Right, well I want to O'C my two 4870s to improve performance in crysis a little bit (not QUITE playable on max settings currently) and just to get better performance basically. However they are running stock coolers, so I was wondering just how far its safe to push them (without raising voltages if possible) and what overclocking apps to use as I am an overclocking noob generally speaking
(Oh and a completely unrelated question is that I'm running a Q6600@3ghz and I was wondering if it was bottlenecking my 4870's, does anyone know if I should O'C that a bit more as well to get the best out of em?)
It's almost a certainty that using ATI-OVERDRIVE in CCC will set the clocks at 790 core, 1190 mem deeming them stable. That's despite the nasty 80-88c idle temps thanks to their brilliant restriction of the fans speed to a max of 30%. But still it's safe to max them to the limit of 800/1100 (4400mhz effective mem) since they have always found the highest stable clock and backed down 10mhz. Those clocks result in a 6.25% core increase and 22% memory increase, depending on the game a 10-15% performance boost on average, though usually closer to 10%. My 4870 was fine running those clocks for very extended periods for 2 months without the benifit of fan control but i got a 4870x2 as a gift...and it just don't appeal so much now =D.
when i came across the fan speed over-rides i already had my x2 card, but tested it on the single first just incase it was going to blow up. With ambient room temp at 78F, using rivatuner to bump the fan speed up to 50% knocked the idle temp of the gpu from 87C, to 52C in under 2 minutes, load temps stayed under 70C at 100% load using overclocked settings
Taking the fan to 75%, while possibly too loud for some, resulted in idle temps of 30C, and load temps of about 42C, even after an hour at 100% load it didn't go above 45C. In all cases the memory chip temps were uniformly 5C higher than the core.
After that i applied the same approach to fixing the 4870x2 heat problem, though suprisingly it wasn't as bad as the single version though another problem is presented. Since there is a single HSF for both gpu's one core is always hotter since it's getting the heat of the other core blown across it, so there is anywhere between 5C and 12C difference. Even with that issue...i currently have the fan set to 70%, ambient temp in my room is 79F, and GPU0 is at 35C, GPU1 29.5C, even with heavy gaming load i've yet to see either core break 50C, and if i bump the fan up to 80% they both stay under 45C.