 |
February 04, 2014, 04:37:59 AM |
|
Well, I am in. Sent 2,600 US + 150 for overnight shipping to asiabtc for the LA3M (3Mh/s total scrypt power, which is a controller, power, and set of 10 of the 300-350kh/s pucks). He is very responsive on PM fyi for those of you wanting to order. He lists price in US, but I sent him BTC using preev rate. Due Feb 15, and he has promised it will ship no later than Feb 20 (I asked for absolute outside latest date come hell or high water, and that is what he gave me).
If one of the testers could post some exact dimensions and ideas for mounting these, that would be great. I am going to fire up my 3D printer and see if I can make a tower something like a wine rack to put each unit in fan first; planning to make it modular so new ones can be snapped in... or any better idea for those who have them?
What do folks think about trying to optimize these for mining BTC vs. scrypt? Power requirements are so much greater for BTC, it would seem that it might often be the case that you would mine only scrypt with these, but maybe sometimes it would be better to only mine BTC, or both?? That there are 3 modes seems to present a great option to be able to optimize cgminer real time...
Starting with the github cgminer code for these devices, I would add a "smart" mining option for cgminer that will switch between the scrypt and 256SHA mode depending on maximum profit. I am thinking that the option would look something like:
--optimize-profit 256SHA-coin, scrypt-coin, averaging-function, time-window, recompute-time
Where:
256SHA-coin is the desired 256-SHA coin for profit comparison purposes (does not need to be coin actually mined). This would probably have to be a lookup table or service/deamon that provides information on current difficulty, average block solve time, block reward, and price for a given actual or fictitious coin... fictitious coins could be used for merged coin mining and/or multipool mining operations to provide average difficulty, solve time, etc. but the cgminer program would not be responsible for computing all that.
scrypt-coin is a reference to a lookup table or service or deamon that provides relevant info on the the desired scrypt based actual or fictitious coin
averaging-function is the desired method of computing profits from the relevant coin data (e.g., moving averages, weighted moving averages, stochastics, etc. I should be able to use open source libraries used in stock data analysis very easily here)
time-window is the time window to use when performing the averaging-function (12 hours, 4 hours, 5 days, etc.)
recompute-time is the time between recomputation of the averaging-function (e.g, every 240 seconds)
Basically, the modification to cgminer would compute the expected profits from mining in each of the three modes (scrypt only, sha256 only, and both) during the recompute period, subtract out power costs, and then select the operating mode that maximizes expected profits.
Thoughts? Better ideas?
|
