So finally got my Sapphire 5850 yesterday, and have been lucky enough to get it pushing 403.35MH/s at 1020 core / 300 memory clocks, using Phoenix / phatk on Windows XP, SDK 2.1 and 11.5 Catalyst. Had to use 1.2 volts to get there, but my core temp is ~70c and my VRM temp just under 80c.

Using GPUTool (old school Furmark, essentially) to test for stability, this is no where near stable for actual 3D graphics / gaming. However I only get rejected blocks when they are submitted around a block switch-over, and only had a driver crash when pushed up to 1030 core - otherwise I'm hashing away happily.

Is stability above the maximum gaming overclock because only part of the graphics core is used while mining? And would the shares generated by Phoenix show as "rejected" if they were invalid due to an unstable overclock?

This is just a random thought that I might share

I'm in slush's pool, using DaCoinMinster's Chrome extension. One of the additional statistics displayed is a calculated MH/s. I assume that this is calculated based on something like: (cluster GH per sec / total shares) * shares I produced. I believe that some other pools also display this information, although I haven't dug too far into many of them.

I push a (stonking) 35MH/s total at the moment (soon to be moderately and prudently improved), and there is pretty large variance in the calculated MH/s displayed for each round - sometimes I'm seeing 100.64MH/s, others 7.30MH/s and anywhere in between. This variance sometimes evens out over longer rounds (which you'd expect) but often it makes no difference whether it is a 2.5 hour round or an 8 minute round.

Do you all think that this could be perceived as a measure of "luck"? As in, for rounds in which I'm fortunate my MH/s is over-represented, and on bung rounds it is lower?

Or am I pushing so few MH/s into the pool that the overall measure is probably going to be completely out of wack anyway?

Keen to hear any insight or general musings