Opinions on Prime95 vs. IBT???

[daveybrat]

For stability testing on my 960T i've been using Prime95 to test with for now. Everyone says that Intel Burn Test is better because it stresses the cpu even harder so you don't need to run it as long.

I just downloaded IBT and plan on trying it out. Just curious as to real world opinions on it.

Thanks guys!

 

[Idontcare]

I'm in favor of using IBT over Prime95 small FFT for the very reasons you mention - it drives the CPU to higher temperatures, and more quickly, such that it flushes out system instability all the quicker in terms of my time when tuning and optimizing a computer for stability purposes.

Note that HT will muck it up, so in my case with my 2600K I not only limit the threads to 4 (versus the default 8) but I also lock the thread affinity to physical cores.

This does make a difference in terms of driving the absolute maximum power-consumption out of the chip which in turn enables higher temperatures during stress testing and all the faster time-to-failure considerations.

 

[daveybrat]

Thanks, i'm gonna give it a try tonight and see how it runs.

 

[Idontcare]

Thanks, i'm gonna give it a try tonight and see how it runs.

Oh, one last thing that matters and I forgot to mention it, IBT becomes more and more efficient in heating up your CPU as you assign more and more ram to it.

So if your goal is to get your CPU as hot as possible then be sure and click that "ALL MEM" button (or however it is labeled in the version you ultimately download).

 

[bryanW1995]

I always use IBT (linX), I don't even bother with prime 95 at all anymore.

Then, when I'm done, I test it out in seti@home for 10 years or so just to make sure that it's really really stable

 

[daveybrat]

I always use IBT (linX), I don't even bother with prime 95 at all anymore.

Then, when I'm done, I test it out in seti@home for 10 years or so just to make sure that it's really really stable

Hehe, so your Pentium 3 computers should be verifiably stable by now then?

 

[Puppies04]

I always use IBT (linX), I don't even bother with prime 95 at all anymore.

Then, when I'm done, I test it out in seti@home for 10 years or so just to make sure that it's really really stable

I will keep my eyes open for the "2500k stability test results" thread in 9 years or so, should make for some good reading

 

[ehume]

I always had the impression that stability testing and heat runs are not the same thing at all. Occasionally you run into people who have run both. They generally agree that Prime95 will bung machines that can run IBT for a whole day.

I use OCCT running Linpack (the software behind IBT) to test the efficiency of fans on heatsinks. I don't expect it to really be doing a stability test for me.

 

[Idontcare]

I always had the impression that stability testing and heat runs are not the same thing at all. Occasionally you run into people who have run both. They generally agree that Prime95 will bung machines that can run IBT for a whole day.

I use OCCT running Linpack (the software behind IBT) to test the efficiency of fans on heatsinks. I don't expect it to really be doing a stability test for me.

Depends where the instability is in the millions of circuits on your cpu. Some are more sensitive to thermal noise, others to the specific sequencing of instructions that are being processed.

Consider that no freely available stress tester programs even come close to testing the mathematical correctness of all 700+ instructions for aggressive stability purposes, Prime95 is merely testing a very small subset of instructions (as is F@H and IBT).

So its a choice of what you are using as a proxy for indicating instability/stability in the rest of the chip that is not being tested for stability or mathematical functioning.

And the program you are using needs to be a mix of instructions that are themselves a reasonable proxy for the types of instructions that you aim to be using with your regular apps.

(this is why stress testing with IBT for an OC'ed rig that is going to be gaming is kinda silly, IBT is a poor proxy for the instruction that will be used in gaming...same with using OCCT for your vcard but you then use the card for gaming, totally different application classes)

But this is the challenge with "free", defining the best is easy, but getting the best for free is impossible. So we try to use the free stuff in ways that are good enough for government work

In the end, whether Prime95 or IBT is first to showing you the weakness in your OC is going to come down to what specifically is failing in your CPU's circuitry.

The fact that IBT seems to give more people that data sooner than Prime95, but not everyone, suggests to me that more often than not a cpu's weakness in OC'ing is going to have something to do with the specific circuits used by IBT's instruction mix (and data accesses) at the temperature with which they are operating versus that for Prime95.

But its not universally true, as it can never be, because neither prime95 nor IBT test all 700+ instructions for correct functioning.

 

[Stayfr0sty]

OCCT all the way!