Anyone else think that these boxes aren't asics? With the size / power consumption i'm willing to bet these things are rocking sophon cores. If that is the case, hard fork all you want, it just takes a cgminer update for these to come back online.

It's a bugfix in the sia network implemented to fix block times, it's was announced quite a long time. Difficulty is still expected to drop quite a bit,

Guys can we stop the miner FUD?

Global hashrate is around 8P right now, which means there were roughly 10K A3's released in batch one. Block time is still around 20 min's, this has been decreasing and will hit 10 min soon. Right now profitability is near 1300SC/day which should be near 2600SC/day when the block time corrects in a few days time.

Until batch 2, 2600SC/day should be stable. If batch 2 is also near 10K units this will halve to 1300SC/day which is still $30 ish per day at current price. If batch 2 is 20K units then we should be seeing around 650SC/day or $15 ish/day at current price.

So the only way batch 2 stands to not be profitable is if there is a major crash in sia price or more than 20K units are shipped. Is it the $600/day that people were getting on day one, no. But it will still likely be one of Bitmain's most profitable miners until batch 3.

Play around with similar clock speeds. Also make sure fans are manually set to 100%. 575mhz was getting 70C for me, but 581mhz got 55C. Something is wrong with their voltage calculations. This is also subzero for me as well.

Controllers and all observable software / firmware are identical. If they are the same miner the change was made at the ASIC / PIC bootloader level which will take quite a bit more work to figure out, let alone figure out how to move algos. It will likely not happen even if it is possible. This is coming from someone who has already torn their A3 apart software and hardware wise to find out.

you can use multiple PSU's, just make sure each hashboard is dedicated to a single PSU, don't split the a single hashboard's connectors across multiple PSU's.

Usually the eeprom's are different, L3/S9/D3 all have different eeprom content. Except the A3 and D3 share the same eeprom content. These eeproms are on the i2c bus and each hash board has one. No clue what the data is actually used for, but considering every miner has been different to this point except the D3/A3 suggest's they are more similar than bitmain would like us to believe.

Also, i traced out that header on the A3 board which all seems to connect to the PIC16F1704 chip.

1 = 14 = VSS
2 = 12 = RA1/ISCPCLK
3 = 13 = RA0/ISCPDAT
4 = 14 = VSS
5 = 1 = VDD
6 = 4 = Vpp/MCLRT/RA3

That basically makes it an ICSP header, which might be able to gather some useful information on the programming. Clearly it isn't locked as cgminer is able to flash this chip on startup if needed. Based on the code it looks like the PIC has a small boot loader program which initializes things, then waits to hear from the controller. Whether the controller actually has an ICSP connection to the PIC itself, or if the small loader program has the ability to read data over i2c for programming i don't know.

/bin/pic.txt has the pic microcode in it, but it may not contain that initial loader program.

Have spent quite a bit of time with mine, I've gone through the whole unit software / hardware wise and made quite a few hardware notes that ill curate and share at a later date.

Voltage control is whack on these, needs custom firmware. Over the past few days of testing i was getting the following.

600mhz, 80C, 10K+ HW errors
581mhz, 55C, 5 HW errors
575mhz, 70C, 7K+ HW errors

How reducing clock speed results in more heat + HW errors is beyond me unless there are serious voltage calculation issues with these things. Also the CPU on the control board is running near full tilt just to keep up with the UART activity on the hashing boards. So even the simple act of refreshing the web page can give you HW errors. These aren't real errors, but just the control board lagging behind.

Also these are identical to the D3 in almost every way. pic firmware, eeprom images, etc... are all identical. The only thing that sets this apart from a D3 is the ASIC's, but considering it has all the same issues that the D3 had it makes me wonder if bitmain has come up with a more universal ASIC chip that can be reconfigured. There are JTAG headers on the hash boards, so who knows?

Yea, got mine on Thursday. While i have been enjoying the past couple days, simple math showed it wouldn't last long. If batch 2 isn't released till march then we may have a good chance of a 30 day ROI. To drop to $45 we would need roughly 14K miners online which so far it seems like they didn't sell that many. So i think somewhere between $50 and $100 is our landing point until batch 2. Just pray that the Chinese new year delays it quite a bit.

For the price of this miner, $50 a day is pretty solid. You'd be hard pressed to put together a GPU miner for that price that could hit similar numbers. I mean surely you didn't expect the $600 / day rate to last very long, especially with how small the global hash rate was.

I have an idea that could improve ROI on the D3, but i don't have a D3 to test on. Will someone who has a D3 do the following and send me the files?

nanddump -f /initramfs.bin.SD /dev/mtd8
nanddump -f /am335x-boneblack-bitmainer.dtb /dev/mtd6
nanddump -f /uImage.bin /dev/mtd7
nanddump -f /u-boot.img /dev/mtd4

You can use an application like winscp to get the files off of the miner. The miner uses a ramdisk, so these files will disappear on reboot so you don't have to worry about storage space.

Actually been working this angle, have a pretty good idea of what makes a miner a D3 or and A3, but i need a D3 to test or someone with a D3 to send me some data to try to turn my A3 into a D3.

There's pretty good reasons for bitmain to do something like this if they could. It would massively increase profits for them to have a ASIC chip that could be re-seeded to different algos.

Also if you want to dump firmware the bitmain way...

nanddump -f /initramfs.bin.SD /dev/mtd8
nanddump -f /am335x-boneblack-bitmainer.dtb /dev/mtd6
nanddump -f /uImage.bin /dev/mtd7
nanddump -f /u-boot.img /dev/mtd4

You'll have to hex edit those files afterwards and remove the trailing FF's to make them exactly like bitmain sends them out.

Looking for a D3 in just about any condition. Can have dead boards / chips. Just need at least one board functioning well enough to mine some amount of dash. Shoot me a PM if you're willing to let one go. Will escrow if needed.

Just based on the global hash rate, it looks like most of these boxes are online now. I would say the tapering is starting. My guess is somewhere in the $50 - $100 / day category until the next batch.

Not only that, but consider that the L3+ still sells for around $3K while only generating around $10 a day in profit, while the A3 is still north of $150 a day in profit.

Of course i'm not biting at that price, ill just wait for the next Bitmain batch. Bitmain prices are actually fairly reasonable most of the time.


And to add to my A3 adventure i ended up under clocking it to 575mhz, still pushing over 800GH/s but my temps are much more reasonable and the HW errors have nearly vanished. Was getting about 10SC/day at 600mhz/850GHS with about 3600HW errors per day. Now im getting 11.5SC/day at 575mhz/810GHS with about 4 HW errors per day. Chip temps went from 80C to 60C

Your miner configuration is wrong

Let this be a PSA to everyone check your temps. If the chips are over 80C you need to ramp up the fans. With 75F/24C ambient temps, mine would hit 75C with 100% fans! I moved the miner into my garage with 32F/0C ambient temps and the chips are still running 65C... If you are hitting 80C or higher you are risking a dead board during the time this thing is most (or possibly at all) profitable.

That's just a dead board plain and simple. Aside from getting warranty work done on it there's not a lot you can do. Could be power circuity, and if your good with soldering iron you might be able to fix it. But i would not advise doing it while under warranty. I'd just get it replaced. Unfortunately profitability is going to be much less by the time you'd get it back. Personally i'd ride it out on two boards till the gold rush is over, then send it in.

Pushed mine to 625mhz with no real gains. I think these things are being pushed to their limits with the factory clock. Tried 650mhz, started getting a ton of HW errors.

Lets say you draw 1400W from the wall which at 120V is 11.6A. (1400W/120V=11.6A). But in this case the 18 gauge wire is too thin which is making it act as a resistor (which is why it gets warm / hot to the touch). Because the wire in this case is acting as a resistor it will drop some voltage under load. Let's say it drop's 15V making the line voltage at the miner 105V. The miner still pulls 1400W regardless (1400W/105V=13.3A). So the amperage went up to compensate for the drop in voltage and the wire / PSU gets hotter as they are dissipating that extra energy as heat.

This is why many recommend 240V, as 1400W/240V=5.8A. The energy consumption of 1400W is the same (actually a bit less as the PSU is slightly more efficient at higher votlages), but the amperage is much lower which means you can fit more devices on a single circuit.

Basically if your power cable is warm, you are dropping voltage and wasting electricity. Not to mention creating a potential fire hazard. For this miner i would say 16 gauge power cables are a minimum requirement at 120v. At 240V 18 gauge should suffice just fine.