So I tried to do something that I'd not suggest anyone else to do
Of course try it if you really know what you are doing and have a 5kRPM fan, high quality power and are using linux
The aim was to see how fast I can get the R909 going - and I've got a record so far, we'll see who can beat it
My production R909 has one chip not as good as the other 5, so someone with 6 perfect chips should be able to beat it.
Anyway, I put in a 5kRPM fan, set the blue screw close to maximum, set 5 chips to ... 725MHz ... and the dodgy chip to 500MHz.
Results after hours of running (that I posted in discord earlier):
(5s):5.067T (1m):5.046T (5m):5.105T (15m):5.098T (avg):5.024Th/s
A:36261324 R:23056 HW:8786576 WU:70192.7/m
Connected to diff 2.88K with stratum as user
Block: 5fd33018... Diff:34.1T Started: [19:45:10.943] Best share: 84.9M
--------------------------------------------------------------------------------
0: GSF 10008346: BM1397:01+ 450.00MHz T:450 P:450 (28:14) | 100% WU:100% | 307.9G / 292.5Gh/s WU: 4086.1/m
1: GSF 10008000: BM1397:06+ 500.00MHz T:500 P:500 (4:2) | 100% WU: 98% | 2.265T / 1.970Th/s WU:27525.2/m
2: GSF 10055348: BM1397:01+ 450.00MHz T:450 P:450 (28:14) | 100% WU:100% | 296.0G / 286.5Gh/s WU: 4001.8/m
3: GSF 10070004: BM1397:06+ 687.50MHz T:500 P:665 (3:2) | 90.4% WU: 89% | 2.433T / 2.475Th/s WU:34579.3/m
Massive >20% hardware errors, but "3: GSF 10070004: BM1397:06+" powering along at 2.475Th/s
(and of course the other number there, total of >5TH for 2xR909 and 2xCompacF)
Mine on!
Edit: I should add, when you change frequencies or it has finished ramping up the frequencies, reset the averages.
I've added the details to
http://kano.is/gekko.php#perf