Stress testing- what's the best software?

[killcomic]

I've been overclocking my new 7700K and I've been using a few programs to test stability and I've been wondering which is the most valid.

 

Realbench seems to have Luxmark crashes. An apparently common problem with Nvidia drivers. This happens when overclocked or stock.

 

Aida64 runs with no problems for 2 hours but it's a trial version. I only got a couple of days left before it locks me out.

 

Intel Extreme Tuning Utility seems very gentle on the CPU and I wonder how reliable are the results.

 

Prime 95... not even touching it, specially the new versions. I don't need another heater in the room.

 

Any of you guys know any good, reliable and free stress tests I can use?

[xDylanio]

I feel like i destroyed 5 years of my cpu's life with those 10 stress tests i took in prime 95...

[aisle9]

There's no "one" best one. A combination of several is the best way to find out what your CPU can take. These days, I've taken to a combination of Intel XTU, Prime95, OCCT, Firestrike, Kombustor, Heaven/Valley and RealBench to push my overclocks. Yes, I use Firestrike, Unigine and Kombustor to test a CPU overclock. A stable 4.8GHz is nothing if your PSU can't drive enough voltage to sustain it while your overclocked GPU is screaming for more.

 

Luxmark crashes in RealBench can usually be fixed by uninstalling your GPU drivers with DDU, then doing a clean install of a fresh download. I've also randomly had Luxmark crashes solved by just exiting Afterburner on a couple of older Kepler cards (I want to say 660 Ti and 730).

[Belgarathian]

Prime95, AIDA64, Furmark, Unigine Heaven (or similar). 

[scottyseng]

I personally do 20 runs of Intel Burn Test at maximum, but it runs very hot / your CPU may never reach these temps.

 

However, since doing so, I have never had a issue with a PC that I overclocked that could survive that many runs of Intel Burn Test.

[killcomic]

8 minutes ago, xDylanio said:

I feel like i destroyed 5 years of my cpu's life with those 10 stress tests i took in prime 95...

using Prime95 is akin to smashing your CPU with a cricket bat then asking it to do 10 laps and 50 push ups.

I used to use Prime95 all the time, then I noticed how over the top temperatures were compared to what I saw during full load with anything else. Needless to say, I stopped using it.

[killcomic]

1 minute ago, scottyseng said:

I personally do 20 runs of Intel Burn Test at maximum, but it runs very hot / your CPU may never reach these temps.

 

However, since doing so, I have never had a issue with a PC that I overclocked that could survive that many runs of Intel Burn Test.

I was wonder about IBT. I've heard it's mostly to test thermal capacity and not very good at testing stability.

[scottyseng]

7 minutes ago, killcomic said:

I was wonder about IBT. I've heard it's mostly to test thermal capacity and not very good at testing stability.

Interesting. I heard that IBT was the super overkill one, even beyond Prime 95 (Though I hear the newer version of Prime95 have gotten even hotter).

 

No experience with RealBench, AIDA64, or Intel XTU either. I kind of render a lot so I kind of go for the overkill stress tests. I think I did RealBench and AIDA64 before and they were "stable" but the overclock would crash when rendering for me. This was back in the Ivy Bridge days though...maybe they've gotten better now.

[killcomic]

Just now, scottyseng said:

Interesting. I heard that IBT was the super overkill one, even beyond Prime 95 (Though I hear the newer version of Prime95 have gotten even hotter).

 

No experience with RealBench, AIDA64, or Intel XTU either. I kind of render a lot so I kind of go for the overkill stress tests. I think I did RealBench and AIDA64 before and they were "stable" but the overclock would crash when rendering for me.

mmmh.... Maybe doing a run of IBT could be in order,,,

[2FA]

While stress tests are useful for finding out if it's initially stable, practical tests also help. I've overclocked many times and the stress tests didn't cause a crash but ended up playing BF4 the next day which did cause a crash due to instability.

[WereCat]

OCCT

[scottyseng]

1 minute ago, killcomic said:

mmmh.... Maybe doing a run of IBT could be in order,,,

Just be weary, it does run really hot. Like my CPU is 64C under Intel Burn Test with a H105 with three delta fans on it (2700K at 1.45v / 4.9 GHz).

 

Under normal rendering load, the temps are only 46C with the fans not even approaching 40% speed.

 

I actually had to upgrade to a H105 because my hyper 212+ with the delta fans couldn't keep it cool enough for overclocking.

 

That being said, nothing's wrong with running multiple different tests.

[killcomic]

Just now, DeadEyePsycho said:

While stress tests are useful for finding out if it's initially stable, practical tests also help. I've overclocked many times and the stress tests didn't cause a crash but ended up playing BF4 the next day which did cause a crash due to instability.

Good point!

[killcomic]

Just now, scottyseng said:

Just be weary, it does run really hot. Like my CPU is 64C under Intel Burn Test with a H105 with three delta fans on it (2700K at 1.45v / 4.9 GHz).

 

Under normal rendering load, the temps are only 46C with the fans not even approaching 40% speed.

 

I actually had to upgrade to a H105 because my hyper 212+ with the delta fans couldn't keep it cool enough for overclocking.

 

That being said, nothing's wrong with running multiple different tests.

Okay, sounds a tad beasty. I might do a run and see where it takes me.

[killcomic]

2 minutes ago, WereCat said:

OCCT

I'm a bit afraid of that one. I've read that it's up there with Prime 95.

[scottyseng]

6 minutes ago, killcomic said:

Okay, sounds a tad beasty. I might do a run and see where it takes me.

Usually in my experience, the first two passes will find a majority of OC instability.

 

I just do the 20 (sometimes 10) run thing because AIO coolers (all the rage these days) take around 30+ min to reach equilibrium on temps.

 

Other note is that apparently IBT gets worse as you give it more RAM / RAM speed to play with.

[WereCat]

7 minutes ago, killcomic said:

I'm a bit afraid of that one. I've read that it's up there with Prime 95.

Only if you enable AVX Capable Linpack

[killcomic]

Just now, scottyseng said:

Usually in my experience, the first two passes will find a majority of OC instability.

 

I just do the 20 (sometimes 10) run thing because AIO coolers (all the rage these days) take around 30+ min to reach equilibrium on temps.

 

Other note is that apparently IBT gets worse as you give it more RAM / RAM speed to play with.

Oh great, IT GROWS!!!
Not sure I want to do it now.

[killcomic]

Just now, WereCat said:

Only if you enable AVX Capable Linpack

Ah, so you can turn off AVX stress. Well, that makes things a tad safer.

[scottyseng]

2 minutes ago, killcomic said:

Oh great, IT GROWS!!!
Not sure I want to do it now.

Haha, I'm not sure how true it is though. I've only read about it on the forums here.

 

Though I do think increasing memory speed will make your overclock back down a bit because the memory controller has more to deal with.

 

I did a build recently with a 5820K (Yeah, we wanted Ryzen, but all motherboards were sold out), and at 2133MHz (default), the overclock was 4.6GHz stable. Running at the RAM's full speed of 3200MHz, the overclock is only stable at 4.4GHz.

 

Have fun cooking the CPU though. haha.

[SpaceGhostC2C]

Personally, I like OCCT. You can monitor everything (or 90%) you need from within, applies a substantial load without just burning it like Prime95, and it fails due to instability faster than other alternatives I've tried.

 

27 minutes ago, killcomic said:

I'm a bit afraid of that one. I've read that it's up there with Prime 95.

Well, I like OCCT because it heats the CPU less than Prime95  It's not a huge difference, but the main point for me is that, at least in my case, it took longer for Prime95 to fail, and it was more likely to fail due to high temps, leaving the question of whether it was just a matter of improving cooling or not. OCCT can eventually generate a lot of heat, but it failed due to instability much quicker, meaning that I got useful information about the OC attempt before heating the CPU to the limit, and without waiting long hours at high temps to "confirm stability".

[killcomic]

45 minutes ago, SpaceGhostC2C said:

Personally, I like OCCT. You can monitor everything (or 90%) you need from within, applies a substantial load without just burning it like Prime95, and it fails due to instability faster than other alternatives I've tried.

 

Well, I like OCCT because it heats the CPU less than Prime95  It's not a huge difference, but the main point for me is that, at least in my case, it took longer for Prime95 to fail, and it was more likely to fail due to high temps, leaving the question of whether it was just a matter of improving cooling or not. OCCT can eventually generate a lot of heat, but it failed due to instability much quicker, meaning that I got useful information about the OC attempt before heating the CPU to the limit, and without waiting long hours at high temps to "confirm stability".

So it's not as bad as Prime 95. Good stuff! Might give it a run to see how well it goes.

[SpaceGhostC2C]

1 hour ago, killcomic said:

So it's not as bad as Prime 95. Good stuff! Might give it a run to see how well it goes.

Disclaimer: I was OCing FX cpus, so you probably want to listen to @WereCat's advice