They haven't put out a lot of hard info on the A3 so far - but they aren't bothering to look for investors for it, and they HAVE said they plan to release the A3 and A4 at the same time.

 I also suspect the A3 is going to be a lot better than 0.2w/GH - based on the efficiency gains of the A4 vs. the A2 and the claimed efficiency of the already-existing full-custom 14/16nm designs, I'm betting on under 0.1 and more likely 0.06 ballpark.


 They probably don't figure there's any significant risk on the A3 of not being able to sell enough of them to make a profit, given how much bigger the SHA256 market is.


 They announced tape-out on the A3 in the SAME "leak" they mentioned it for the A4 back around July or so.


 Innosilicon has been a miner manufacturer before - they built the A2 in 3 versions as a miner, the 88MH and 110MH units and that 60ish MH Farmboy thing.


 I'd suspect that they'll have a fairly high MOQ on the chips for a while, but will drop it down eventually when they're not competing with themselves as much by selling smaller quantities of chips.

The Compaq is probably the only miner on a USB stick with a prayer of actually being profitable for the medium term, but the probability of ANY USB-based SHA256 miner actually making enough back to pay for itself is pretty much zero at this point, unless you have FREE electric.

 ALL of them cost quite a bit more per GH (and most are still best measured as cost per MH!!!!) than same-generation larger miners using the same chips, which makes profitability on them quite harder to achieve.



 No, I don't get the "for fun" aspect some folks talk about. I can think of a LOT of things that are more fun than waiting on some low-performance miner to actually do something noticeable.

120v vs 240v are pretty much equally dangerous.
 Both are well over the threshold to electrocute you, both have plenty of amperage, both are AC so they tend to "vibrate you" rather than having a 50/50 chance to blow you away like DC has (the other part of DC is the possibility to lock you TO the line).


 Transformers tend to be VERY high efficiency - 98% or BETTER is quite common if you're not throwing wierd waveforms at them.

 The input cord to these things look VERY VERY iffy for the current they're rated to handle - look like *maybe* 10 amp cords when they should be AT LEAST 25 amp rated for the wattage these devices alegedly handle.

 Cost is also fairly high - it's cheaper to wire up a 220 circuit, breaker and all, even adding in retail pricing on a 6-15 or 6-20 socket/plug pair.

 They also seem very very small for a supposed 2500 watt solution - I suspect that's an INTERMITTANT rating, not a continuous one like a miner NEEDS.


 I'd be very cautious trying to run a miner via one of these things, monitor the temperatures on them like a hawk on both the device itself AND the cord.

google "innnosilicon" and either "A3" or "A4"

 Quite a few links that will find you, though a lot are reporting of the same few press releases.

 Tapeout is normally a step MONTHS in advance of producing actual sellable/manufacture with chips.
 Sometimes, if the design wasn't done right, tape-out is just "oops, we designed these but the design don't WORK back to the drawing board to figure out how to FIX what's not working right", though that's not common with modern design tools. More commonly, the design works but doesn't meet the intended specs at which point you have to try to figure out how to fix the design to meet target specs, or you decide to live with the specs achieved if they're close enough to target.

 Even in cases where the tape-out prototype chips meet the target specs, actually ramping up to production is still usually a "months to go" thing.


 It's not hard to picture Bitfury having working chips NOW though, all of their announcements have been a few months ahead of the corresponding Innosilicon ones, and I'm expecting full production out of Innosilicon 1Q next year timeframe.

The damned Sleeve bearing crap I've had to deal with wasn't "my pick" - they're FAR too common on video cards, and in power supplies, and I've had to replace WAY too many of the MANY DIFFERENT BRANDS of that sort of crap over the decades.

 The fact that they NEED "cleaned" so damned often should be a bad sign in and of itself - I don't remember the last time I had to "clean" ANY ball-bearing design fan out, they just keep running and running and running and they don't DIE AND TAKE VALUABLE COMPUTER GEAR WITH THEM WHEN IT OVERHEATS BECAUSE THEY DIED VERY YOUNG.

 YOU are the one that said "suddenly" - I realise quite well that the gear will usually survive a short while after the fan dies and lets the gear start overheating.

 You also ASSUME that I only have 1 or 2 fans running - I can just look behind me and see at least 20, ONE fan going out is not going to change the noise level in my computer/miner gear room noticeable at all.



 If a sleeve bearing fan actually LASTED a year reliably, it wouldn't be so bad - but I've had WAY too many of the cheap pieces of junk die on me in a few MONTHS of use, sometimes *A* month of use. They are NOT reliable. They do NOT last. They are shit by design. There is a REASON they are cheap.

 More like they realise that by the time the Batch 9 units ship, they'll have serious competition AND they're going to be impossible to RoI at the older overcharge price even with VERY VERY cheap electric.

 It does appear that the "28nm full custom" generation is going to have a rather short profitable lifetime - the units of that generation came out WAY too close to the halfing, and it looks like the first 14/16nm "available to the public" gear should be out before the halfing as well.

 You can't buy an SF100 either, they've had none available for MONTHS now.

 I'm not sure if Innosilicon is still selling their "A2 Farm Boy" units, if they are that's the ONLY new Scrypt gear for sale right now as I'm pretty sure they mentioned selling out the last of the new A2 Terminator units they had.

14/16nm full custom will be the last major efficiency improvement - for a few years.

 IBM has already announced they're researching 8nm (IIRC) for the next generation process.

I've had too many Deltas last 10+ years without NEEDING "cleaning".
Can't say if they'll last longer than that, I've still never had a Delta die and the oldest of them I've got would be around 12 years old.

 I've had WAY too many shit "sleeve bearing" fans that didn't even make it to ONE year before they jammed up and died - and caused the gear they were on to overheat and die as well.

 You're ASSUMING I have only used Deltas on miner gear - I was using them years before Bitcoin EXISTED.
 Does "Alpha 8045" mean anything to you?
 No, I didn't use the 80cfm screamers, but I do still have some of the one-step down 68CFM Deltas, currently in use in 2u rack-mount cases.
 

I do miss the days of Rotron metal-frame fans - I've got a couple of those that I bought used back in the 1980s THAT ARE STILL RUNNING.
 THOSE were good, reliable fans.


 Also, not all Deltas are super-noisy, just the ones that move massive amounts of air - as are ANY of the few 120mm fan models in the 200CFM+ range from ANY manufacturer.

 Not part of the ATX spec, and would add significant cost to them.
 They would also be less efficient, as they can't be designed to maximise efficiency at a specific operating voltage.



 The whole point to a string design is to not NEED a voltage regulator on the board at all, dropping the cost substantially. The tradeoff is flexability to tailor the voltage for max output or max efficiency.


 I don't see Bitfury ever being listed on the NYSE, as they're too "secretative", they're too SMALL, and they are not a US company anyway (depository shares at MOST, similar to what some Canadian companies like Pengrowth do, but more likely on NASDAQ than NYSE).

 I vastly prefer fans that don't NEED cleaning, as fans that get dirty and stop running tend to KILL electronic gear before you get the chance to notice the fan died.
 Ball bearing fans are a TON better at "don't need cleaning" than ANY sleeve-bearing or varient design, as they're designed to NOT NEED LUBRICANT JUST TO RUN AT ALL.


 I'll take a "40000 hour MBTF" Delta over any "160000 MBTF" piece of junk sleeve/hydro/rifle/twister type design, as the Delta might actually last that 40000 hours where the sleeve/hydro/rifle/twister/etc. type design will be lucky to make 10000 before it locks up the first time, and will probably have completely died before 20000.

Any miner can be "recycled" as a space heater. I suspect quite a few older ones are already used that way.

 Granted the USB stick miners aren't very good at it.....

 Twister bearing, aka "rifle" "hydro" etc is nothing more than a fancy sleeve bearing type.
 I recommend AGAINST any version of a sleeve bearing on a fan.


 Manufacturer MBF ratings from most fan makers tend to be wildly optimistic, assuming lab-level CLEAN AND COOL conditions among other things - ball bearing fans tend to be the ONLY ones that actually achieve their rated MBF in REAL WORLD usage, and even some of THOSE are rated somewhat optimistically.

 That's one of the reasons "push" is almost always better than "pull" for cooling - the "pull" configuration runs the fan a lot hotter, making it a lot more likely to overheat and fail faster.

 ROFLMAO!!!!!

 Not this year.
 Perhaps middle of NEXT year, with a drop at the halfing for a short while.

 You've OBVIOUSLY not been paying attention to "new gear in production" or "new gear available SOON" announcements.

Y'all are forgetting something.

 I'd BET that their "for domestic use" means those LIGHTBULBS of theirs, NOT "home miners".

 More fury and noise meaning NOTHING NEW AFAIK (other than the "now in production" part of the announcement).



 They're not the only folks claiming 0.1w/GH or less on their hardware, BTW.
 KnC made that claim about their "Solar" chips and stated "in production" months ago, yet they seem to be having major yield issues or a cash shortage since they've not ramped up their hashrate much since that "in production" claim.


 They're also not the only claim of "full custom" at the 14/16nm level.



 It would appear that the "28nm full custom" generation is going to have a VERY SHORT viable lifespan - I'd say less than a year for ANY of it unless you have VERY VERY VERY cheap or free electric.

See Innosilicon, they seem to be the only company to date (of the majors) that has done ANY "build new units to be able to reuse old hardware" at all (Farmboy/Zeus stuff).

 Despite Bitmain claims, they have never actually done so, and I don't think anyone else has ever even talked about "upgrading".

Don't bother, the A3 / A4 are going to blow this thing out of the water *IF* SFARDS ever bothers trying to sell any more.

 The whole "dual miner" concept has always been a very bad one - the SHA256 side gets outdated VERY quickly leaving you with an overpriced, not all that great performance Scrypt miner.

 They MIGHT have a faint prayer of making a competative miner if they move to 14/16nm in the next year or two - THAT miner would have a fairly long time at "current tech" and might stay competative long enough to make up for the SERIOUS HANDICAP of pushing 2 different algorythms on one chip.

 PM sent, since you still haven't replied to the emails I've sent.

 It's odd, since I mentioned the wrong unit shipped issue in the SAME email and THAT part you replied to quite promptly.

Titans tend to die too much, lose die, etc - the reliability of the things is junk.

 They ARE quite a bit more efficient than anything else CURRENTLY available.