Originally Posted by juano
I don't know about that. It certainly wouldn't hurt to try, assuming you do it correctly and patiently so you don't fail a dozen WUs, but I wouldn't get your hopes up too high. If you really believe that the instabilities were related web page use and not just general instabilities then it might be worth trying.
I got a single NAN at idle folding instantly at 962 MHz core within just a few moments of the client starting on -advmethods. (This was way back when I first tried to fold on my "gaming stable" OC.)
I reduced my clocks to 950 MHz and had a single NAN after 9 successful WUs. The NAN was when the PC was unattended.
I reduced my clocks by a much more significant value to 925 MHz and had a single NAN after 20 successful WUs. The NAN happened while I was web browsing (with hardware acceleration enabled) on this site.
I reduced my clocks to 918 MHz and completed about 35 to 40 WUs and then had two NANs while browsing on this site, back to back. When I stopped browsing, the NANs didn't occur.
To counteract this, I raised my core voltage in single increments, getting three more NANs (with several good WUs between each NAN) until I came to 904.5 MHz core and 1.138 V being a truly error-free operating point. I've completed over 250 WUs on my Fermi card, most of them -advmethods 7620, 7621, and 7622 projects and I have had a total of 9 NANs. (With the first 5 happening early on, when I was still learning.)
Unfortunately, there's no other program that I've found that can stress the card at this level where I can safely test without hurting Stanford's research. I've kept it to a minimum and I'd like to think that I've run enough successful WUs (with 98% or more time remaining to deadline) that I've made up for the few bad WUs that I caused.