the free version to one hour usage, with prices at $5 per month or $25 per year. That's a little steep when there are totally free stress testing packages out there.

OCBASE/OCCT : Free, all-in-one stability, stress test, benchmark and monitoring tool for your PC

Ocbase is the home of OCCT, the most popular all-in-one stability / stress testing / benchmarking / monitoring tool available for PC

www.ocbase.com