Topic: Buying Bitmain or Bitfury ASIC chips directly (Read 4457 times)

Hahaha... niice.

Finally proof of the Dragonmint!!!!!

---

(Moderator's note: This post was edited by frodocooper to trim the quote from fanatic26.)

Dang, where was this a year ago?

I mean someone has to have some data on these chips right?

Bitmain won't give out the datasheet until the next chip comes out is my guess. 

Great idea. Maybe we can start another Bitconnect and label it as ASIC designers instead of bitcoin trading robots.

You won't need a datasheet. Just copy or build an electrically equivalent board to mount a stack of chips on. The only think tricky about this ASIC is the secret recipe inside.

All the circuitry of miners is very, very simple.

If you get ASICs to sample, we can just replace one ASIC on a working board to test. easy, easy.

Head's up. Common Chinesium scam method:
- send working samples, aka actual ASICs pulled from S9 miner, to build trust
- continue to send pulled chips to get the pump primed
- receive money for bulk order
- send 20% real chips, 80% fakes
- Oh, so sorry for your trouble, please return failed units so we can analyze
- replace the 80% fakes with another 20/80 mix
- Oh, so sorry for your trouble, our analysis indicates you have installation method problem. We will offer you another replacement. Please return the chips that failed.
- return a 30/70 mix
Oh, so sorry. It looks like you did not fully correct you installation techniques. Perhaps we should build these boards for you here. Send us more money so we can start.

You see how this works? Many people there get very, very rich with this scam.

a. Cloudmining discussions are verboten in this area and thread
b. Cloudmining in general sucks the sweat off of a hogs balls because a good part of the income goes to the operators and NOT the miners

Bitmain has never sold their s9 chips. Period. That ^^ reseller is either selling stolen chips, ones salvaged from improperly disposed scrap, or relabled dummy chips. No matter what it is a highly ILLEGAL operation. DO NOT SUPPORT ANY GRAY-MARKET ACTIVITIES LIKE THAT.

Hello, I have a reseller in China who is able to give me BM1387 (the chip used in S9) but without the datasheet we can't create our own miner.
Is there someone able to find the datasheet? Clearly BITMAIN hasn't released it

Looks like we are getting forced to make our own community sourced miners and because chips are not available from big guys... our own ASIC chips as well. Bummer.

I work for a company that manufactures custom electronics, automation and various controls.
We have clients who want some specific custom miners to be made but we can not manufacture those without chips. We have access to manufacturing facilities in Asia and Europe and I am familiar with outsourcing manufacturing on larger scale (do this every day). There’s a definite demand for miners on the market, as we all know. Nothing stops us to make these units available for community as well (with quantities increase - cost per unit decreases).
 
Looks like we will be making our own ASIC chips at this point. We are talking to some fellow miners to join the effort. PM for more info.

Anyways, as sidehack said, 500W medium miners would be the units we will be going after as one of product in the portfolio. 500W on standard 120v North American power source will draw just a bit more than 4A. You can safely plug this one in any standard outlet at home. They would be waaaaaay quieter than typical Antminer and such.
If you really want, you can run 2 or even 3 or them on same circuit.
Something to talk about, for sure...

Fan of what you've done thus far. I have 12 2Pacs grinding away at my home.

I don't know how much hashing power one can eek out of 500 watts, but don't hate the idea. We'd not yet conversed about power consumption on our side, except to say that we wanted the unit to be meaningful/functional. Rough math (please don't hate me for not being an engineering guru) is that with 500 watts you could get somewhere around 4.5 - 4.9 TH, give or take?? That's more than reasonable, and if if manufacturing can be scalable it'd be sweet to pull off.

We'd even talked (this is quite a ways down the rabbit hole, but why not) about how even if this "mini miner" wasn't as efficient as Antminers or some of the others... the idea that if we can get/make this accessible... have a shelf full of mini miners because you can actually get ahold of them, and replace them later with the bigger/badder units when possible... and make the minis available to others at affordable rates in the aftermarkets.

I know our group would invest some cash and energy into trying to make it happen, though I have to admit we've not been far enough down the road to know what kind of undertaking it is, or whom we could even partner with, but we damn sure love to try.

Would love to pick your brain some day, and I'd happily fly to you and pay for a nice dinner to do that. Unless you live in Nairobi or something. Haha.

I've been interested in (and actively working on) building miners for about 3 years now, but I have zero interest in building something over about 500 watts. That's a practical draw for noise and cooling concerns of non-industrial customers.

Yeah, I kind of feel the same way.

My small group of affiliated miners (kind of a brain trust group, all separate, but sharing collective knowledge) has been toying with the idea of some kind of "mini miner" that packs enough hashing power to be worthwhile (5-7 TH) but also affordable to manufacture, so one could have the ability to scale on a budget, but more importantly on their own terms. A shelf full of mini miners, in theory, may not be as efficient as the Antminers of the world (hey, maybe they could be with some strong teamwork, and some good folks behind it) but at least we'd have access to them on our own terms vs. being at the mercy of these larger corps. and their mood swing, batch releases, shipping delays, and other complexities.

Feels nearly impossible, without access to chips, though.

So much for decentralization.

Well let me throw a new voice in the pile...

I reached out to BF this past week, and received a response Monday from [email protected] to contact Hut 8 Mining who now controls all BF product sales inside North America.

When I reached out to Hut 8, I received an almost immediate response that plans to sell chips in bulks are currently suspended. So talk about mixed messages, I got no such response about $2m minimums or anything.

I Google’d a bit and learned that BF and Hut 8 are essentially merging, and it appears the focus is now on creating the largest mining facility in North America, using BF hardware technologies I’m assuming.

With a 6-7 month lead time, I feel like we would be seeing next gen product (S10?) from Bitmain by then, which makes having a B8 Integrator Kit entirely pointless.

I've been informed that due to surging demand the price of the integrator kit is now $30k and still requires a $2M order and a 6/7 month lead time.

You wish

Performance
Hashrate: from 30 TH/s ±5% (max power efficiency) to 49 TH/s ±5% (max performance)
Power Efficiency: variable, from 0.101 to 0.133 J/GH
Power Consumption: variable, 6.4 KW (max)

Features
576 Bitfury 16nm BF8162B ASIC chips
Variable DC voltage for variable efficiency
Upgradable design
Proprietary Bitfury string power design
Proprietary Bitfury mining soſtware running on Debian Linux

Dimensions and Weight
Form factor: Rack (6U)
H × W × D: 19” × 17.32” × 25.6” (270 × 440 × 650 mm)
Weight: 37 kgs

Environmental Requirements
AC Power (2 inputs): 85–305 VAC, 200–277 VAC (nominal), 45–66 Hz, 16.5 ARMS
Networking: Ethernet 100BASE-TX, IEEE 802.3u
Operating Temperature: -20°C to 40°C (5–95% RH)

$20k for an Integrator Kit? What the hell? Is it like 100TH/s machine?

Once you get your PM working, give me your email, we will talk there without restrictions.

I saw your PM, but apparently, I am unable to respond because I am over my limit of 2 pms since I registered today and am a new user (I've been lurking here since 2010).

I asked for the B8 and Asic spec sheets and they sent them over. I was told, but not provided documentation about the Integrator Pack.

It was a general promo-sheet about integrator pack and B8. Did they give you spec sheets?

I sent you PM about it.

Wow, thats pretty silly. It appears there are a few things going on here.

I wasn't told anything about getting reimbursed for the Integrator Pack purchase. So for the policy to change from that to what I was quoted in a span of two days is odd.

Spend 2 Mil and wait for a year to get your product. I don't think I have that kind of trust for vendors in general, let alone here.

What was attached for the integrator kit information? I had to ask for the B8 and asic chip spec sheets.

Conversation took place on Dec 18, 2017.
“We sell integrator kit (see attached) for $20k. This is reimbursed against final invoice when $2M+ chips purchased.

They also said that bulk order chips will be ready by September 2018 (if ordered now).  Thus, I’m very confused about their pricing and ability to ‘actually’ deliver.

When (time of year) did you ask for a quote?

The messaging was rather disjointed, but I'm still not sold that more than two people (BF side) were involved in the exchange. I watched an interview with a BitFury guy at some conference from Oct 2017 where he said the price for B8 units was 5200 and I was quoted $7k. Maybe they do it to see if you balk at the price tag or ask more questions?

I was quoted $20k for integrator pack. Do you have hey just throw numbers around to see what sticks?

Thank you. Sorry for not doing any research first. Shame they don't sell directly for less than $2m, I would love their B8.

The B8 itself houses 576 chips according to their specsheet.

The 100 chips included in the order are for prototyping purposes.

I asked them that same question since they advertise the chip differently on the site and got this as a response:

"The 75GH/s I stated is based on very conservative test results. In the right environment, the chips can run up to 100GH/s."

The specs are different for 100 chips specified hashrate above.

Perhaps their unit uses different chips? Have I done a mistake with hashrate?

edit - their B8 unit has almost 600 chips

I should have probably included this:

"The integrator pack consists of 1 full B8 server, 100 chips and 15 control chips."

I believe the B8 Included with the pack is ~50TH

A standalone B8 is $7000 is available as well but needs to be part of that $2M order.

So when you consider the price of an S9 (~14TH @ $2830) compared to a B8 (~50TH @$7000) you get ~$202/TH (s9) vs ~$140/TH (B8). Previous to BitMains price increase the S9 was at ~$107/TH. None of these numbers factor in shipping costs.

If I understand correctly, with their integrator pack you can build your own?

$7500 for 7.5 th/s doesn't seem like such a bargain.

Anyone have a sick or dead s9 board they would sacrifice for the reverse engineering gods?  There is some chance it would be brought back to life after pulling a few chips and tracing, but no promises.

I've recently reached out myself (12/20) and can corroborate this. I was quoted "We can offer our 75GH/s chips at USD8.15 per chip, with a minimum order value of USD2M."

Also offered was an Integrator Pack:

"The integrator pack consists of 1 full B8 server, 100 chips and 15 control chips. Price for the integrator pack is USD7500."

I asked further questions regarding delivery, pickup, and timelines:

"Due to current market demand it can take up to 6/7 months"

"The integrator pack can usually be delivered within 25 days"

"Delivery time on the B8 at the moment can be up to 6/7 months"

"All products are sold ex works, so will require pick up. The integrator pack will need to be picked up in Amsterdam, the subsequent chip order in korea"

To purchase any of this, it all needs to be a part of the $2M order, which is a bummer since now that BitMain raised the price of their miners in the last batch the $/TH is in their(BitFury's) favor.

I would bet it's not much different from previous generations. We used the BM1385's datasheet's comm specs to write a BM1384 driver which previously had been adapted from BM1382 code. Seems likely they wouldn't start over from scratch when four generations have already proven it works.

I might have torn into an S9 by now if they hadn't stayed balls expensive for the last year.

Sure, lots of suppliers claim to have access, but if you press the story is that they will try to get the chip if you prepay for 10K units using wire transfer.  Tread carefully, ali does not have protection like aliexpress.

I have a quantity of s9 that were purchased around $1385 and $1425 USD.  I could disassemble them and get several thousand chips at roughly $7/unit. But that's not a good deal, of course. 

Let's see if anything pans out.  Has anyone RE'd the s9 to the point that the comms format is known?

https://m.alibaba.com/product/60722789277/BM1387-ASIC-Chip-Bitcoin-Miner-Chip.html?spm=a2706.7843667.1998817009.1.3x97EL#show_specifications

This is the link

Ali is a notorious liars den, but who knows.  Just like the sales page here, let's see if any who claim to hold the chip can produce a photo of the goods. 

Some seller in china and indonesia sell this chip about 7 to 9 $
I think it is expensive and big problem is NO DATASHEET
What do u think about this?
With no data sheet it seems very dificault to do

https://ibb.co/mVCoAR
https://ibb.co/ewpJAR

A 12" wafer gives you about 65kmm2 after accounting for edge die and defects.  My die size estimate is 8mm2, or about 8500 dpw.  A standard lot is 25w, or roughly 215K units.

This would suggest that bitfury is probably selling one wafer lot for $2M, and suggests my estimates are conservative or their yield is optimistic.   Their price is quite ridiculous as the wafer cost is probably $5K/w, although that is my own guess as I don't work in that node and don't know the latest prices.

My guess is that they throw this number out to weed out punters and fish for suckers.  

That's certainly interesting.

Where did you hear that?
Link?

Hi all
It seems bitmain sell bm1387 chips
About 7 $ for 10k pieces
As mentioned above if any body wants to buy i am in

How many chips per-wafer mainly depends on 2 things: The wafer size and the die size. A decent write up on this is found at Wikipedia https://en.wikipedia.org/wiki/Wafer_(electronics) I would hazard that TSMC and Samsung are both using 11 or 12 inch (300mm) wafers. What die sizes that Bitemain, Canaan and, BitFury use so far is a secret... Possibly 2x2 or 4x4mm?

As for spec testing for performance binning, safe bet BitFury will only guarantee a minimum range of speed/Vcore that 'work's' along with quote-unquote Average performance expected from any given lot of chips. Given that even the Avalon chips constantly tweak their speeds when running I doubt any single narrow range of specs can be given. The 14/16nm node seems still just too unpredictable, at least for hot-running miner chips.

An interesting question: Over what time frame would BitFury supply your roughly 245,000 chips if you were to cut them a check for $2M USD? I can understand why you wouldn't have pursued it, but I would think that 245K chips would be a pretty good sized run at a Fab facility, no? I also wonder what level of testing/screening has been done for those parts? I am hardly an ASIC guy, but I have seen the results of poorly screened processors, before the processor vendor has figured out what has to be screened for. Does the recipient of the parts have to speed sort (aka "bin") them?

Maybe NotFuzzy would care to speculate on this (keeping his flame on "low")?

Even *if* does one get hold of chips you then still have to write a driver for them. It is not just a matter of talking to them via I2C or SPI. You need to format the data and establish the work protocols the chips use.

Ever since the s7 Bitmain has stopped providing any of that information. If you do manage to get chips from BitFury only then will they provide the needed resources to write a driver.

I reached out to Bitfury recently.  They quoted 8.15 U.S. a chip and minimum buy of 2 million dollars. not realistic for the hobby engineer.....

BF says $7/chip and $2M order.
I have connections to make a buy like this. BF salesguy said: "I've forwarded your info to my colleague and he will be in touch shortly".
It has been 3 weeks since then...
Hmmm
Maybe BF is not very keen on selling their 16nm chips?

Even though it's a very expensive process to build a batch of chips, since creating masks require cleanrooms for starters...
It's not feasable for a startup company, so why bother

If this continues, I doubt they will grow (as fast as competitor) except keep luring deep pocket customer to acquire their customized mining chips/system.

For 2 mill $, chip of similar efficiency than those in S9, with chances of design failure to get it working as a system  / not working at all versus S9 which is already in existing market, well known (good & not so good experiences). It's bleak future for those chips.

Technology must be accessible by means of cost and availability. If it's too expansive, very less exposure to it's technology, least likely it would grow.

I got the same reply from BF, $2M min order.

I have the ability and the contacts to produce the PCB and the machine, but the budget is far beyond my disposable cash.

Anyone who is serious, get in touch.  I'm wondering if we can get 50 units of the BF chip from some other buyer to debug the design.  Once that's done, there is much less risk is going to mass production.

As for starting a fab, no that's not possible.  I have been designing IC's for 15 years, that is not feasible.  Better to buy a used fab from a company going fabless.  BTW that's also not feasible, but way better than trying to build one.

Hello. We here in Russia also want to work on new machines, Bitfury B8, 5200 usd but minimum order 2 mln usd, now I put the machine on the boards, Bitfury with Georgia and Iceland on old 28 nm chips and consumes 6.5 KW of electricity, but you can disperse 20 to 28 Th, they fails there change to a new 16 nm. But how to buy nobody sells that sold me the grey party is not clear how and where brought. Maybe someone knows where to get and then I have a lot of energy of 1 MW to 4 rubles, and if to do with gas 1.26 ruble

I am seriously thinking about doing a group-buy of chips and making our own boxes. I'm seriously sick of being held at the neck by BITMAIN.

Who's interested? I think we should be able to group together $4mm (in a pseudo serious tone) 

He probably meant design and have produced a batch of chips, not build a factory from scratch.

Leastways I hope that's what he meant.

All would change is knocking off a couple zeroes from the campaign - still pretty unattainable.

Would make for a record-breaking GoFundMe....
Objective: Build stete-of-the-art IC fab for producing 16/14nm node Crypto coin mining ASIC's
Target to be Raised: $4,000,000,000
Time-frame for building fab after raising capital: 4 years.

Do you have ANY idea what it takes to produce integrated circuits?

From what I understand you can't buy the Bitmain chips direct.  You'd need to order miners with BCH and break them out.  I suppose there is a chance if the order is big enough, but considering demand far outweighs supply I wouldn't expect a deviation without paying a sizable premium.

maybe we can crowdfund our own fab.

If it's possible to get the chip, then is there anyone who can complete the puzzle to get the chip as working hardware ?

Because ASIC chips are not cheap, that's it. Once I've contacted a manufacturer and they said don't write us if you don't have 4M dollars

Only Bitfury sells chips along that companies. Minimum order is $500k and $7 per chip.
They were talking about sample orders but since last year there's no update yet. I think it's better to go their HQ and check.

You will have 71.4xx chips in total each can be o/c'ed upto 185 GH/s.

Not a bad idea, this is what I am thinking since a few months but I don't have that much capital yet.

Hmm  I guess the circuitry isn't that complicated when I looked at my Antminer S9. Why does it take mega bucks $$ to do this?

Considering it would take months to build prototypes with the naked chips and all sorts of other things it makes no sense whatsoever since bitmains shipping time is normally under 60 days from the time of order.

Why instead of having a modicum of patience are you wanting to tank your entire business with some poorly thought out scheme to build your own stuff?

I believe Bitfury's minimum order size is $1m. Bitmain doesn't sell their naked chips - even if they did, since there is wait for their units, why would they sell a key component so someone can beat them to market?

Been tried, didn't work then, probably won't now unless you gots mega$$ and inside contacts!

Does anyone know how to buy bitmain or bitfury chips directly? I am thinking about doing a group-buy of custom rigs. The Bitmain wait is absolutely insane.

What do people think about this idea?