Testing the Tests – CPU Stability

Conclusion

If my results haven’t proven anything to you, think about it this way. You recently built your computer and you’re trying to overclock it. Since overclocking is an “YMMV” sport, you cannot assume that it’s going to hit 5GHz because another similar binned processor is hitting 5GHz. You have to apply the guess and check model. You then need to test at certain intervals for stability. This can be a long and tedious process with a ton of frustration and BSODs (my counter is at 29). The idea is to pick a program designed to find errors, and do it as quickly as possible so you can move onto your next step. I have found that Intel Burn Test is the right test for that job. After completing Test 1, I knew my CPU was fully stable at 190BLCK, but I also knew it failed at 195BLCK on all tests. Test 2 was derived by setting the BLCK value to 190 + 1, and then checked for stability – rinse and repeat. I chose IBT since it has the shortest duration. IBT passed all the way up to 193 BLCK, but failed immediately at 194. This process took me roughly 20 minutes to complete. I then wanted to see if either Prime95 or OCCT could detect an error at 194 BLCK as quickly as IBT did. OCCT said my computer was stable after 15 minutes of running it, and Prime95 was no different. There obviously was a discrepancy between the programs.

What I would like you to take away from this test is that 3 minutes of stability testing is by no means an accurate representation of true system stability, but it can quickly get you on your way to the next and final step.

I personally think Prime95 is an aging test. It shouldn’t be the main stability test for your CPU. In fact, none of them should be. From here on out, I will personally be using a combination of OCCT and IBT. The reason for my choice is as such; IBT wins out at quickly determining CPU stability so you can move onto the next test. It offers the ability to benchmark the CPU to test if the overclock provides any actual advantages to system speed. It does lack the ability to stress the CPU continuously over a set period of time.

If you recall the results from IBT’s test one, you can see the CPU temperatures along with CPU usage are varying from 100% during the test, to 0% during its “break time.” This is where OCCT shines. Besides the 1 minute monitoring at the beginning and the 4 minute cool down period at the end (which as MrK pointed out, this does not happen if you choose “infinite loop”), the time in the middle is full load. As you can see in my Prime95 first test, over a long duration my water temperatures increased dramatically. With IBT, your cooling solution has the time to recuperate, with OCCT – it does not.

Nobody wants to build a computer and then spend weeks or months checking for stability. My recommendation to you is to use IBT with 1-5 iterations at certain increments of overclocking. Keep going until you’re satisfied with the overclock. Once you’ve reached that stage, run OCCT for 24 hours, and then run IBT for 150 iterations. If you have no errors – you’re free to go.

If you’re not overclocking, just run OCCT and IBT for 24 hours and 150 iterations respectively.

Lastly, go and enjoy what you just paid money for. That’s the reason why we should check for stability. If it continuously fails – you’re not going to enjoy it.

Hope you found this useful!

This review was done by me for ABT exclusively. Please honor my request to keep it exclusive. You may link to this thread, or even use my results as long as you give credit where credit is due – to both ABT and me.

– Michael Turner