Each overclock is different for each card. What you should do is open up precision and increase the core clock by 10-15 Mhz and stress test using evga OCCT for at least half an hour or more. If you don't get and artifacts then you're good to go. Keep doing this till you get some artifacts and bring it down to the last known safe clock speed.

thanks, I will try increasing the core clock little by little to see what i can get. Im not running games that are extremely demanding. Just SC2 for now and maybe Crysis. Just want a little performance boost :)

I got stable results at 815/1630/2100.  I got no artifacts at this setting.  I also wanted to keep stock voltage because as a general rule of thumb, I don't feel I know enough about it to mess with it.  Here is little picture of what was happening during the test.  I hit a max GPU temp of 68°.  I kept the fan on Auto here as well.  I noticed that with this setting, with benchmarks from OCCT I had a 10 FPS increase over stock settings.  I'm not sure whether this is significant for actual gameplay however.  I am gonna keep the settings as is and try OC'ing my processor next, I just installed a new after market CPU cooler (CM Hyper 212 +) with fantastic results as far as temps are concerned.  Let me know how you do.
 

I am unsure to be honest with you.  I tried it both ways and I ended with the same results as far as stability was concerned.  I think with the heavy draw it puts a lot larger load on you GPU.  I know that when I had this mode active I was only going 44 FPS, then when I ran the next test, I was in the 388 range.  This is my best guess, perhaps someone else can explain more.  
 
Also, as a side not, I looked at what other people claimed to have OC'ed to and I started there.  I just payed close attention.  I literally would start a 10 minute test, and watch intently for artifacts, then I would stop the test as soon as artifacts started to show up.  I noticed there was a cascade effect that would happen and when one artifact showed up, more would come in short order and then numbers of them would multiply.  So, I started high but never let my driver actually crash.  I made a list of settings that lasted the ten minute mark, then I started at the fastest speed and ran a half hour test.  I just payed attention again and looked for artifacts.  Two showed up right after the ten minute mark on my highest, then I brought it down to the next setting.  Second on my list was stable beyond 30 minutes.  I checked this with OC Scanner and OCCT.   I also made sure to run it at my native screen resolution because that is where I do all my gaming.  
 
As another note, when I made my list, I went in intervals of 5 Mhz for the clock speeds.  I may be able to tweak it a bit higher, but I think I am perfectly happy here.  I literally picked up 10 FPS on Crysis with this setting.   I honestly didn't expect a turn around like that.  It was crazy.  I noticed that the memory clock could be advanced a lot greater than the core clock or the shader clock.  I found stable core and shader clock settings, then started moving the memory clock up.  I was dumbfounded when i was able to go a full 200 Mhz on this.  And all this on stock voltages too.

I am not 100% sure on what I'm gonna say next, so I would appreciate it if someone would inform me if I'm wrong.  
 
My logic tells me that stress testing the GPU for a whole day is excessive.  It is a stress test, so it is meant to use all the resources of the GPU for a set amount of time.  I get that 40 minutes isn't enough, I only used shorter times to get a general area, then longer times to find stability issues.  I stress tested using OC Scanner for 4 hours today while at work, I was artifact/error free.  The thing is this, it only produces errors when pushed, and since I will never push my card for longer than probably two hours at a shot before it gets a break, is it really necessary to run a stress test for such a large amount of time?
 
It's like a stress test on the human body.  They work you up to a level that coincides with a set of standards, in this case what a gamer or computer graphic artist would push the card to for a standard amount of time.  Then they push you slightly over that limit and check your heartbeat and blood pressure and oxygen uptake in the blood.  Here that coincides with the various clock speeds a card runs at and tests stuff like card temperature and graphic performance.  In the human stress test they look for things like irregular heart beat, heart rate return, and excessively high blood pressure.  Here that means calculation errors and artifacts.  
 
So, I guess I am saying that since you will never push your card for a strait 24 hour period (most likely) it seems excessive to me.  Go over whatever the average amount of time you would use your card for, but to multiply that time by 8 for my test seems wrong to me.  I mean, if you are using your card for something that runs 24/7, go for the long haul, but if your a gamer that plays games a clip for 2-3 hours, don't bother testing for a whole day.  
 
That is my take on it, feel free to make a liar out of me.  

Ah, I see.  Haha, we are not liars, it's true, and you totally fit my if/then instance.   I stand corrected good sir.  I find the folding idea very fascinating.  How long does a single folding process take actually take?