Stress test tool that allows testing of only designated cores?

[darkewaffle]

I've been running into some stability issues lately and I'm trying to isolate the problem. Does anyone know of any tools that let you pick and choose which cores specifically are utilized during a test?

The scenario I'd like to accomplish is essentially "use core 1 only", "use core 3 only", "use cores 2 and 4 only", that kind of thing.

I read that AIDA64 is capable of doing this, however does not have a free version and is priced at $40. Has anyone found any free tools that may serve this purpose?

Thanks.

 

[Ryun]

You might be able to get away with using something like Intel Burn Test, setting the number of threads to 1, and then setting the core affinity to the cores you want to test using task manager.

You can do this by right clicking the process in the task manager and clicking on "Set Affinity...". Then click the core(s) you want to use.

I don't think this will totally use only those selected core(s) though, and you may introduce instability doing this, but it may be what you're after.

 

[Puppies04]

You might be able to get away with using something like Intel Burn Test, setting the number of threads to 1, and then setting the core affinity to the cores you want to test using task manager.

You can do this by right clicking the process in the task manager and clicking on "Set Affinity...". Then click the core(s) you want to use.

I don't think this will totally use only those selected core(s) though, and you may introduce instability doing this, but it may be what you're after.

I( was just playing around with this myself, doesn't seem to work. For some reason it gave me access denied on the core affinity for IBT most of the time (even when running as administrator) the couple of times it did let me set the affinity IBT still loaded all cores evenly.

 

[Kyle.]

Prime95 will show you the individual core that fails the test.

 

[grimpr]

Use Process Lasso for setting affinity and avoiding hyperthreaded logical cores.

 

[kevinsbane]

Start Prime95 testing all cores; then stop the test on the cores you don't want to stress.

 

[taltamir]

I've been running into some stability issues lately and I'm trying to isolate the problem. Does anyone know of any tools that let you pick and choose which cores specifically are utilized during a test?

The scenario I'd like to accomplish is essentially "use core 1 only", "use core 3 only", "use cores 2 and 4 only", that kind of thing.

I read that AIDA64 is capable of doing this, however does not have a free version and is priced at $40. Has anyone found any free tools that may serve this purpose?

Thanks.

1. You are assuming that the problem is at a single core.
2. You are assuming there is anything you can do about it if it is a single defective core.
3. You are assuming that OC stability problems will manifest when only one core is being stress tested (much lower temperature on the chip as a whole)

1 is probably wrong, 2 and 3 are definitively wrong.

 

[WhoFarted]

Start Prime95 testing all cores; then stop the test on the cores you don't want to stress.

+1 Boom!

 

[darkewaffle]

1. You are assuming that the problem is at a single core.
2. You are assuming there is anything you can do about it if it is a single defective core.
3. You are assuming that OC stability problems will manifest when only one core is being stress tested (much lower temperature on the chip as a whole)

1 is probably wrong, 2 and 3 are definitively wrong.

It's just a matter of trying to narrow down possible problems. I'm not testing an overclock, I'm testing why a system with long-term stability would suddenly experience daily hard crashes under load. Somewhere something has abruptly gone 'bad', and at this point it seems most likely in the PSU, CPU, or on the motherboard. Unfortunately I do not have an appropriate replacement for any of those components available to me short of buying new ones on nothing more than a hunch.

Also thanks for the prime suggestion, I was using it for general testing but didn't realize I could just stop specific threads.

 

[taltamir]

It's just a matter of trying to narrow down possible problems. I'm not testing an overclock, I'm testing why a system with long-term stability would suddenly experience daily hard crashes under load. Somewhere something has abruptly gone 'bad', and at this point it seems most likely in the PSU, CPU, or on the motherboard. Unfortunately I do not have an appropriate replacement for any of those components available to me short of buying new ones on nothing more than a hunch.

That doesn't change my assessment of your assumptions. (and there is nothing wrong with being wrong, we are not born knowledgeable)
Weather testing an OC or a stock clock system, if the CPU is the issue then your 3 assumptions are wrong as I explained.

What you need to do is stress test all cores simultaneously and if it proves the CPU to be defective you replace it.

I suggest you get the program OCCT http://www.ocbase.com/
It can be used to test your various components. Then replace components until the issue is resolved. No matter how long the system was stable for, something can break, all electronics are degraded over time when a current runs through them and eventually they will need replacing.

 

[darkewaffle]

I've already run "full" CPU stress tests which is why I am where I am now. Safe mode with Prime 95 on four threads hard crashes in a couple of minutes, but that still does me no good telling me if that's a result of the PSU, CPU, or motherboard.

Testing individual cores has the potential to at least provide more information. If only combinations including core 4 crash, it points to the CPU. If loading up one or two cores does not cause a crash but 3 or 4 does, I would suspect the PSU. If it crashes under all conditions, I'm back to square one. Definitive? No, but it's new data.

 

[Hatisherrif]

Weather testing an OC or a stock clock system, if the CPU is the issue then your 3 assumptions are wrong as I explained.

Weather testing as in setting your system up outside while it's raining to see if it will survive?

 

[podspi]

1. You are assuming that the problem is at a single core.
2. You are assuming there is anything you can do about it if it is a single defective core.
3. You are assuming that OC stability problems will manifest when only one core is being stress tested (much lower temperature on the chip as a whole)

1 is probably wrong, 2 and 3 are definitively wrong.

You can disable individual cores... I've done it before to get a higher OC (I have an X6 and was playing Starcraft 2)...

 

[taltamir]

Weather testing as in setting your system up outside while it's raining to see if it will survive?

thanks for the correction. whether

You can disable individual cores... I've done it before to get a higher OC (I have an X6 and was playing Starcraft 2)...

No, you can set core affinity.
There is no way to disable individual cores. Please prove me wrong by posting a link on how to do it.

I've already run "full" CPU stress tests which is why I am where I am now. Safe mode with Prime 95 on four threads hard crashes in a couple of minutes, but that still does me no good telling me if that's a result of the PSU, CPU, or motherboard.

You cannot find out which part is broken by running tests. You can find out which part is erroring.
If the CPU is erroring then it could be: CPU, mobo, PSU, RAM
If the RAM is erroring then it could be: CPU, mobo, PSU, RAM
If the HDD is erroring then it could be: CPU, mobo, PSU, RAM, HDD, SATA Cable.

What you do is replace parts until it works, the last part you replaced is the one that was broken.

Lets go back to your three assumptions
1. You are assuming that the problem is at a single core. - This is extremely unlikely, most likely the whole CPU is the problem (if the CPU is the problem)
2. You are assuming there is anything you can do about it if it is a single defective core. - If there is a defective core and you identify it, there is jack you can do about it except replacing the whole CPU. Only intel and AMD have access to their die harvesting tools.
3. You are assuming that OC stability problems will manifest when only one core is being stress tested (much lower temperature on the chip as a whole) - While not OCing, you are still assuming that using software to stress test the CPU one core at a time will give results. Most likely testing one core at a time will not give you errors while all cores simultaneously will.