JLCPCB Prototype for $2(Any Color): https://jlcpcb.com
Stripdown of a Bitmain Antminer T9+ cryptocurrency mining rig. This monster can calculate 10.5 trillion SHA-256 hashes per second. I thought it would make a nice shed heater (with cashback).
Stripdown of a Bitmain Antminer T9+ cryptocurrency mining rig. This monster can calculate 10.5 trillion SHA-256 hashes per second. I thought it would make a nice shed heater (with cashback).
Rams? uh, RAM stands for Random Access Memory
10.5TH in the box, 64.8GH per ASIC. In 10 mins. Means 17.5GH / sec which is quite quick even for an ASIC. So my guess is that each ASIC is loaded with the start and runs like shit, multiple calculators in parallel (as many as will fit in) to get an answer, send them all back to the controller which then tests for a hit. If there is one, fills out the complete packet and sends it back into the pool. So this is the question: will a standard LCA (ie not a Zinc) have enough CLBs to make a sensible calculator for a sensible power consumption for your room heater ? ๐
1400w at 12v… Sounds like the consumption of a boat 12v to mains inverter.. or a small plasma cutter. easy stuff ๐ So did you fix the controller?
Does this mean 5amps at 240?
Or does it not take 240
could you run this bitmain miner directly off of a battery bank with a range of voltage at 12-16vdc?
Hi. Are you share that this t9+ hashboard has 54 ASICS on each side? I thought it was 54 ASICS PER hashboard, but again what's the other side for?
Hurray, splendid. I was a bit concerned when I could not find the hardware specs of this setup online. And it got me quite curious.
The new KD5 miners sell for around 40000$.
I am wondering now; how about connecting up these miners and using the extracted heat to power a generator of sort, perhaps a stirling engine, feeding the energy back into the system. That might make them more efficient as such.
Again, thank you for the investment and share.
Highest regards,
Mac
Sir can you sell me this
Question: Of all the watts that are used by a miner, do you know how much each component/board/fan uses?
Great teardown video, Thanks.
what's inside a bitcoin? a board with a microprocessor wrapped with a bronze cover?
Is there any possibility to use a hashing board like this in connection with another antminer running a different algorithm? E.g. a x3 or something like that? Because I was wondering where the algorith is located on that thing (I thought it would be just on the control board)… Any ideas?
What a waste
I'm working on repairing S9 hashboards and see from this video that there is a strong similarity to the T9+ boards. Along the way, I've learned alot about the ASICs and boards so may be I can add some color to the commentary in this video. The T9+ has 54 asics per board for a total of 162 asics. To get the total hashrate, you need to multiply the number of asics (162) by a) the clock frequency used to drive the chips and b) the number of parallel SHA-256 hashing circuits inside each chip. Apparently, each SHA-256 circuit is able to output 1 hash for each clock cycle. I don't know the clock frequency or number of parallel circuits in the T9 asics but for the S9 with 192 total asics, operating at 13.5 TH/sec with a clock frequency of 625 MHz, you can back-calculate the number of parallel circuits within each asic as about 112. I'd bet that the T9 asics aren't all that dissimilar to the S9 asics – different generations of the same theme.
The controller communicates with the asics, the pic and likely a temperature sensing chip via a serial bus which is at least similar to I2C if not the same as. The pic may have its own serial bus. It looks like the T9+ may use 3 serial busses for each board – not sure, the S9 uses 1. Because there are so many chips, the communication signals are connected to the asics in daisy-chain fashion. The likely reason for this is to keep the electrical load on the serial bus within reason. The controller's serial bus signals are directly connected only to the first asic. The first asic then (apparently) contains a driver circuit that passes the serial bus signals to the next chip in the chain and so on for each successive asic. With this arrangement, the controller only sees the load of 1 asic on its serial bus and each asic also sees the load of 1 asic (the next in the chain) on its serial I/O pins.
Since each asic depends on I/O signals from another asic immediately "upstream" or "downstream" in the chain, any asic in the "chain" that fails prevents other chips that are further down the chain from being able to communicate with the controller. So, if the monitoring software built into the miner reports that it found 37 asics whereas 54 are physically installed, most likely what is happening is that chip #38 has failed and is therefore unable to pass the communication signals further down the chain thus isolating and incapacitating those asics. The chips past #38 are probably good (no guarantees since they can't be tested) but the control board can't determine that because it cannot communicate with those chips. The remedy in that situation is to replace #38.
In the S9, and I suspect also in the T9+, there is an I2C based temperature sensing chip. That chip is able to connect to 2 temperature sensors. The "sensor" is or can be a special type of transistor that has a known temperature profile vs. voltage, resistance or current. One sensor of this type is built into the temperature sensing chip. This sensor measurement roughly reflects the "board" temperature. Each asic also has one of these special transistors built in to be used as a temperature sensor. However, only the sensor in the last asic in the chain is connected to the temperature sensing chip. This sensor reflects the "chip" temperature but keep in mind it represents only the chip at the end of the chain. The asics upstream "should" be operating a little cooler since they are getting cooler air – the air warms as it progresses down the chain. If an asic in the middle of the chain fails and thereby prevents communication with everything further along the chain, then the temperature measurements also become unavailable. Without a temperature measurement, the software cannot properly modulate the speed of the fans so they just go to 100% when there is no temperature measurement – the only safe option other than shutting down the whole chain.
The voltage regulation circuit with the 4 mosfets mentioned in the video steps the voltage down from 12 volts to about 9 V on the S9 – probably a little lower on the T9+. This 9V is then divided up by additional voltage regulators that are closer to the chips – in series fashion. In the S9, there is an LDO voltage regulator for each group of 3 asics and as there are 21 groups of 3, each group gets about 9/21 = 0.425 V to feed the power hungry portions of the asics. The PIC controller outputs a signal that inhibits/enables the dc-dc converter, another signal that allows for slight adjustment of the dc-dc converter output voltage and also a clock signal that determines the clock frequency used by the asics. So, overclocking and undervolting are possible with the right software access since the PIC is able to adjust the voltage and frequency.
There are a variety of problems that crop up with hashboards. With so many asics, the probability of an asic failing is arguably the most likely thing to happen. The mosfets can also fail and there are several capacitors that may present weak links. A problem I"m perplexed has to do with one of the communication signals named BO. It is a signal that indicates an asic is "busy" communicating with the controller – to avoid bus contention. Normally, this signal should be logic low but on some of my boards it is continuously logic high thereby preventing communication – so the monitoring software reports that it finds 0 asics. If anyone has any insight on that matter, please leave a comment here. I've not been able to find any information about that issue anywhere.
I stopped watching this video at the point where this dude said he bought a Bitcoin miner to heat his home… Iโd laugh but itโs not really funny…
positives feedback from kyle.hackss on IG they are legitimate