Quote from: lordmykl on October 28, 2014, 01:43:42 AM

So I did a little experiment with one of my Antminer S3+ units (One of the better performing ones that is able to hold any frequency setting for long durations without any 'Xs.' I essentially tested out each of the frequency settings and compared them to the wattage numbers used at that setting with 110v power, then tested again at 230v. The results of my testing are below:

In essence, what I found was somewhat obvious...that you get better Joules/Gh at 230V than you do at 110V, regardless of the frequency the miner is set to. I also found that the 'sweet spot' of energy used per GH/s is around the 200 MHz mark (200 MHz @ 230v and 206.25 MHz @ 110v.) I noted that in green, which incidently was the only reading that matched the 'published' efficiency numbers. A few notes about my testing:

1. I left each of them at the specific MHz frequency for exactly 3 hours, then recorded the Wattage and Average hashrate at that time. I was off by a few minutes sometimes, but suffice it to say I left them for at least three hours. I did note that this number is more accurate the longer you wait...however, I didn't have so much time to play with this, so there may be some variation that I couldn't account for or the numbers may change more significantly over time.

2. I used the same Antminer for every setting, the same PS (Corsair CX750 with all four PCI-E cables plugged in) and the same Kil-a-watt unit for each test. I was using the 8/11 firmware on the unit with the updated cgminer version (4.6.0-140908) posted by ckolivas.

3. The one major variable I couldn't control for was ambient air temperature, which was generally about 5C warmer in the 230V location than in the 110V location, plus during the day there were fluctuations. This likely would affect efficiency, though probably not by too great of a deal.

4. I used three variables which you don't see here...the price per kilowatt/hour ($0.10,) the current price of BTC in USD ($351.59,) and the amount of BTC/Gh/day based on difficulty (0.00001398 currently)

5. I included the power usage of the unit above when it is not hashing (noted as 0) just to note the overhead that the system takes up if it is on but not running.

Anyway, hope you can get some use out of this...at a minimum it taught me that it's not just a myth that 230V is more cost-efficient than 110V, but when it comes down to it, it's only more efficient by 1 or 2 cents a day...

wow nice and ty man I applied this some what to my last 2 S1 till i replace them sense i don't like the idea of penciling them . it actually dropped my power usage like 100 watts or a little more per S1 at these setting .

#option 'freq_value' '0b81' #300M
#option 'chip_freq' '300'
#option 'timeout' '45'

and my two S3 are set to

231 i get little oC in there with kind of nice saving on power .