Best Burn/Stability Test Software

[Garretta]

.

 

[Rossi~]

You'll get loads of different replies on this, but here we go.
I use the following:

IBT 10 runs (CPU)
Prime 95 Blend (12 hours) (CPU)
OCCT Error Check (GPU)

Whatever for temperature readings.

 

[martinje]

Everyone has their own methoed. The main thing is keep an eye on temps.


IBT 25 passes
Prime 95 Blend 4 hours
OCCT 1 hour random CPU test
OCCT Error Check GPU

 

[rive22]

I do these in this order throughout the duration of finding my overclock.

Mobo & Disk Checks
Memtest : Overnight
Furmark : 15 minutes
OCCT : 10 Minutes
P95 Large FFT : 30 Minutes
OCCT : 1 Hour
P95 Small FFT : 2 Hours
Intel BT : 25 runs
P95 Blend : 12 hours
Intel BT : 60 runs

 

[mmmmmdonuts]

LinX - 10-15 passes
F@H - overnight
Furmark - 1 hour
memtest86+ - 2 passes

 

[nzaneb]

Furmark and Kombustor are a sure way to kill a GPU. You're new 580 is built to sense those programs and throttle down to save itself. DO NOT use these useless programs. Run the applications you will be using.

CPU - 10 passes of IBT maximum is just fine for a daily driver.

 

[sadbuttrue]

Passing stability tests does not always mean a stable system. When the voltage/load varies is where most problems occur, so running a 100% load test will only help show an unstable system. This is especially true for graphics cards. You can run OCCT GPU all day long with zero errors, only for an overclocked card to crash in certain games.

 

[Rossi~]

Passing stability tests does not always mean a stable system. When the voltage/load varies is where most problems occur, so running a 100% load test will only help show an unstable system. This is especially true for graphics cards. You can run OCCT GPU all day long with zero errors, only for an overclocked card to crash in certain games.

I'm not entirely sure that's correct regarding OCT Error Check and GPU stressing. It's the the most efficient and stressful way to test a graphics card as it is crunching numbers. You could be "stable" playing a game for for hours yet run this and get 50 errors in 5 minutes.

 

[Forceman]

I'm not entirely sure that's correct regarding OCT Error Check and GPU stressing. It's the the most efficient and stressful way to test a graphics card as it is crunching numbers. You could be "stable" playing a game for for hours yet run this and get 50 errors in 5 minutes.

And for many people that's just fine. If you can't see the errors on screen, and it doesn't affect stability, then the presence of OCCT errors isn't necessarily a show-stopper.

 

[cyclone3d]

For all of you still spouting that memtest86+ is a good way to test RAM stability.. IT IS NOT.

It is only good to show blatently faulty RAM.

If it is stability you are having trouble with and it shows up in memtest86+, it should have been super easy to tell even without running memtest86+.

Memtest86+ does not put enough stress on the RAM to show much of anything.

If you want to do an actual RAM only stability test, use HCI memtest. It will show up errors within a few minutes instead of possibly an error after many hours or days that memtest86+ would require.

And I still wouldn't even trust HCI memtest for RAM stability testing on an overclocked system.

 

[sadbuttrue]

I'm not entirely sure that's correct regarding OCT Error Check and GPU stressing. It's the the most efficient and stressful way to test a graphics card as it is crunching numbers. You could be "stable" playing a game for for hours yet run this and get 50 errors in 5 minutes.

I wasn't implying OCCT being bad for stressing a card. Clearly you should always strive to be stable in every stress test. If it fails in OCCT it will likely fail in game too.

What I am saying is that being OCCT stable does not mean you are 100% stable, due to where most components fail (during voltage/load transitions). By the same token, being Prime95+IBT stable also does not mean you are 100% cpu stable. Overclocking forums are littered with people complaining about random crashes, insisting they are stable because they can past x number of tests.

 

[nderscore]

When I worked for a boutique system vendor, this is what we would to ensure system after an OC:

IBT: 16 runs to test CPU/RAM (once ran Prime95 for 80 hours and same system failed within 10 min of IBT)
FurMark: atleast 1 hour for GPU

Monitoring programs: CPU-Z, HWmonitor (though I take the temp readings with a grain of salt)

 

[Term-X]

While I used IntelBurnTest to test my CPU overclocking, just remember that these programs do not guarantee that it is stable. For example, I ran IBT fine for up to 2 hours or even just 20 minutes and at certain settings it would crash AFTER the tests were complete.

You just have to toy around and use all the common apps you would use.

 

[Tsumi]

Yeah, not all stability tests are accurate.

My system was completely stable with LinX run 20 times at the maximum memory I can get. However, as soon as I started SMP FAH, I get artifact errors, and had to reboot the system. Backing off the overclock by 100 mhz completely cleared up the artifact issue with FAH.

 

[Rossi~]

I wasn't implying OCCT being bad for stressing a card. Clearly you should always strive to be stable in every stress test. If it fails in OCCT it will likely fail in game too.

What I am saying is that being OCCT stable does not mean you are 100% stable, due to where most components fail (during voltage/load transitions). By the same token, being Prime95+IBT stable also does not mean you are 100% cpu stable. Overclocking forums are littered with people complaining about random crashes, insisting they are stable because they can past x number of tests.

Ah right, gotcha sorry read your post wrong.
Indeed I found my last system was 12 hour Prime95 stable and 12 runs of IBT was fine but random crashed once a fortnight or so. In the end I ran 24 hours of Prime95 and it failed after 18ish.


Also, those people who run Furmark for an hour, do you sit there staring at it for an hour looking for artifacts? :/ It's not really the best stability test to start with let alone leaving it for an hour unwatched.

 

[sirmonkey1985]

10-20 passes of IBT at highest memory setting and thats it.

Also, those people who run Furmark for an hour, do you sit there staring at it for an hour looking for artifacts? :/ It's not really the best stability test to start with let alone leaving it for an hour unwatched.

heck no. i just watch the evga precision monitor or wait for the drivers to crash. typically the drivers crash before i even see any artifacts. but usually i do my research before hand and know exactly what the card can or can't do.

 

[AdvanSuper]

It really comes down to personal preference between LinX, Prime95, OCCT and IBT for the CPU. With GPU's there are personal preferences as well such as OCCT, Kombuster and FurMark. Also with GPU's some people usually just run 3D Mark and wait for drivers/clocks to crash as error checking.

Folding will put your rig through hell though and let you know if it's stable.

 

[djoye]

10-20 passes of IBT at highest memory setting and thats it.

I'm scared of 20 passes of IBT. heh I was at [unspecified location] where I have access to equipment and replacement parts and I decided to stress test your average business laptop. These things idle at 40-48C (RealTemp) and hit up to 90C at full load probably because of everything being so compacted together. Also note that this was an i5 CPU, not sure the exact model.

Anyway, I let it do 20 runs while I was out doing something else, came back and it was locked up. Rebooted, locked up after loading Windows. Replaced the CPU and memory, still locks up after getting into Windows. Replaced the motherboard and the system then ran fine. heheheh Nothing like being able to break things and have the parts to fix it. I bet the laptop ran 90C for the duration IBT ran until the motherboard died.

For IBT, I've found that 5 runs doesn't seem adequate so 10 is probably a good number. I also have to manually set it to 8 threads (i7) or else it won't fully load the CPU when set to 'auto'.

 

[mmmmmdonuts]

Folding will put your rig through hell though and let you know if it's stable.

I can't agree with you more. Some of my rigs have passed all my tests and then once I fire up F@H it will crash in an hour or two. To me its the ultimate benchmark.