As I mentioned in the s9 thread but certainly more on-topic here,
<snip>
IMHO it could well be that since Canaan Creative chose to be late to the 16/14nm node they (major stakeholders) decided to take the money and retire to a nice private island somewhere. As long as their IP does now belong to an "Angel" company with it's eye on the miner market (and we don't know that yet), then cool.

Even cooler would be if Avalon has low-node designs in the works but never put to silicon. With the new ownership it opens the door to them maybe using Global Foundries and go 14nm. Nice safe location (upstate NY) and as far as I know, GF is open to all customers.

Also add AMT/Bitmine.ch to the Failed list.

Assuming the Founders/major stake holders in Bitmaintech do not wish to cashout of the business (yet) then they should be staying as a privately held company.

As soon as a business goes Public they must then conform to the Public investors wishes which all too often results in hostile buyout offers by companies that only want to strip and selloff the IP from their target.

IMHO it could well be that since Canaan Creative chose to be late to the 16/14nm node they (major stakeholders) decided to take the money and retire to a nice private island somewhere. As long as their IP does now belong to an "Angel" company with it's eye on the miner market, then cool.

Even cooler would be if Avalon has low-node designs in the works but never put to silicon. With the new ownership it opens the door to them maybe using Global Foundries and go 14nm. Nice safe location (upstate NY) and as far as I know, GF is open to all customers.

If that is the case then it would explain the paucity of chips in the hands of other integrators. The container/data farm construction division of BitFury after all is a sister company of the chip division. Bet they got/are getting the lions share of chips from the limited number of wafers BF has received from TSMC so far.

No doubt the smaller integrators got some to play with and test out but....

About 16-17kw giving ~86.4THs is at work (free) the rest is at home @ 15cents/kwh (ouch!) but a s7 still makes about $5.86/day there after feeding. All depending on pool luck of course. On Kano the $5+ something seems to be the going long-term average. Historically I've always ran 5-6kw worth of miners at home at home since the late s1 days.

With all the talk on ROi, all I can say is:
Jan 2014, bought 2x 10GHs BFL miners from TigerDirect for $349 ea.. Quickly was hooked.

Feb 28 2014, Saw AMT's 512GHs miner for $4,000. Considering that Antminer s1's were around $2,300 at this time and from an unknown to me at-that-time Chinese company, I ordered the AMT A1 rig, eventually was sent a 1.2THs Dragon clone from AMT in Sept. 2014.

Late March 2014, getting tired of AMT's stall tactics but since still hooked, took a deep breath and ordered 2x Ant s1's  now @ $1,200ea through Amazon (Buyer protection and all that).

Mid to late April 2014, getting good returns, bought 2x more Ant s1's @ $750 each, again through Amazon.

Those are all the miners bought with $USD for a grand total of $8,598.00

Since then all expansion and upgrades of my farm has been solely financed using BTC earned from mining with the upgrades covering all Bitmain models except the s5+, most of the time getting at least 1 batch-1 and paying the price for it. Naturally all sales were/are now directly through Bitmain aside from what was apparently the very last new-unused-in-box s4+ that I got through Zoomhash.

I've bought probably close to $10k worth of consumer and IT goods from Tigerdirect using BTC -- which more than covers my total cash outlay when I got started -- before their consumer div went under last Dec. (and their CEO along with former Owner went to prison...)

Now running 110THs worth of mostly s7's (and have two b18 spares doh!) and even after ordering a s9 I still have >25BTC in the active kitty. So mining never makes ROI much less NET income how?

Sidehack, if you need a 135 chip s7 to play with, looks like I'll have a couple spares after my s9 arrives  (limited power budget) so I can always send you one.

btw: if you need more s5's to strip for making Compacs I have 3 more I can send ya. As for old farts.... Turning 60 here next mo right around the Halving.

+1 for thinking out-of-the-box but...
To answer directly - it is NOT a good idea to use an Ant hashboard as a reference for a different miner. Probably the only major function block they would have in common is the Vcore power regulators and even that would only serve as a gross reference to how string load balancing is handled.

The best you will get from forensics of an existing miner is knowledge of how that particular design addressed their particular design problems targets. Yes you will be able to tell how thick various copper layers need to be and how they chose to route signals along with seeing if they took care to shield the signal lines. All that is nice to know but not directly helpful with a new design.

It would be like using an Intel PC mobo as a reference to make a mobo that uses AMD cpu's. Yes the concept of what happens on the motherboards and the input/output from them is the same but the signals methodology and support circuits used with the chips are very very different.

More to the point is that so far no chip maker is releasing data sheets detailing pinouts and com protocols for their chips. That vital piece of information is the starting point for any miner design. There are folks here that know how to make miners (Sidehack for one) and can design all the support circuits the ASIC's use as well as do board layout. Then there is the micro-controller driver coding for everything between the ASIC and the outside world. That is not an off-the-shelf bit of software and is not part of CGminer or BFGminer. But again, there are people here who can do that.

While Bitmain does release data sheets for their chips they do not provide reference designs. Understandable since they actually make miners. But BitFury and BW -- they make chips, not miners so if they want folks to buy their chips (to make miners) why do they not, a) provide full data sheets and b) provide a simple 2-4 chip reference design so folks can see how it should work and then expand on the reference with their own ideas for more/longer string nodes? That is how every other semiconductor mfgr on the planet works...

Having a reference board (or not) and knowledge of hashboard design is not the main problem. Getting the data sheets and chips is.  

I think that Mr. Kashif is referring to is the PCB house taking x-rays of the pcb's and then generating layout files from them along with figuring out the schematics.

Problem is, that's fine and dandy if you want to clone/improve the s7 design but usless for using any other ASIC other than what the board was made for. Pinouts for signals and power are no doubt different as are the com protocols. Then there is the micro-controller driver issue which is pretty much unique to each ASIC chip family and the mc's they use..

Again, the only reusable parts are heatsinks, endplates/fans, and perhaps the RasPi/BB control boards.

As has already repeated ad-nauseum here and in other 16nm threads, 'tisn't Bitmain that limited the supply, it's TSMC. Pretty safe bet that given a choice Bitmain would much rather be pumping these things out the door as fast as possible.

I don't doubt that Bitmain is on an allocation schedule from TSMC so other 16nm boutique chips can be ran (maybe BitFury's? Been mighty quiet in that camp...) until all the delayed production is being covered and eventually is caught up. Until then it will be interesting to see how batch timing goes.

Ditto. Was afraid I missed getting one as with Summer upon us I want to take 3 of the s7's at home offline and replace with the s9. 'Bout the same TH/s but knocking 2.6kw off of the heat load my AC has to handle. Happy Happy, Joy Joy!

I have 2x batch-1 and 3x b2 s7's as well as many other up to batch-17  that have been running non stop since getting them all in a very warm area. As i wintertime ambient was 85F and now in summer is banging around 95-98F with zero issues.

That said, I *do* have 2 later batch s5's that do the same thing. From time to time would drop a a card and on soft reboot either the entire card was gone or would show a few missing chips in 2 rows. usually a hard boot would bring them back to life for weeks-months and at the most inconvenient time drop out again. Finally 1 always showed missing chips so it and the other were among the 1st replaced by s7's.

If you mean redesigned to reuse the main heatsinks/frame layout and possibly the controller boards, yes. It entails just making new hashboards to fit the heatsinks. This is along the lines for what Sidehack thinking of doing to begin with. The old RasPi boards from s1 and s3's might not be handle the data throughput but looks like a BeagleBone as used on the s5/7/9 can.

Which comes back to the problem of getting the chips themselves....

Doubt they ever will.

With the A1 neither Bitmine.ch nor any of the more ah, successful, integrators using the A1 ever released a BOM for the rigs to the Public. All InnoSilicon has ever released is their 2-chip ref design. Sad part is, I have said documents for AMT/Bitmine.ch's original attempts at the CoinCraft rigs and as much later backed by Inno themselves, Bitmine.ch simply took the 2-chip layout and expanded the nodes per-buck regulator with zero regard to what that does to the PDN and even less to the what should have been very obvious thermal points eg needing real full sized heatsinks on top of the chips. .

Since AMT was BItmine's single Authorized dealer of the A1 and rigs based on it (AMT's verbatium clones of all boards used in the CoinCraft Desk) it was the blind leading the even blinder (AMT). Starting June 30 2014 under NDA I did forensics on them and made clear to AMT/Bitmine.ch where they and the PCB house AMT was using blew it. Had they let me look at things a couple mo earlier who knows - they might have recovered from the A1 kerfuffle. Main problem was needing more attention to detail on feeding stable power to the ASIC's as well severe underestimating of the thermals. Both easily solved as proven with the Dragons layouts. That said I still must professionally consider said BOM's including schematics and PCB layers as not for Publication since AMT/Bitmine.ch (well, what is left og them) still hold Copyright on that information. Technically I suppose I should delete it but -- it's safely archived and still gives a good lesson on crossing all the t's and i's. Especially when it comes to brand new node size (28nm) chip releases

I'd say someone find a Wayback archived copy of the site and any mention of support time.

Joshua Zipkin former owner of AMT for months spanning August 2013 through their closure in Sept. 2014 touted on their website pushing their 'American Made' licensed copies of Bitmine.ch's doomed to failure A1 chip boards a 'Lifetime FREE upgrade as miner tech evolves'. We know how that went... AFAIK Josh is still stalling the Court on getting new representation since his original lawyers were relieved of Burden last year due to Josh not paying them...  Think the latest stall makes 3 or 4. Hopefully the Court will eventually tire of this and issue a Summary Judgment?

edit: I found a Wayback to AMT's 1.2THs rigs ($5,599. ea) http://web.archive.org/web/20140321031122/http://advancedminers.com/bitcoin-mining-hardware/1-2ths-bitcoin-miner/ That particular page from March 2014 claims 'Lifetime Warranty on the hash boards.  

DISCLAIMER: For whatever reason, in Sept of 2014 I was one of the damn Lucky Few that actually got A1 based rigs from Joshua Zipkin. I actually went 'cheap' and had ordered just a 512GHs rig (advertised as in-production vs the 1.2THs being pre-order only) but as all the other lucky few received, I got a 1.2THs Dragon clone thanks to contacts he made during his trip to Shenzhen in the last days (well, 3mo actually) of AMT. So I ended up with 2x the purchased hashrate with only a 7mo delay vs the others (the 1.2THs rig customers) only getting as-advertised and most (who actually got a working one) having waited since late 2013. Thank you BTC gods fer that one! Mine ran perfectly 24x7x365 up to last month when I took it off line to use its 1.3-4 kw power budget for a s7 elsewhere.

Or thanks to quantum fluctuation the gates decide to often be somewhere else when observed eg read or written to

"Tapeout only means that they have functioning chips -- which back then would mean strictly very low yield per-wafer engineering samples to play with. Maybe they work per expected or (mostly likely) not. Considering TSMC first announced 16nm starting full tilt production for AMD around July last year (ref link?) it is possible KnC had earlier finagled getting a few test wafers slipped in but no way any large - and decent yield - amount of wafers that were production-grade.

I won't fault them on being, well, call it "overly-optimistic" about when 16/14nm would open up to customers other than Tier-1 simply because said T1 customers had for the last few years been transfixed by the light at the end of the 16/14nm tunnel as well. Problem is, it kept slipping further and further back... That said, KnC should have been closely following what was happening at the Foundries back then and tempered their enthusiasm with experience from past forays into new sizes.

Edit: Given the history of (finally) hitting 14/16nm nodes with decent yields only around Nov. of last year it sounds like that since TSMC at least ran samples for KnC to evaluate and given the general simplicity that 2112 says sha mining ASIC's have vs huge complexity that full up SOC or CPU's have they would have made nice tests for any foundry to use as process response benchmarks so that would be an incentive to run a batch or few of 'test' wafers with (again) someone else footing most of the foundry's actual process cost...

Wanna bet that what the foundry learned from running KnC's chips was applied to Bitmains and BitFury's? (and where applicable the Tier-1 folks as well) A foundry may not be able to suggest layouts 'other folks' used but they *can* apply information regarding how the various process steps are performed and the general results from their 'secret sauce' . That information is IP for the foundry to internally use as as desired

Of course back in Feb last year making the actual state of process readiness to run boutique chips known -- *that* would have heavily dampened any VC and pre-order/other funding...

To have had enough 16nm chips to fill a data center a year ago would have meant getting chips  in quantity say around Jan 2015. That is impossible. At that point in time all research and production at ALL the Fabs just starting to be capable of producing 16/14nm chips (not counting IBM & Intel who only make chips for internal use) was exclusively for AMD, Apple, Cisco, and Samsung (and Samsung was 14nm, internal use only back then).

Unless KnC was willing to maybe pony up several 100 million dollars to join the que for low-node chips back then all they would have gotten is nice smiling nods in discussions with the Foundries and that's it.

+1 ^
As far as anyone knows they have not taken a single Satoshi from the Public. All costs from the earthquake delay at TSMC has been born by BitFury, it's investors, and whoever they have lined up as integrators to actually build miners using their chip.

Now if any integrators counting on using the chips have taken pre-orders for miners from the general public -- dunna know. Certainly cannot blame BitFury if any did -- that would be between said miner mfgr's  and their customers.

Since Summer is upon us, as soon as I get notice that my last 2x s7b18's I ordered last night are shipped (great timing there ay what?) I will be ordering an s9.

Right now I have 4x s7's running at home being 5.2kw worth of heat load being added to what my AC has to handle so is a no brainer there: take 3 of the s7's off line and replace with the 1 s9. Introductory pricing be damned the combined miner and AC power savings make it a winner for me. Best part is that the ~1,350w load fits perfectly with the 1,800w (2.2kVA) Cyberpower always on-line UPS at home so no worries about dropping >14THs from low power line or blackouts (also have a permanent home generator to handle extended outages) . For those wondering, got the GE 11kw whole-house genset ~ 3 years ago after my area of Detroit had a 2-week long blackout -- during a big summertime storm the local substation literally blew which in-turn damaged the feeders to it ... Of course since then the longest outage has been maybe 1/2hr at worst  

At least for now I can replace the last s5's I have at work with a couple of the displaced s7's and will have spares...

I would think it will be about BitFury's chip.
Considering Bitmain now has (a limited supply) of their 16nm chips from TSMC it should be safe to assume that BitFury and through them their integrators are now getting their chips as well.

Since Bitmain has now set a benchmark with their s9 it will be interesting to see what BitFury's integrators come up with to match that