Topic: Building Your Own ASIC Miner? Yes You Can! (Read 1029 times)

Yeah i do not think you are crazy to come up with this idea.

Even i was thinking the samething back in last year NOV 2021.

It is possible to clone a PSB and the we can start with the S19 Pro. I was in china more than 7 years i can speak the language and move on the project .

i would like to work with someone would like to gang up with me.

and even i can manage to get some second hand control boards for cloning . and the second hand chips for the first prototype virsion .

if anyone interested to work together . i would love to work on this project.

i think we need a good electronics guy with Circuit drawing skills .
my whats App (0086 185 2251 0643  )

Well, of course you can if you can get all the parts, design a board yourself, and build it.

For example that's basically what the Compac F from GekkoScience is. It's a custom board that uses one BM1397 chip, which is the one used in the S17 miner.

Here's the main component: https://asicminerparts.com/products/bm1397ag-asic-chip-s17-t17

But of course, most people won't have the skills, time, or desire to do all that work, so they just buy the miner pre-built by someone else.

For short it's no, here's why

Application-specific integrated circuits, or ASICs, are circuits that are tailored to a certain application. The logic gates are connected in such a way that they can process data in a specific fashion, as the name implies. What's the best way to get them? You can't, as far as I know, because the makers haven't placed it up for sale on ebay or Amazon. Instead, they were utilised in tandem to build circuits and offer them as minors in pre-configured configurations. Is it possible for you to construct one on your own? Yes, but not in a straightforward manner! To begin, you must understand how to programme FPGAs (Field Programmable Gate Arrays). It's a semiconductor made up of Arrays of Programmable logic Gates, as the name implies.

You can programme the gates to do a certain computation if you like. (An aha256 hashcode is calculated in bitcoin). The result will be a chip that is extremely efficient in terms of power usage and cost. Then you should place an order for your chip with a business for production, for which you will have to pay a large sum of money. So, to make your order cost-effective, you need order thousands of them! So, unless you want to invest a large sum of money and start your own pool/miner firm, creating your DIY ASIC miner right now is pointless in terms of cost-benefit analysis!

https://pureweb3.blogspot.com/2022/04/can-i-build-my-own-mining-asic-from.html

Um you'd better read the pinned post at the top here, mainly point-3.

I would like to build my own Bitcoin Mining system with AMD .   : )

@mellosoft
Maybe find the maintenance manual and reverse-engineer it with the manual and inspecting an actual miner?
Zeusbtc.com might have the manual. It's worth a while to ask them and check through their website.

EDIT: I found this guide --> https://zeusbtc.com/manuals/Antminer-S19-Pro-Hash-Board-Repair-Guide.asp

^^ Does not exist outside of Bitmain. The s9 chip data sheets are the last that BM released to the public.

Can Anyone help my find the BM1398 BM1398BB datasheet please, I have an Idea!

Apparently the company that made the cloned S9 hash boards almost doubled the hashrate on them. However that means there would either have to be more chips on board or they essentially overclocked a board, which would mean that modifications would need to be made to the firmware and to the microcontroller on the hash board. The companies which produce cloned FPGA board clones (like arduino clones) usually tend to use a slightly different version of the ram chips or some other chip on board which then they can sell for a profit.

Having a trace somewhere that is slightly longer wouldnt effect things too much as far as i know. Damaged traces or vias could cause problems but in most cases, any problems to those mean that theres not gonna be a continuity through the board at all to begin with.
Now of course there are different guages of traces that can be used on PCBs and the thicker the guage the more the resistance but then the difference would be so small if any that it would be unnoticeable.

Making a one to one copy of the hash board like i suggested/plan on doing shouldnt have any considerable difference in the hashrate or efficiency, provided that the PCB used matches the appropriate ratings for the amount of heat and current that is going to be passed through it because, in the end, its the control board thats in charge of making sure that the chips are running at their required potential and giving out the required hashrate.

It would have to be at least some light modifications made to the PCB because I don't think companies can get away with making a token change to some circuit. Which makes me wonder if the cloned PCBs you get from this will be as efficient as the original, since chances are the design houses will not make a perfect clone of the PCB. Maybe they'll do stuff like making some path longer (I think I'm a developer, not an electrician or circuit designer) that makes some part of operation less efficient, to avoid infringing on some random company's IP.

Its obvious that the custom made hash boards wont hold as much value as the original hash boards if you're trying to sell them. Because essentially what you're doing is substituting the physical labor, time and effort that you are going to put into assembling the thing to save yourself some cash which you would need to buy a new hash board (in my case anyway, but i am sure it was the same for that company). I wouldnt even be surprised if a company started selling custom made hash boards for more than what the original ones cost because in the end they had to spend more time modifying the board (which would be a terrible decision but it happens).

If you thought i was making this post because i wanted to mass produce cloned hash boards and sell them for a profit because i claim to be able to produce them for cheap, then no, thats far from what i meant. I am aware thats not going to be at all profitable and moreover theres always a legal caveat to the story (which is why the companies tend to "modify" their clones instead of making a copy). I was more of planning to get back the ROI from the boards themselves by using them, if they worked of course.

However mass producing and selling custom made hash boards is gonna be way less profitable than making a few of them for yourself to use for mining (which the companies are doing as well like i said before) because, as you said, people would rather go for the original ones or even an original different miner alltogether from a trusted source since the price isnt gonna be much more different.

As for the custom firmware, it was most likely because they produced a modified version of the hash board to run on the control board for the S9. I am pretty certain that making a one to one clone of a board will not require a custom firmware to be flashed on the control board and it should just work because every single bit of data is freely accessible on the microcontroller.

Moreover, coincidentally enough, i was contacted by a chinese company to analyze the firmware on the control board for an s17 back in 2019, i was able to get to the linux shell on it in about two hours. Since then there have been a few writeups on this (not by me), one of which can be found here : https://hackaday.com/2020/12/10/a-xilinx-zynq-linux-fpga-board-for-under-20-the-windfall-of-decommissioned-crypto-mining/

Yeah I was sent three testing boards.

They did 53 watts a th on lowest power setting and 62 watts a th on highest power setting.

maybe 20th on the high power setting in a s9 frame. I liked it but moq was high and we passed.

Another member(he can reveal himself if he wants to) and I were contacted by a Chinese manufacturer who did something almost identical to what you are trying to archive, they made their own hash boards which were more efficient than the regular S9 hash boards, and their own firmware which you had to deploy on the default controller of the S9 (the hash boards won't work with the default firmware).

The company wanted economical consultation in terms of pricing, I contacted a few major resellers in China, Russia, and the U.S, and sadly the final sale price was not tempting enough to convince anyone to get those hash boards, mind you the company that did this is located in china, and they paid us just for the consultation, they clearly had the money and the manpower needed for that job, and the results were not so great.

I am not aware of where did they go wrong or how difficult it was, but I am certain they had more skills, money, and resources than you do, which is one of the main reasons why I told you in my first comment that this "won't work", it wasn't a blind baseless statement, I hope the above information will be beneficial to you oneway or the other.

Good luck.

I agree with that one, hidden costs could and do usually come up at any point in time that i dont know of since the price is always fluctuating.

The reason i am confident and almost a little too confident about this is because i've worked on repairing asic miners and particularly the control and hash boards for a long time, i essentially built two hash boards by parts for an s17 pro that was struck by lightning where i used the chips and microcontroller from one and the pcb from another and sourced the other components on the internet, this is how i am certain that they are not hardware locked. This is essentially the exact same thing with the difference that the pcb is required to be manufactured rather than sourced from a different miner and i'd have to manually copy over the data on the microcontroller.

But this, indeed, is quite an expensive endeavour so i definitely should try it out with an older model first. I am gonna keep searching for someone who could help me copy the PCB in the meantime and i'll update this if i find any useful information on it. Thanks!

I don't why but I think you are under the impression that I am trying to make you change your mind as if I own a huge chunk of the mining industry, it is actually the exact opposite, I am a buyer I would most certainly love to see the likes of your idea get implemented so I can buy cheaper gears.

But I know for a fact that it won't make any economical sense, ignore the little details and just look at the big picture, you are trying to compete with multi-billion dollar companies, if Bitmain, MicroBT, Canaan or Inno had a way of making an S19 equivalent for less than the market price they would have done just that, they won't leave this money making opportunity to the average joe.

This means there are some hidden and unknown costs that you don't seem to be aware of (I am not aware of them either), you seem to be simplifying the matter way too much, this is either overconfidence or pure misunderstanding of how things are.
 
You claim that you can make a mining gear that goes for 14k for just under 6k + the PCB cost, then my words hold no meaning at all, go ahead and do it, meanwhile, you might want to test on a cheaper gear like the S17 or S9 given that the success rate of this operation isn't too high.

Good luck

I respect your opinions but i just simply disagree with most of what you said. Some of your opinions seem to be dependant on more of a "what can/would" instead of "what is"  while the others are flat out contradicting my research without a valid reason.
I asked the sellers to make me an invoice along with shipping and handling and i am literally able to purchase the chips for $10 to $14 depending on the quantity with around $30 shipping (not including customs and handling).
They offer to send you a paid sample ($25) after your order is booked to make sure of the authenticity of the chip before making the full payment.

No ones talking about cloning an s19 pro miner, this is about cloning the main PCB thats found in that miner. As for the time and soldering of the chips on the board, i am a software developer and i do hardware repairing job as a hobby on the side and hold over 6 years of experience. I can easily get access to BGA rework stations if need be. As stated before, getting 3 hash boards with 342 chips ready would take me no longer than around a day or two provided that i have a good PCB, excluding the hours it takes to reprogram the microcontroller.

As for everyone making their own cars and Phones. Think of it as replacing the wheels of an already intact car with aftermarket ones and replacing the memory chip of a new phone with a different one of another phone of the same make-model, and yes, i am not talking outta my ass on this one, this is actually being done for iphones nowadays but thats besides the point.

You clearly dont have your research on the parts market and your statements simply state, "this will not work because... no" or, "because... economy" instead of a good explanation as to why while every bit of information i have shared can be traced back and confirmed at least at the time of writing this post, if you do your research.
The resellers are selling the machines at such high profit margins that assembling it from parts has become cheaper.

Now of course all this is theoretical and its hard to say for certain without knowing the cost of cloning the pcb and the shipping costs might fluctuate a bit, but right now at the time of writing this post i, and even you can purchase EVERY SINGLE thing you need to assemble a working S19 Pro miner including shipping OTHER THAN the hashboard PCB for (and yes i have calculated this after getting in stock ready to ship part quotations from sellers) : $5890

The lowest price on a S19 Pro IF in stock that i could find in my three hour long parts hunt on alibaba is $14800 and that does not include shipping.

Now i expect you to, yet again, "highly doubt" this. But i have no reason to flat out lie on a public forum while the prices are out there on the internet for everyone to see and find out.

I read every single bit of it, twice, and I have no changed my mind, this will NOT work out as far as the economy is concerned.


This does not make sense, there is chip shortage and high demand, your "someone" can sell the chips for a large profit if they are indeed getting them near or at cost price, so no, nobody is going to give you cheap chips, and then even if that was the case, you are better of reselling the chips for a premium than trying to reinvent the wheel.


That is not a fact, it's an assumption, there is a technology edge you seem to miss here, just by accruing the parts needed to make something does not mean you can "assemble" it cheaper than the manufacturer if this was the case, everyone would be making their own cars, TVs, phones and whatnot, your theory is only partially valid if you had nothing else to do, in other words, your time is worthless and thus there is no loss of opportunity cost.



I highly doubt that, they just don't know the latest prices and don't even bother asking their sources, try to make an order and you will be hit by the real prices, but then really the exact number does not matter, the price will be a reflection of the whole market, and the important part is that the total cost of the gear you are going to assemble is not going to be lower than the market price, be it $20 per chip of 20 chips for $1.


This does not make sense, if the profit margin is as HUGE as you think, then nobody would have enough resources to run all those gears, someone would have made tens of thousands of them and fully utilized every bit of power they have and will eventually have to start selling, also there are no-name gears coming online, we are watching the hashrate very closely, everything points to a massive shortage in gear making across all the major manufacturers, I hate to break it to you but, NO, nobody is assembling S19s for a fraction of they go for.



And I did not disagree with this, my whole point was economical, I know the technical part will be workable somehow, everyone can manufacturer almost everything, it all boils down to the cost and quality of the end product, there is no way on planet earth that a single person like you or I can beat the market, do you have the slightest clue how long it is going to take you to solder 342 chips which is probably the easiest part of your experiment?

I don't want to sound rude, but it's funny that you think you found a glitch in the game or a money printing machine which ONLY you know about, it's crazy because it sounds just like perpetual motion.

With that being said, if this is to be done for the sole purpose of learning or having fun, I'd go for it, but if you think you will assemble a miner at a reasonably lower price than the current markets, well, good luck.

Would really appreciate if you at least read the whole thread before jumping to conclusions, you clearly half assed through the convo before writing your response and have no clue what the post is about.
And moreover 114*3*20 = 6840 rather than 8640

My whole point of writing the post was the fact that i personally know someone and i can maybe get my hands on at least some chips for near cost price and the fact that assembling the miner on your own from parts is somehow costing much much less than what its going for. (assuming that cost of cloning a pcb isnt astronomically high which i am certain its not)
Moreover that aside, checking with other sellers on alibaba at the time of writing this message, the chips are going for around $10 to $14 maximum, dont have to take my word for it, you could inquire with alibaba sellers yourself.

114*10*3 = 3420
add around 200 for the control board and 250 to 280 for a psu.
There is a microcontroller, some SMD components, some electrolytic capasitors and an inductor on the board, all of which can be arranged for no more than $50 to $100.
The peripherals including the fans and wiring cost no more than $100.
The rest depends on the cost of cloning the PCB .

Give or take $1k on the whole total, you are still looking at less than half the price of what the miner is going for and there is NO possible way you're ever going to reach that amount of money if something like this ends up working (14 to 15k).

Now am i saying that everything is going to be 100% genuine and is going to work on the first try?
No, nothing is certain, it might even end up being a total waste of time and money.
But then, that goes for the whole crypto scene/market alltogether anyway.

As for the market not getting flooded with no-name miners, if the company is able to self-assemble the miner at less than half the amount of what it costs then the ROI is so quick that its not worth the risk to lose money by putting it up on the market. Moreover companies know that if they decide to put it up on the market they would need to claim the product to be their own and bitmain is a closed source project plus it would lower the ASIC miner sales and that poses the risk of getting sued by bitmain and/or quite possibly a supply cut off of chips and miners from them for that company. The distributor companies who do these things tend to work along with bitmain rather than working against them.
There are numerous other companies who claim to be producing/manufacturing ASIC miners but refuse to put them up for sale.

All i am saying is that, In theory there is no reason why it wouldnt work and i might put some funds into this in the coming months and it would be great if someone with pcb manufacturing ability is willing to help.

That's far from accurate, the  BM1398BB chips which are used on the S19 cost over $20 (ignore prices on Alibaba/Aliexpress because those are either scammer or failed to update the price list).

So for an S19 pro you going to need 114*3*20 = $8,640 in Chips alone, you might find them a bit cheaper, but nothing remotely close to the 2$ mark.


I disagree, and I am certain that it will cost you above $10k, you are trying to beat bitmain into making a "cheaper" gear is daydreaming at best, honestly, I think any rational thoughts should be based on the fact that ALL markets on planet earth eventually get to an equilibrium, whereby the "arbitrage" profit becomes too small.

So with that in mind, you have to keep in mind that there are a dozen of businesses there, including some Chinese BILLIONAIRES who have access to more resources than you do, they have hundreds if not thousands of staff and robots, if "cloning" an S19 pro today costs less than 10k, everyone and their grandmother in China would be doing that, the S19 pro sells for 14-15k in China, and it's selling HOT.

That 14-15k isn't a random number that Chines sellers decide over their wechat chat-room, it's a perfect mix of supply and demand, NOBODY was able to increase the supply and that's why the price hasn't dropped if what you think is true, the markets would be flooded with a no-name 110TH miners or so, but why aren't there any? unless you think you know things that nobody else does, which is unlikely.

I really admire your ambition, but what you are trying to do is worse than reinventing the wheel.

The chips are running in parallel to one another, if the temperature on any of the chips goes above 85C (185F), the chip stops functioning until its cooled back down to around 60C (140F) by the continuously running fans which ramp up upon a surge in temperature.
Now even though the thermals are well managed by software alone, i am sure manufacturing the PCB would require some specific things to be taken care of but, then again, thats where my knowledge of PCBs end and thats exactly what i need help with.
My point by the above statement was assuming that if we have a properly cloned PCB, slapping on the components isnt really too hard of a task.

Now as for the cost of cloning the pcb correctly and also the claims of alibaba sellers saying they are able to clone any pcbs for cheap if you ship one to them, i am in the dark.
But what i do know for sure, and i am sure you'd agree, is that its still not going to cost anywhere close to $10k.

That is a vast oversimplification of what goes into making hash boards...

This is a very high power density application that requires special attention to the thermal properties of the boards and where the chips mount. Specifically, the need for very thick power planes and a dense pattern of thermal vias under each of the chips to pull heat away from them and transfer it to the backside heat sinks. Many PCB houses have no idea how setup their reflow ovens to do that properly.

May I refer you to part of the results from forensics I did for AMT back in 2014 regarding why their hash boards were literally going up in flames https://bitcointalksearch.org/topic/m.7610684

The PSU can supply a max of 12-15V 233A to the hash board while it only supplies 12.3V 15A to the control board as seen from the datasheet of the PSU on bitmain's website linked below :
https://shop.bitmain.com/product/detail?pid=00020210302152412332Xrq3G3E60644

As for the mounting of all the components, they are cheap and available to buy from any component website or even alibaba including the main hashing chips and all other ICs as i stated before. The board has at most 5 kinds of different components on it other than the main hashing chips, tracking those exact components down is a piece of cake.
For someone with a tiny bit of experience with BGA and SMD components, getting a hash board ready would take no longer than an hour.

This, however does come with a caveat. The boards do have a microcontroller on them if i remember correctly to communicate with the control board. (i'll update with the exact controller model once i get the chance). But the data pins on the controller are freely accessible on the board itself (they soldered on a connector for it XD). Copying over the code from the controller to another should not be a hard task at all.

I'm pretty sure the hash board will draw a lot more current than the control board, right?

Somehow it's hard to find digital PSUs on Amazon that can output 150A and not some low amount like 5A or 10A, which makes me wonder whether high-power drawing PSUs are are only readily available from generic sellers on Alibaba (considering most of the PSUs end up in China anyway)?


But then how are you going to mount all the components onto the cloned PCB when you don't have duplicates of them? That's assuming you can somehow find out the model numbers of them all.

The control board and the hash board require separate voltages. In my opinion the best bet would be to spend the $300 to get an actual PSU thats meant to work with the miner. This then means that the only component thats not factory is the hash boards (and the fans). The hash boards simply have some SMD diodes for reverse current protection and some capacitors along with the BM1398BB chips.

Now speaking from experience, i've replaced hashing chips on antminer s17 pro in the past and they tend to work normally, the hash board does not look much different than s17 pro's hash board so i am pretty certain that the chips themselves are not tied to the control board in any way. Moreover even if they were, the chips are going to be from bitmain themselves so with a little bit of luck, they should end up working.

If someone manages to help with copying a hash board pcb for cheap then we're basically looking at an s19 pro for around $2000 if not cheaper, with added work of soldering the BGA chips and the SMD components on the board of course.
Alibaba does have services where you can send in a pcb and they'll clone it for you for cheap (again, pcb = just the board no components on it). But i sadly dont have a board i can sacrifice, hell there isnt even a proper image of the board on google yet -_-

As long as that PSU is externally controllable/programmable, it is exactly what is needed. Of course readouts of voltage & current are not required  
The key point is one needs to know how to communicate between the controller and PSU: Is it digital (I2C) or analog (0-10v = 0-100%)? Does the PSU report back to the controller to give basic status and actual voltage & current output?

Using programmable PSU's is one of the tricks used to make miners more efficient. Using 2nd stage hash board regulators to provide Vcore to the chips like the S9, A841 and other earlier miners is at BEST 95% efficient in converting 12v power to Vcore and more often around only 85-95%. By using a few more chips per-string you can use a higher baseline voltage and having the PSU itself do all regulation to set Vcore applied to the chip strings eliminates that drop in power efficiency.

Ah then to jerry rig something together one would need two power supplies with the outputs as mentioned in the link below, one to power the chips and the other one to power the control board.
https://shop.bitmain.com/product/detail?pid=00020210302152412332Xrq3G3E60644

Still, not outta the equation considering that the variable output power supplies are going for quite cheap on chinese websites.
The main chips receive around 12-15v at 233A apparently and the control board is 12.3V at 15A (i think)

Now that i think about it, the control board is supposed to communicate with the power supply through a separate ribbon cable, most likely to shut off a board in case of abnormal hashrate or a surge in current. I will ask about how they power their oil cooled miners in their company and update the main post once i get the info.

Thankfully the power supply is not that hard to reverse engineer either and the control chip is most likely available to buy just like all the other chips, i'll take a look at my psu and see if i find anything as well.

Edit :
After taking another close look at the back of the psu,
The power supply uses a micron pic16f1704-I/SL microcontroller to communicate with the control board and uses a few other ICs for protection such as ix4340n.
The fans get direct input from the transformers so the fan speed on the power supply is not variable.
Now the microcontroller is programmable and the pinouts are conveniently available to access on the psu.
this micron IC manages the voltage regulation and also helps with the "sleep" feature that s19 pro has.


Now talking to my friend again i learned that their company buys as many PSU's they can from bitmain since they're relitively cheap but then they also have a third party manufacturer who makes these PSU's for them.
The PSU's, original or aftermarket, are going from anywhere between $240 to $300 at the time of writing this. which, then again, is not out of the equation considering the price of the whole unit.

Do you know whether Bitmain's firmware checks for unofficial parts at runtime (perhaps like what Apple does) and powers off if there are non-Bitmain parts inside? I think they could use that to force people to buy their own miners.


I wonder if a part like this would work: https://aliexpress.com/i/4000307508670.html ? This is an external part so I'm not too sure how you'd connect it to the casing.

The miner does not use a generic 12v PSU and what they use delivers a lot more than 20A. 12v*20a=240w so not even close to the over 3kw the s19 pulls. Modern miners use variable output PSU's that adjust the voltage to between 15-20vdc at well over 150amps.

Hi there!
Recently i bought two S19 Pros and ended up blowing the PSU on one of them which lead me to open it up and replace some mosfets to fix it.
While i was at it i decided to give the rest of the miner a clean so i dove in and began doing that. While in the process of cleaning i took a note of the components inside on the main hash boards and the control board.

I have a friend in china who used to be a business partner of my cousin and she now works for a ASIC miner distributor company via alibaba. I got in touch with her to talk about the making of these machines to get some more insight.
Now in the two hour long conversation i learned quite a bit about it and i will be sharing a few of the things i learned.

The reason its almost impossible to buy miners straight from bitmain or other mining companies is because they look for high profits rather than customer satisfaction, there are underground bids which take place between large distributors and the highest bidding company gets the cake. This happens weeks before the miners are released on their website for purchase.

Cutting to the chase. The distributor company that my friend works in used to be a PCB manufacturing company. They buy BM1398BB hashing chips straight from bitmain and manufacture their own hashing pcb's by cloning one of the original ones and end up oil cooling them. Now to me, all of it makes sense since all an ASIC miner has is a control board, a power supply and a few hashing boards. None of the components are tied to one another when it comes to the hardware.

Now while its possible to clone the control board and the information on it is as follows, it looks like its quite a tedious task considering that the actual board is quite cheap to buy straight from bitmain.
CPU :  XC7Z007S-1CLG225C
DRAM : M15T2G16128A (2L)
NAND Flash Memory : MT29F2G08ABAEAWP 2GB
Network Chip : PPT PM44-11BP
The linux based firmware that is stored on the flash memory is later witten to the DRAM at runtime and is available to download on bitmain's website and can be flashed using the sd card slot on the control board.

Anyhoo, oil cooling being a whole different story, I was told that each hashing chip costs about $2 and some select few companies can hook you up with the original ones. Which leaves me with needing a control board which costs around $200 and wiring which costs around $50 including the power cords which are not included with the power supply or miner purchase. A generic 12v 20a power supply that s19 pro needs costs a mere $60 where i live. The power supply costs around $240 to $300.

Now of course i did not include the small components like heat sinks, housing, tax and shipping costs and such.
But in hindsight if i were to build the whole thing on my own, it would cost a considerably less amount of money considering that these machines are going for around $10,000 if not more, per unit, at the time of writing this post.

I've been interested in this since i learned about it a few months back, but i lack the ability to clone boards and funds to get access to a board.
My point of writing this post is to learn more about things that i might have missed or gotten wrong and/or to maybe try out this project with someone who thinks they can help or have experience in pcb cloning/manufacturing without a schematic and finds this interesting.