A glance through this thread will show that it is true: overclocking increases the risk of invalid results.

Obviously, we're not going to stop anyone overclocking, and what you prefer to do and what risks you are happy with is entirely up to you.

But the official position is, and must remain, please don't overclock.

I get errors a couple times a week usually, but most of the time it's on the machines that AREN'T oc'ed.

Is it possible that I'm going to shorten the lifespan of my cpu? Maybe, although I don't believe that. But even if it's true, what exactly does that mean? I have an old AMD K7 that has been oc'ed for years and has survived clogged heatsinks and a variety of other abuse. If it's original design life was 20 years, do I really care if that gets cut in half?

Plus, as long as you keep the temperature of the integrated heatsink (little metal cap on the cpu package) below the Tcase and Tjunction values specified by the manufacturer, the chip should last as long as it is supposed to.

If you have some particularly intense cpu activity on a system that is overclocked you can get errors, but that's where your stability testing regime comes in. I like to run boinc AND Prime95 at the same time. Together, they push core temps up by about 10 degrees celsius on the 3.6ghz Q6600 - as compared to running just Prime95. If the cpu can handle that and not produce any errors, I'm pretty confident that just running boinc and doing normal computing is never going to cause a problem.

However, keeping an open mind, it would be interesting to test the theory. One way would be to compare the error rate for team XS to the overall error rate. I would expect it to be a little higher than average but not more than a couple percent - if that. Since XS'ers are forever tweaking their systems, it's likely that at least sometimes they'll be running on a less than stable system.

I think it's probably safe to assume that the vast majority of team members have at least one rig (might need to add a couple zeros) that is oc'ed. Stats are still updating, so can't see if this is possible.

Oh BTW, Recently found another nice stability tool called OCCT that you can run with SpeedFan to produce some nice graphs of your stability testing. It also seems to find errors a bit quicker. I find that a 20min run while tweaking will get u to that 98% stable point and when you think you're there a longer run (4 hours) will confirm it. Will also run while BOINC is going but that can throw your temp graphing off a bit (idle only).

Well this old ancient machine I'm on now has been overclocked to 3.2 Ghz since I built it in early 2003. It has done Grid/UD, FAD, Rosetta and WCG and still works like new.

[Edit]BTW, it has been on 24/7 for all of those projects. Lol[/Edit]

I tried overclocking as an experiment my gain was only 400 Mhz more 200 mhz on 2 cores and it was n't all that stable... so using stock speed at 3.6Ghz a core the system puts out very well.. other systems that run slower have their profiles adjusted to run DDDT since now the WU are lighter weight and thus will have better results output.