WCG do NOT recommend to OC. If you do, you're on your own and strongly advise to keep a close watch at the reported jobs in the Result Status page. 1-2 'Invalid' is a hint, many 'Invalid' becomes obvious, 'Error' tells you to take a long rain check from this practice.

I second vaio's comments.

If you overclock to the point of causing system instability, you will almost certainly notice this long before you ding any work units. Either your system won't start at all or it will start to reboot for no reason or your thermal alarm will go off, or . . . . you get the idea.

I'm not an expert on hyperthreading, but if you have a dual core, I would disable it completely. My understanding of HT is that it allows you to emulate multi-cores on a single core. But if you've already got a multicore, I don't see what you have to gain. Plus, I've read that it causes a lot of memory "churning" that is just going to slow you down. Try disabling it and see if it affects your results.

Also, check out some of the overclocking forums. I don't have a favorite. I generally just put the word overclock and a few other search terms into google and see what pops up.

I wouldn't play with the memory timings unless you're sure that the sticks you've got will handle them. Unlike CPU manufacturers, it seems that memory makers don't give you a lot of wiggle room over and above what the chips are spec'ed out at. That seems to be true for newer DDR and DDR2 chips. I HAVE gotten away with ramping up the speed on older PC-100 and PC133 chips.

I have HT on an Intel machine which I've disabled. Just assumed that AMD had something similar although I never have noticed the option on my AMD X2 4200+ - guess that's why.

not new to oc'ing..
this rig in question is set up for just that...
my deal is that i dont see much more productivity for boinc oc or not...
can sure see the difference on benchies and gaming...

I guess the best place to start is to find out if the calculations being done by either WCG agent are mostly floating point, mostly integer, mostly logic (lots of branching, memory access). The CA's can probably answer that.

Once we know primarily what kinds of calcs are being done, then you can get a frame of reference from your benchmark tests. So if your machine kicks @ss on some benchmarks but not on the ones that measure the type of performance needed for WCG, that would explain the apparent disparity.

On the other hand, if your benching high on the types of calcs WCG needs, then maybe there is a problem that needs to be investigated.