Just took a look at the results from 12+ hours of logging yesterday after pulling chainminer v1.2. The board looks quite stable and error rates dropped significantly (~1% over the entire peroid). I'm looking at 36 GH/s sustained over long periods just dc fans and no heat sinks. I've manually clocked all chips at 54 and have not noted any benefit in using the auto feature in v1.2. Now someone push the board up to 40 
how are you making that sort of log? Im moving to manual tuning now, since most chips are stable at 54/55 with 1% errors and 36Ghash/37Ghash-nonce.
if it runs well for the next few hours i will consider moving the resistance from 1.178 to 1.165 (before it gave me issues, but that was prior to chainminer update)
https://bitcointalk.org/index.php?topic=287590.msg3108736#msg3108736
Send Isokivi a tip for saving you the headache.
ooh, not being a linux guy that still looks like a headache! (massive respect for the method/result though!). I can see 1.5hr averages at bitminter that work okay, but a longer-spanning solution would be nice. maybe the next sd update ill try to get the logger set up.
right now, im seeing a very steady 36.5 Ghash at the pool over the last 3 hours, with the bitfury page suggesting it might be closer to 37
EDIT: after about 16hrs, looks like 36.8Ghash is the average hashrate, with the last 3hrs above 37Ghash
EDIT 2: some new tweaking, and it looks like i am averaging 37.7Ghash
I have gotten 43-44GH/s frome one board. Only 1 of my 5 h-boards did this for 24+ hours. My assumption is the chips were from a good bin. What I also notice is there is a HUGE variance in the performance of these chips. My question is: why?