The schems, BOM and PCB files remain property of Bitmine.ch and I have to respct that part of the NDA I had with AMT.

What I can say is that Bitmine pretty much took the 2-chip dev board that was being used in the A1 dev thread here and multiplied it into a lot of 2-chp layouts on 1 big board. Also apparently never paid attention to what was being said there about thermals when the chips are pushed to published specs.

As to Vcore... Due to product availability AMT used a different model of the regulator chip that had a higher value internal ref divider (100k vs 10k). But - dinna change the external resistor values to match the difference. Also AMT/Bitmine never tried to set the internal flash ram in the regulator to startup at a safe voltage. Result: For several seconds the ASICS are hit with a Vcore voltage above their max ratings until a bootstrap program is run to set it to the correct voltage.  

EDI: Correction. Looking through my notes was the DPOT feeding the regulators set point that was substituted w/o correcting the fixed resistor values in the reference divider tied to it. End result was the same.

All in all I side with Inno's OP explanation of what went down in the Bitmine.ch/AMT A1 miners fiasco.

As an attempt to help out with the A1 mess from AMT/Bitmine, I did forensics on several of the A1 boards that Bitmine/AMT were using from late Feb thru around mid-May of 2014. They went through several revisions in that period. These boards and a testbed were provided courtesy of Joshua Zipkin owner of AMT. Since aside from a different name on them they are the same boards, my findings should apply.

Bottom line is that whoever designed those boards and ESPECIALLY the thermals layout should be shot (figuratively of course). As evidence, re the difference in the large topside heatsinks that the Dragons used: Bitmine/AMT 1st tried none at all then tiny chipsinks held on by thermal tape. 1 guess as to which hashboards had highly heat-discolored sinks falling off and chips literally going poof letting out their Majik Smoke.

Trying to get information from Bitmine's engineer on even the most basic things like how the ASICs are reset on power up generally resulted in, 'uh, what?' By digging deeper through the schems provided I found how they did it and their method of choice often/mostly worked but was certainly not 100% reliable. I won't even get into the Vcore regulator issues I found...

Thanks! I'l check that on Mon when I get back to work. Very good pic of the controller board at that link showing which connector is what chain.
Assuming it goes offline again that is. I reseated the PCIe plugs & dropped it to stock speed late Fri and it's still ticking along per Pool reporting. Und ja when chain 3 went out speed dropped proportionately. Before asked, the highest I OC to is 218.75mHz due to PSU limits.

When this started going south last Thursday, the 1st 2 times it died in both cases the red Fault light on front was lit and controller was unresponsive to ssh or web access. Odd thing is - in both cases the LCD display was lit and showing results from its last update: full speed and temps in the low 60c range. All fans sounded like running at normal speeds. Both those things suggest the PSU is/was working fine.

This 3rd incident was the first time it stayed alive enough to get an inkling of what was up.

Tried that. At least for the versions that used PCIe power connectors like this one, there are no LED's on the boards. There *is* a row of very tiny ones on the controller board that could serve the same functions but is no way these eyes can see what they are labeled.

As usual after it has gone down, after a power down and short rest it again came back up at full speed. The 2 other times it went down the controller was dead - no lan plug lights, no blinking CPU light, no red light.

To cover other questions, all fans are fine, free spinning, no axial play to the blades. Temps 55-59c (room is 90F). Psu light is green.

Looking at it from the front, in the s4, which board is which chain?
One of my original B2 s4's is acting up and finally failed in a manner that kept the controller/GUI accessible. Says that chain 3 is all ----------
Cheers!

Very very nice!
Only thing I would suggest changing is that for fire safety, use polycarbonate (Lexan) vs acrylic. Acrylic does not self-extinguish whereas polycarb does.
Only problem though is esthetics as polycarb gives a bit bubbly & brown edge when laser cut.

And the key word above is 'GLUE'. Do not just use sticky thermal tape. From past experience using it on my retired s1's I can tell ya after a month or so the heatsinks start sliding towards the center of the earth... Fortunately I caught that before things started shorting out!

Always use a thermal epoxy so they stay put. Just a small dollop per sink is all that is needed. Also conducts heat better than thermal tape.

It refers to how much resistance the fan can overcome while moving x-amount of air. In our case, the resistance comes from pushing the air through the narrow passageways of the heatsinks. Another term for it would be back-pressure.

The higher that resistance the harder the fans has to work to move air through it.

Fans for miners or any other device with high-density fins on the power devices have to over come a substantial back-pressure to be able to move the required airflow through the heatsinks.

By and large to produce high pressure a fan either A) has to be fairly thick to control how much air leaks between the fan blades and the housing. Is why most PC fans @ 25mm thick suck at producing any real pressure. Miner fans are generally >35mm thick. Or B) the other way to get the pressure is speed. Put the 2 together and yer Golden.

I wonder if Bitmain changed vendors for their fans in the last s4 production runs? I got my 1st 2 s4's shortly after B2 came out and so far fans in those are running perfect.

The s4 with short-lived fans I got a few months ago from Zoomhash - it was the last brand-new in box one they had We are talking around only 2.5k hours run time vs at least 3x that on my other s4's. Hmm.

Well I finally joined the dead S4 Fan(s) club. Had only 1 out the 4 reliably working so the 80c cut out got a good testing and passed with flying colors.

While waiting for Delta replacements to come in I did a little experimenting and found that the fans from my retired S1's WORK PERFECT! Back to running just over stock speed, hottest blade is 69-70C which is just fine. (room is 90F).
Happy happy joy joy! The only caveat is that ya have to use 4-pin fan extension cables but already have them on hand.

Also dug up the post on parts/how-to repair the fans. https://bitcointalk.org/index.php?topic=796839.msg10827831#msg10827831

As Bitmain states in the product description, we need to supply our own cord.
Gets them off the hook in supporting the gazillion different national plug types out there.

If you are willing to cut the male end off of a 14ga 110v IEC cord and replace it with a 6-15P that can be used. Barring that try Amazon/eBay or here http://www.stayonline.com/208v-straight-blade-cords.aspx

And by default since Bitmine's sole US representative was AMT ran by Giorgio's family friend Joshua Zipkin, some of the AMT customers as well though in our case continuing outright fraud is more the point.

Aside from now having confirmation about what I had heard here and there about what Innosilicon was doing, what happened pretty much mirrors what I was posting early on in the AMT threads: I indirectly work with the Semi industry and know the timelines for new chip development pretty well.

The timeline from Bitmine/AMT announcing the A1 and them almost immediately starting to take Investomer monies - sorry, 'pre-orders' - followed by the wildly over-optimistic delivery dates was simply not possible.

Sadly I dug up all this distressing information after I had placed an order with AMT on Feb. 26, 2014 for their 540GHs A1 rig On the bright side, I was one of the very Lucky ones and finally *did* get my miner from AMT on August 22, 2014 - and it is a 1.2THs version only because the folks in China (Iltek) that Josh Zipkin found would only make that speed.

Thankfully after going through almost 2 months of lies from Joshua Zipkin in the AMT threads about how AMT/Bitmine were progressing with the A1, I had the nerve to fall back 3 and punt. Got my 1st Ants from Bitmain and have done very very well by them ever since:)

On miners and UPS's... Miners do NOT like yer normal PC stand-by UPS's. Been there, tried that with several different makes. Even with a miners PSU at 1/2 load every one I tried caused the miner to hang when they switch or adjusted the voltage. Reason is the relay switch transfer time which for most UPS's is spec'd at <10ms. That blip is long enough to often cause a big enough dip in the 12v output to lockup the controller.

If you are going to uses a UPS is MUST be what is called a on-line UPS. Less efficient as they are always 'on' changing AC>DC>AC but there is zero transfer time. For several of my S4's and S2's I use http://www.cyberpowersystems.com/products/ups-systems/smart-app-ups/paragon-series/ol2200rtxl2u.html Pricey? Yes. However, are rock-solid.

btw, the s2's & 4's are the only ones on UPS's. When spiked (well, dipped really) they can be (are) the most susceptible to the sd card getting munged.. S3's and s5's just hang so the majority of the farm is not on UPS's. Waayyy too expensive to do that.

Heh heh.. I'm one of the lucky ones. I pay 15 cents kW-hr at home where I run 4.4Ths but at work my electric used for miners is 'free' so have 28Ths running there.  Our light-industrial rate there is 7.5 cents/kW-hr.
Nice being one of the owners

Or most folks already are well stocked with HP and/or IBM server supplies to take care of immediate future needs (like me)  

But ja, unless folks have several 'spare' kw to use at home and a way to take care of the heat, home mining could well be kaput. Now, that doesn't mean that the only customers left for Bitmain and any other mfgrs still around will be only Peta-farms, just that it's now past the hobbyist stage and in the hands of folks who can afford & locate larger setups.

Re: The Skype chats, They came from my chatroom originally setup to find out a solution to the A1 miner problem AMT/Bitmine.ch had. ref  https://bitcointalk.org/index.php?topic=830189.msg9277338#msg9277338 and https://bitcointalk.org/index.php?topic=830189.msg9277558#msg9277558

As for the hack, yes, no smoking gun. Just intent and motive though the article does (briefly) point out that other parties are just as er, miffed, at Bitcointalk. Only difference we have Joshua Zipkin himself on record actually making an offer for someone to do it. But - still not definitive proof.

Forget thee not Joshua Zipkin also has (now publicly) stated intent per the Skype chats. And before ya get into that aspect (the chats) ref to https://bitcointalk.org/index.php?topic=830189.msg9277338#msg9277338 and https://bitcointalk.org/index.php?topic=830189.msg9277558#msg9277558

However, I agree that there certainly is no smoking gun and in fact Jeff Owenby of BFL has some recent though again, so far let's say 'insetting, links to the hack as well.

As for the article, I tend to agree that CCN may not have been the best place to go to BUT the piece does start out saying that is is speculative article.

While the Skype convos do point a finger at Josh as a possible perp, that is all it says. No where does the article conclude that Joshua Zipkin had anything to do with it. It only brings up that he *and/or others* could well have been involved.

No.
The OP picked a bad name for the thread. However they were hacked via Social Engineering Hack. Once their intrusion was detected the Forum went off-line for cleanup.

Post corrected about the plug, thanks. Might as well add that one source for 240v C19/20 cords is http://www.stayonline.com/detail.aspx?id=38741

Bummer about the current signal. It isn't derived inside the PSU and just bought out via the high-density plug?