Quote from: cedivad on September 18, 2014, 10:02:30 PM

Quote from: Minor Miner on September 18, 2014, 09:50:53 PM

Quote from: RoadStress on September 18, 2014, 08:42:23 PM

Quote from: cedivad on September 18, 2014, 07:11:20 PM

http://cointerra.com/product/aire-miner-bitcoin-miner/
0.3J/GH.

Do you think anyone will buy from Cointerra after their fail? I don't think so...

Look like a Spondoolies design to me.

The Spondoolies design is simply smart, imho; a huge surface to dissipiate heat from, air cooled, it's easier to bring power to the chips, etc. It's just the same design as avalon/bitmain, just more compact and more datacenter oriented. I guess that the different chip same wafer design HF should have been working of should have been similar? (it's all again in the examinations).

Spondoolies does good designs, but theyre not the only people that do impressive and simple designs.

to me the most impressive board ive seen in a long time is the bitfury board thats in the 3500BF box. http://bitfury.org/products

its an extremely elegant and simple system. The boards are huge, but they contain practically nothing except bitfury chips and are sparsely populated, allowing huge swathes of exposed pcb to be the cooling system (no heatsinks required!).

i think ct were influenced by several of their contemporaries for their next generation miner.

In their first system the goal was clearly to push as much performance - PER ASIC - as they could (regardless of the power that it used, since bitcoin price was rising rapidly and performance per box ($/GH) was a higher priority than performance per watt. At the time of launch ct's was the lowest cost box around.

Like their contemporaries of the time (KnC & HashFast) they aimed for maxing out the asic... a big chip made up of four dies... and they targetted 500 GH per chip - at the time it mightve been one of the fastest asics.. but alas they missed by 20% (but being fair, which asic companies DIDNT miss their targets.. by a little or a lot...? (AsicMiner, Spondoolies, BitFury, BFL.. many missed by far more than 20% etc etc) i cant think of any that didnt miss (except hf, but they failed in other ways).

so.. ct initially delivered 400 GH per chip for the first board - but the asic was hampered by the board not giving it enough power to run flat out and the 2nd board (but same asic) four months later finally just about hit their orig goal of 500 GH (but too late to be sold at retail). Unfortunately, with practically 400 watts per chip ct needed an exotic liquid cooling system to keep it cool. The more exotic, the more moving parts, the more complexity... the more chance there were for things to be unreliable. practically any problems people have witnessed with the ct units are related to the liquid cooling (pump failure, contacts with cooling head etc). it would be the same with any other similarly complicated system that contained moving parts (and liquid) and seems to be pretty much the nature of the beast. They also had an issue that the system was designed for a fixed four chip solution, and that precluded them from just adding another chip when they wanted to recover the missing 20%. on a more flexible system design, they wouldve simply added another asic and the market wouldnt have cared whether there were four or five asics as long as it delivered the full 2 TH... but that wasnt to be.

The new cointerra design is in some ways the complete opposite. Instead of eeking out the highest performance they can from each asic... they've re-architected it to be an extremely low power design with a scalable system design.. so instead of a low number of very hot chips, they can have an infinite number of very cool chips and just dial in as many chips as they want on a board to achieve a desired hashrate ( to be fair, thats pretty much what others like bitfury, avalon, spondoolies have done in the past). but the key is that by being first on 16nm, they are ahead of others in terms of power consumption. The two metrics that matter most... $/GH and more importantly, J/GH (running costs). They are also in the right place on Moore's Law to continue manufacturing in 2015 and beyond. There's no other asics out there that are sub 0.3 J/GH right now and it might even be possible to underclock to reach lower than 0.2 J/GH if the need ever arose to go that low.

At present, the mining network is about 200-250 PH... and most of that is at 1W/GH or greater. Those systems (as everyone knows) are getting a bit long in the tooth and all of them will need to be replaced as they consume too much power for the next year of mining. i doubt any of those systems will be running beyond year end and most likely they will be switched off in october/november because no one (who cares about profitability) can afford to run them if theyre consuming more cost of power than they generate in bitcoins (unless there's a steep rise in the price of bitcoin in the near future)

The new designs from the likes of KnC's and Spondoolies will allow those guys a few extra months of profitable mining... but the power really needs to go down significantly from those 0.5-0.6 J/GH's asics to be able to mine for much of 2015.

The key now is power consumption... and the only way to mine profitably is to not only be able to buy the miners for the right cost in $/GH... but more importantly, for them to last long enough on the mining network to be sustainably profitable thus the power consumption (J/GH aka W/GH/s) needs to be low enough to justify buying any new miner.

Its a tough decision to think about buying a 0.6 or 0.7 J/GH miner right now.... after all, the running costs just dont stack up... how long will it last? a month or two before its unprofitable? Even a 0.5 J/GH one wont last much into 2015.

As mentioned, the two key metrics... $/GH and J/GH are all that matters. Its pretty easy to downvolt any existing chip to the point that it runs super low power consumption, but its cost of silicon is too high to make that work. It has to be Low $/GH at the same time as Low J/GH, or there's no longevity in the solution. Expect the lowest price GH's to be ones that are high power consumption and they will cost more to run than they do to buy.

my prediction for 2015 is that this will be the year of the sub 0.3 J/GH miners... Mid 2015 its likely there will be a bunch of options in the sub 0.2 J/GH and late 2015 will probably be closing in on 0.1 J/GH. And since the bulk of the network will be made up of miners in this efficiency range, the older and less efficient systems wont be left switched on. Just like the old FPGAs and 65nm+ ASICs last year.