Zero HW errors is not valuable under normal conditions (unless HW errors are otherwise excessive). What is important is accepted hash rate by the pool.
The correct formula for hw errors is "hw / (hw + diffa + diffr)." Should be 1% or under normally. I have a few that get zero HW at normal clocks and a few that get 1% or so, but I don't notice a difference in delivered hash rate. In fact my most productive unit (by only a small margin though) is the one with the most hw errors. All are running at stock clocks.
The correct formula for hw errors is "hw / (hw + diffa + diffr)." Should be 1% or under normally. I have a few that get zero HW at normal clocks and a few that get 1% or so, but I don't notice a difference in delivered hash rate. In fact my most productive unit (by only a small margin though) is the one with the most hw errors. All are running at stock clocks.
with that in mind I have been playing around and of the four I have one seems to be flakey I will leave it at the low setting of 325 as it seems to do some strange things. when trying to over clock it.
the other three are now set at
1)375 freq this machine is giving me very low hw numbers very nice temps and a good hash of 190gh
2) 393 freq this machine gives higher hw numbers good temps and a hash of 198GH
3) 400 freq this machine gives highest hw numbers good temps and a hash of 201gh
the 4th machine give really high hw at 350 375 400 pops an x or 2 in the asics and in general is the lessor of the 4 machines but set at 325 it runs great.
these 4 settings give me 755gh at 1513 watts stock gives me 716gh at 1440 watts. so 73 extra watts = 39gh
I think what has thrown me off was that I have 1 machine that is lacking when compared to the others. I will keep these settings for a while
