All this time it worked at normal (nominal) speed with no overclock because the web interface says the warranty will be broken if I enter advanced mode, so I did not play any stupid games. My money are earned with hard work. And now I am really pissed.
Underclock works on my machine. I played with their settings (650mV/800MHz/4000kbps) and it works perfectly. Check my previous posts to see how to modify the internal python program.
Since I designed one miner with those chips I can confirm that documentation is terrible. Errors and unknowns...
Regarding SPI, well, SPI in string works that way. Any string design work that way, one chip dies, whole chain i unusable. Without manual rerouting chain will remain dead. Of course is possible to design PCB such way that manual rerouting can be done via simple jumper desolder/solder but with this type of equipment (obsolete in half a year) is simple uneconomical...
Well.... SPI can be addressed using another SPI bus and a 16-bit SPI port-expander like MCP23S17. This means that 16 individual A1 chips can be addressed individually and the data can be passed on the common bus.
4 wires (SCK, SI, SO, RST) for a MCP23S17 addressing 16 A1-chips and another 4 wires (SCK1, SI-1, SO-1, RST-1) to collect the data from those 16 A1-chips.
One chip fails? we get a timeout reading its results and move on to the next one. And the internet is full of procedures from assembly to visual-C, from a simple PIC to x86 and ARM32 architectures, to get such a MCP23S17 up and running. Or the datasheet can be read and the brainzzzz can be squeezed and scorched to write a fully-pure-ASM (assembly language) high-speed software function.
I don't know what speed has this MCP23S17 (around 10MHz?) but a specialised port-expander could be built for this purpose, like they did their custom-made A1.
I design industrial-grade hardware devices for monitoring machines in extremely inflammable areas at petroleum drilling and extraction sites, with tons of arse- and brain-f* papers and documentation approved by high authorities in companies and governments after months of headaches, checks, re-checks, f* tests and arse-f* conclusions on hundreds of megabytes of PDFs and tons of f* papers with my f* signature wrote with my own hand on each of the f* sheets. These equipments must be perfect because people life depends of them, families earn money, they enjoy life, holidays, wife, children, job, careers. Failures and mistakes gets people in hospitals, morgues, tears, families destroyed, me in the f* prison and mein arse-f*ed for decades due to a wrong calculated f* resistor/capacitor/inductance value. I don't play childish electronic games, every night I'm cursing the f* day I decided to get this f* job so believe me I know what I am saying. Their "official" miners were manufactured to fail in the near future. Nobody can believe that smart-arse who invented the A1 chip can be responsable for such a failure in the coincraft miner PCB design and specifications. This failure was intended from the start as their long-term strategy. It may look like bitmine partially kept their promises and terms, but they still sold bitsh*tminers. This skunk-arse-stinks like FRAUD. They are really LUCKY these miners are not life-depending equipments, really-really-lucky they are. Very lucky.
By the way - right after the previous post I overclocked my coincraft desk to these custom parameters: clock-speed 1500MHz, core voltage 930mV, SPI speed 8000 kbps. With 4 working modules left I got 1.4TH/s - around 350....360GH/s per module (I am unable to repair the 5th one). I disassembled each PCB from the sink and I applied that old legendary soviet heat-transfer white paste. 4 pieces - 12 volts, 8 amps ventilators (coolers) from some old-school mainframe server (from the era of open reel-to-reel tape storage), two pushing air inside, two pulling the air out to circulate the atmosfere inside the unit. The noise is un-f*-believable low, wind-only, no buzzing or humming. One such ventilator flies off the f* table if it pushes the air downwards and it lifts itself off the ground if guided through its holes by vertical thick wires. I am lucky that in my country at this time of the year the temperature is around -32 degree celsius (-26F) in my area and the coincraft works inside the hogsty, or henhouse, hen-roost, use google-translate to understand this word. It stays where the chickens used to sleep at night, doors and windows are removed. The temperatures recorded by both sensors are around +44 degree celsius (+111F) with -32C (-26F) ambient temperature. No more power connectors to the modules, the wires are soldered directly on the board. The PSU does not show any problems yet. This is COOL and from me the single A++++ and respect for bitmine. I'm melting ice on 40 x A1 chips. For the rest? arse-f* them. I want to see if the unit survives this night.
The A-1 is f* strong and well designed. But the miner design they chose to get fast money..... that is the job for a trial with judges, lawyers and jury. And bitmine's arse in jail for sh*ing on this wonderful chip design and of course on people's heart and trust.