SNIP
If you're averaging anywhere near 3.5Th/s you should be happy. Most Don't.
My Brothers' Neppies "top out" at about 3.45Th/s each.
We figured any downtime of the machines spent on cooling tweaks wouldn't pay for itself by overcoming what was lost during the downtime, there just aren't enough gains to justify it.
With that in mind, We don't even like re-booting for the same reasons. The way things are at start-up, it makes my asshole pucker every time hoping the PSU's provide enough kick to get all the asics going. But I'm pretty satisfied on the performance being what it is, they seem to chug right along without any hiccups once they get going. We have over 10 days nonstop "Uptime" since the air-con kicked the breaker on day 2, which wasn't any fault of the Neppies; and despite running in a HOT garage, they continue to perform without any need for "tweaking."
If you're averaging anywhere near 3.5Th/s you should be happy. Most Don't.
My Brothers' Neppies "top out" at about 3.45Th/s each.
We figured any downtime of the machines spent on cooling tweaks wouldn't pay for itself by overcoming what was lost during the downtime, there just aren't enough gains to justify it.
With that in mind, We don't even like re-booting for the same reasons. The way things are at start-up, it makes my asshole pucker every time hoping the PSU's provide enough kick to get all the asics going. But I'm pretty satisfied on the performance being what it is, they seem to chug right along without any hiccups once they get going. We have over 10 days nonstop "Uptime" since the air-con kicked the breaker on day 2, which wasn't any fault of the Neppies; and despite running in a HOT garage, they continue to perform without any need for "tweaking."
3.45 to 3.55 is ~ 2%
2% of 1 day is 28.8 minutes.
If 28.8 minutes of downtime 'gross' 2% you come out even 'net' in a day or two.
(assume days are within same diff period)
From the break even point forward is free of downtime 2% gain.
This ALSO assumes you let the other 4 ASIC idle the whole time you mod.
So I guess it depends on if you need parts also and cost of parts.
Agree, no NEED for the extra 2%, just extend date of (possible?) ROI with 2% less GH/s in calcs.
PLEASE NOTE
You must first subtract percentage of non functioning core from 3.6 to know what is available.
Each 72 sleeping core is 1% off the top.
I have 16 sleepers so <0.25% or ~3.59 possible.
I have <0.7% HW so ~3.56 is expected.
I am within an "RCH" of expected.
From where I am NOW a few minutes of downtime may matter.
From where I WAS my downtime was worth it.
To ever go beyond 3.6 will require a larger percentage of the power budget.
Hot VRM are less efficient and overbuget already.
They are spending 'lifetime' for the interest on loan from 'current bank'.
You may have 10 days uptime on the BBB, I doubt you have 864,000 seconds uptime in cgminer.
You can easily check with
'api-cgminer devs'
Disclaimer:
Decimal points in examples do not indicate accuracy.
They just ballpark approximations.
I have been known to slip a bit here and there but usually round in direction of whatever is worst case.
YMMV
