Universal Hash Board: Bitfury 28 nM "UHB_BF28NM"

2112

legendary

Activity: 2128

Merit: 1073

Quote from: allten on August 01, 2016, 01:56:55 PM

Glad you like the "art". I've corrected the spelling on our internal files for the next revision. Thanks.

The n-channel device Q5 is a unidirectional High-to-Low Level translator. The High signal is on the left.
Took me a while to grasp how it works. Admittedly, I've never used it in practice; It was referenced from
Bitfury's design files that they publicly shared.

I'm not really buying the "it does work" explanation. This is a https://en.wikipedia.org/wiki/Common_gate topology, the equivalent of https://en.wikipedia.org/wiki/Common_base for BJTs and common-grid for vacuum tubes.

The technically valid applications of this topology are extremely rare. By pure probability the more likely explanations are non-technical:

1) designer went mad
2) unintentional mistake in the schematics
3) intentional mistake in the schematics (to confuse copiers or to track the copiers)
4) some other things that I can't think of right now

All of them are more probable than a valid technical reason.

philipma1957

legendary

Activity: 4256

Merit: 8551

'The right to privacy matters'

Quote from: NotFuzzyWarm on August 01, 2016, 07:10:29 PM

Quote from: Unacceptable on August 01, 2016, 06:40:24 PM

Quote from: allten on August 01, 2016, 02:13:43 PM

Quote from: Unacceptable on July 08, 2016, 01:50:21 AM

Very cool...but a little late...no chips will be sold to anyone by any manufacturer to make a miner for the general public Sad

I totally understand the pessimism. I've been watching Bitcoin since the GPU days and all the drama that has unfolded.
IMO, if it doesn't change (ASIC companies not selling their chips) then they are going to be "holding the bag" and it won't be worth much.
If they don't embrace us "little guys" sooner rather than later then I truly believe POW blockchains utilizing ASICs are doomed.

We've chosen to be optimistic. We've chosen to work hard and share our efforts as much as possible with a bright hope of ASIC manufactures
embracing us and working closely with us.
We desire to become a 3rd party hardware manufacture that will also empower this community and give it a voice that cannot be ignored.

All well & good,except...there are only 2 or 3 ASIC companies left,Avalon & Bitmain & BW(maybe Bitfury,I think they are dead or dying).

They will not sell chips to anyone (in any meaningful amount),cause.....money,they make MORE profit keeping their chips & manufacturing in their CONTROL Roll Eyes

Not to mention the self mining they are doing to increase profit even further.

So,unless YOU make YOUR OWN chips,YOU will have nothing to build or sell to us "little guys"............& your "voice" will be basically ignored,they'll be nice about it though I'm sure.

Bitcoin has been in the commercial phase for a few years now & will get even more commercialized as the the few with chips & miners consolidate the hashrate in their favor.

Best of luck,just don't hold your breath!!!! Grin

On BitFury, ja it is looking like their chip making divisions foray into 16nm will not make it to the Public. But BF as a whole... They do a helluva lot of other things including blockchain analytics so bottom line is probably still pretty good for them. The losses from 16nm chip making and of course related to that, any containerized farms they sold based on it will just be write off's to them and backers/investors.

As for BTC being a commercial phase and becoming more so -- Satoshi or whoever came up with with it designed that in from day-1. What it ultimately boils down to is that BTC block-found payout halving will always happen until finally all income is derived solely from Tx fees vs miner's finding blocks.

Anyone looked at what the typical tx fees per block are currently? Hint: Not much vs the 12btc miners bounties... On CK pool typical best additional payout from Tx fee's that CK distributes is around 1BTC.

all my economic calculations say this will need to be addressed the issue is there is a hard transaction cap of about 700,000 per day . It does not happen since every transaction would need to be 227 bytes to get that high.

the whole network does 300,000 on average each day.

there are several ways or methods that could fix this issue .

I am hoping it to be a fusion between BTC and ETC-ETH actually any gpu non basic coin will due.

the idea is the 300000 average could be slowly increased for BTC no need to rush since ETH/ETC can do a real lot of transactions each day.

I am watching to see this idea or another come about before the next 2 ½ings occur.

NotFuzzyWarm

legendary

Activity: 3822

Merit: 2703

Evil beware: We have waffles!

Quote from: Unacceptable on August 01, 2016, 06:40:24 PM

Quote from: allten on August 01, 2016, 02:13:43 PM

Quote from: Unacceptable on July 08, 2016, 01:50:21 AM

Very cool...but a little late...no chips will be sold to anyone by any manufacturer to make a miner for the general public Sad

I totally understand the pessimism. I've been watching Bitcoin since the GPU days and all the drama that has unfolded.
IMO, if it doesn't change (ASIC companies not selling their chips) then they are going to be "holding the bag" and it won't be worth much.
If they don't embrace us "little guys" sooner rather than later then I truly believe POW blockchains utilizing ASICs are doomed.

We've chosen to be optimistic. We've chosen to work hard and share our efforts as much as possible with a bright hope of ASIC manufactures
embracing us and working closely with us.
We desire to become a 3rd party hardware manufacture that will also empower this community and give it a voice that cannot be ignored.

All well & good,except...there are only 2 or 3 ASIC companies left,Avalon & Bitmain & BW(maybe Bitfury,I think they are dead or dying).

They will not sell chips to anyone (in any meaningful amount),cause.....money,they make MORE profit keeping their chips & manufacturing in their CONTROL Roll Eyes

Not to mention the self mining they are doing to increase profit even further.

So,unless YOU make YOUR OWN chips,YOU will have nothing to build or sell to us "little guys"............& your "voice" will be basically ignored,they'll be nice about it though I'm sure.

Bitcoin has been in the commercial phase for a few years now & will get even more commercialized as the the few with chips & miners consolidate the hashrate in their favor.

Best of luck,just don't hold your breath!!!! Grin

On BitFury, ja it is looking like their chip making divisions foray into 16nm will not make it to the Public. But BF as a whole... They do a helluva lot of other things including blockchain analytics so bottom line is probably still pretty good for them. The losses from 16nm chip making and of course related to that, any containerized farms they sold based on it will just be write off's to them and backers/investors.

As for BTC being a commercial phase and becoming more so -- Satoshi or whoever came up with it designed that in from day-1. What it ultimately boils down to is that BTC block-found payout halving will always happen until finally all income is derived solely from Tx fees vs miner's finding blocks. The idea of course is that folks will have to pay reasonable Tx fees to use BTC -- like the CC card / banks/ PayPal. et al charge.

Anyone looked at what the typical tx fees per block are currently? Hint: Not much vs the 12.5BTC miners bounties... On CK pool typical best additional payout from Tx fee's that CK collects & distributes to us is around 1BTC.

Unacceptable

legendary

Activity: 2212

Merit: 1001

Quote from: allten on August 01, 2016, 02:13:43 PM

Quote from: Unacceptable on July 08, 2016, 01:50:21 AM

Very cool...but a little late...no chips will be sold to anyone by any manufacturer to make a miner for the general public Sad

I totally understand the pessimism. I've been watching Bitcoin since the GPU days and all the drama that has unfolded.
IMO, if it doesn't change (ASIC companies not selling their chips) then they are going to be "holding the bag" and it won't be worth much.
If they don't embrace us "little guys" sooner rather than later then I truly believe POW blockchains utilizing ASICs are doomed.

We've chosen to be optimistic. We've chosen to work hard and share our efforts as much as possible with a bright hope of ASIC manufactures
embracing us and working closely with us.
We desire to become a 3rd party hardware manufacture that will also empower this community and give it a voice that cannot be ignored.

All well & good,except...there are only 2 or 3 ASIC companies left,Avalon & Bitmain & BW(maybe Bitfury,I think they are dead or dying).

They will not sell chips to anyone (in any meaningful amount),cause.....money,they make MORE profit keeping their chips & manufacturing in their CONTROL Roll Eyes

Not to mention the self mining they are doing to increase profit even further.

So,unless YOU make YOUR OWN chips,YOU will have nothing to build or sell to us "little guys"............& your "voice" will be basically ignored,they'll be nice about it though I'm sure.

Bitcoin has been in the commercial phase for a few years now & will get even more commercialized as the the few with chips & miners consolidate the hashrate in their favor.

Best of luck,just don't hold your breath!!!! Grin

NotFuzzyWarm

legendary

Activity: 3822

Merit: 2703

Evil beware: We have waffles!

Quote from: allten on August 01, 2016, 03:25:36 PM

Quote from: dogie on July 08, 2016, 05:09:37 AM

I've heard the easily replaceable hash board before (including from Bitmain) and it never quite works out the way you think it will. Have a look between the generations and try and work out how many Bitmain generations are a. board compatible and b. heatsink compatible and that'll give you an answer as to the sustainability.

Yes. This has by far been the most challenging aspect of the design.
Varying sizes of chips and thickness.
Surrounding components.
Heat dissipation through the top of the chip or through the PCB.
Different PCB layout where chips are now in new locations.

It's no surprise where Bitmain is now: Individual Heat sinks for each ASIC.

Our solution may not be elegant, but should work.

Not ready to go into detail yet and the topic deserves its own thread.

The surrounding components is a pretty important point -- mainly the decoupling caps around the chips and the bypass used for each node when powering a string. It may look like, "well just put some (random value) caps here and..." BUT -- the actual values and more importantly, where they are located and what type they are (material used) is of vital importance. You use different ASIC's, you will need to find a new set of optimum values for the caps or at a minimum, verify how the different ASIC's work with a compromise set of values.

In many ways the power planes feeding ASIC's are much like RF circutry and must be treated as such or else you get some mighty weird and very localized dips/spikes in the 'DC' power feed to them.
ref: http://powerelectronics.com/power-electronics-systems/five-things-every-engineer-should-know-about-pdn

sidehack

legendary

Activity: 3374

Merit: 1859

Curmudgeonly hardware guy

Why is that necessary? Are you looking for signalling specifics for the different chips?

allten

sr. member

Activity: 455

Merit: 250

You Don't Bitcoin 'till You Mint Coin

Please Help. We need some more hardware to reverse engineer and make sure this standard is truly universal.

Looking for broken S7 and S9 Hash boards and controllers.
Looking for a broken Avalon 6 hash board and controller.

If you have any sitting around collecting dust, please donate soon!

THANKS!!

allten

sr. member

Activity: 455

Merit: 250

You Don't Bitcoin 'till You Mint Coin

Quote from: dogie on July 08, 2016, 05:09:37 AM

I've heard the easily replaceable hash board before (including from Bitmain) and it never quite works out the way you think it will. Have a look between the generations and try and work out how many Bitmain generations are a. board compatible and b. heatsink compatible and that'll give you an answer as to the sustainability.

Yes. This has by far been the most challenging aspect of the design.
Varying sizes of chips and thickness.
Surrounding components.
Heat dissipation through the top of the chip or through the PCB.
Different PCB layout where chips are now in new locations.

It's no surprise where Bitmain is now: Individual Heat sinks for each ASIC.

Our solution may not be elegant, but should work.

Not ready to go into detail yet and the topic deserves its own thread.

allten

sr. member

Activity: 455

Merit: 250

You Don't Bitcoin 'till You Mint Coin

Quote from: Unacceptable on July 08, 2016, 01:50:21 AM

Very cool...but a little late...no chips will be sold to anyone by any manufacturer to make a miner for the general public Sad

I totally understand the pessimism. I've been watching Bitcoin since the GPU days and all the drama that has unfolded.
IMO, if it doesn't change (ASIC companies not selling their chips) then they are going to be "holding the bag" and it won't be worth much.
If they don't embrace us "little guys" sooner rather than later then I truly believe POW blockchains utilizing ASICs are doomed.

We've chosen to be optimistic. We've chosen to work hard and share our efforts as much as possible with a bright hope of ASIC manufactures
embracing us and working closely with us.
We desire to become a 3rd party hardware manufacture that will also empower this community and give it a voice that cannot be ignored.

allten

sr. member

Activity: 455

Merit: 250

You Don't Bitcoin 'till You Mint Coin

Quote from: 2112 on July 08, 2016, 08:19:45 AM

The schematics look like art project, not a technical project. It seems like the preparer wasn't understanding what he was doing or for other reasons made strange mistakes. E.g. there is an n-channel device Q5 with gate tied directly to +3.3V.

I've seen electronic schematics prepared around 1960-1970 by a school dropout who was dictated by an engineer who was injured and blinded in an accident. They had similar completely incomprehensible mistakes. E.g. no sane technician or engineer would consistently spell "nM" for "nanometer".

It could also be a joke, because "Bitfury" is consistently spelled "Bitfurry".

Glad you like the "art". I've corrected the spelling on our internal files for the next revision. Thanks.

The n-channel device Q5 is a unidirectional High-to-Low Level translator. The High signal is on the left.
Took me a while to grasp how it works. Admittedly, I've never used it in practice; It was referenced from
Bitfury's design files that they publicly shared.

allten

sr. member

Activity: 455

Merit: 250

You Don't Bitcoin 'till You Mint Coin

Quote from: klondike_bar on July 08, 2016, 10:19:42 AM

4 fans for only a single hashing board? catch up to the times

someting like this needs to be capable of >2KW and be easy to swap boards in. Look at the antminer S2 or the first-gen bitfury 56nm as an example of what it should look like (and then realize that desipte that, both companies never released compatible upgrade modules)

The max Wattage in this standard is at 500 Watts.
The wattage range is much more reasonable for new home users; people who want to investigate the technology.
We still believe that bitcoin mining is the #1 way to introduce this technology to the masses. This is our motivation.

The four fan design as seen in the OP teaser photo offers the ability to cool while being very quiet.
Again, catered towards a reasonable home mining experience for newbies. The design is a large step
from the obsolesce seen so far. All the surrounding infrastructure will survive several future generations.

allten

sr. member

Activity: 455

Merit: 250

You Don't Bitcoin 'till You Mint Coin

Quote from: sidehack on July 08, 2016, 08:30:58 AM

I haven't looked over schematics (not downloading on this browser, I'll check it at the shop), but on that breakout image I don't see any provision for regulation; couple that with the need to be "stringable" and I assume the design is fixed core voltage? In which case the designer needs to be reminded that home users don't really appreciate planned obsolescence, therefore "no thank you".

There is no regulator on the UHB boards in the standard that is being created. The regulator is moved to the Control Board.
There is no "planned obsolescence", but just the opposite. Thanks for taking a look.

klondike_bar

legendary

Activity: 2128

Merit: 1005

ASIC Wannabe

Quote from: RadekG on July 10, 2016, 06:19:27 AM

Here are my simple points why universal PCB can't work.
1) different power density. Just compare KnC or Spondoolies chips with BF, Bitmain or others. Two different ways of power density. SFARDS just in the middle. Different power density needs different PCB design and different heatsink design.

3) total output power. It is based on power density. We can see extremely increasing power density with the same case size in Bitmain's products beginning from Antminer S1 to Antminer S7 or S9. One single heatsink can't handle 700W on multiple chips. Even 300W is too much as we seen on S5 failure rate.

I dont think manufacturers will standardize a design anytime soon - by offering a full product their margins are probably improved

as for power density/draw, a universal heatsink isnt ideal but it is pretty sufficient. simply upgrading the fan/fanspeed could allow a +/- 50% power capacity.

The bigger issue is board layout, where all components were set to a specific maximum thickness in heatsinked areas (so that only critical parts (ASIC, regulators) are snug to heatsink and other components are not), and this could prove inconvenient to some layouts

2112

legendary

Activity: 2128

Merit: 1073

Quote from: RadekG on July 10, 2016, 05:55:51 AM

I do not understand where you see the problem with Q5. Please explain us what's wrong. BSS138 is logic level N-fet. Gate is connected to 3.3V and input goes to source. G+S resistor keeps Q5 default off. If input is H Q5's G+S=0V, Q5 is off. If input is L, G+S=3.3V, Q5 is on.

So what is the point? It is neither inverter nor amplifier. To convert push-pull output to open-drain output? Connecting that open-drain in series with the push-pull output? What for?

Also Q1-Q4 are two inverter pairs. They legitimately convert single-ended output to the balanced outputs. But they only have active pull-down. The pull-ups are resistors. Why using resistors instead of normal CMOS inverters with both pull-down and pull-up active?

I see signal integrity issues with all of those five elements. And they seem destined for off-board routing through a connector. Is that some sort of weird ESD protection by using cheap discrete transistors as sacrificial to protect rare/expensive integrated circuits?

Edit: spelling and US/English terminology fixes

RadekG

hero member

Activity: 924

Merit: 500

Quote from: dogie on July 08, 2016, 05:09:37 AM

I've heard the easily replaceable hash board before (including from Bitmain) and it never quite works out the way you think it will. Have a look between the generations and try and work out how many Bitmain generations are a. board compatible and b. heatsink compatible and that'll give you an answer as to the sustainability.

I think there is a good reason to "standardize" hash boards, but it needs a lot of standards to be followed by all manufacturers. Here are my simple points why universal PCB can't work.

1) different power density. Just compare KnC or Spondoolies chips with BF, Bitmain or others. Two different ways of power density. SFARDS just in the middle. Different power density needs different PCB design and different heatsink design.

2) string design. It is already required, but not all designs are so simple. Bitmain S7 and S9 are hybrid string designs which needs separate buck controller (not necessary when changing design). BF250NM is already prepared for string design, but it is not so easy with BM1384.

3) total output power. It is based on power density. We can see extremely increasing power density with the same case size in Bitmain's products beginning from Antminer S1 to Antminer S7 or S9. One single heatsink can't handle 700W on multiple chips. Even 300W is too much as we seen on S5 failure rate.

It can be win for customer, but it makes so many design rules for manufacturers that it is nearly impossible to implant any standards.

I am happy that anybody spend time thinking about it, but there is still not enough manufacurer competition to do it.

RadekG

hero member

Activity: 924

Merit: 500

Quote from: 2112 on July 08, 2016, 08:19:45 AM

The schematics look like art project, not a technical project. It seems like the preparer wasn't understanding what he was doing or for other reasons made strange mistakes. E.g. there is an n-channel device Q5 with gate tied directly to +3.3V.

I do not understand where you see the problem with Q5. Please explain us what's wrong. BSS138 is logic level N-fet. Gate is connected to 3.3V and input goes to source. G+S resistor keeps Q5 default off. If input is H Q5's G+S=0V, Q5 is off. If input is L, G+S=3.3V, Q5 is on.

klondike_bar

legendary

Activity: 2128

Merit: 1005

ASIC Wannabe

4 fans for only a single hashing board? catch up to the times

someting like this needs to be capable of >2KW and be easy to swap boards in. Look at the antminer S2 or the first-gen bitfury 56nm as an example of what it should look like (and then realize that desipte that, both companies never released compatible upgrade modules)

sidehack

legendary

Activity: 3374

Merit: 1859

Curmudgeonly hardware guy

I haven't looked over schematics (not downloading on this browser, I'll check it at the shop), but on that breakout image I don't see any provision for regulation; couple that with the need to be "stringable" and I assume the design is fixed core voltage? In which case the designer needs to be reminded that home users don't really appreciate planned obsolescence, therefore "no thank you".

2112

legendary

Activity: 2128

Merit: 1073

The schematics look like art project, not a technical project. It seems like the preparer wasn't understanding what he was doing or for other reasons made strange mistakes. E.g. there is an n-channel device Q5 with gate tied directly to +3.3V.

I've seen electronic schematics prepared around 1960-1970 by a school dropout who was dictated by an engineer who was injured and blinded in an accident. They had similar completely incomprehensible mistakes. E.g. no sane technician or engineer would consistently spell "nM" for "nanometer".

It could also be a joke, because "Bitfury" is consistently spelled "Bitfurry".

dogie

legendary

Activity: 1666

Merit: 1185

dogiecoin.com

I've heard the easily replaceable hash board before (including from Bitmain) and it never quite works out the way you think it will. Have a look between the generations and try and work out how many Bitmain generations are a. board compatible and b. heatsink compatible and that'll give you an answer as to the sustainability.

Topic: Universal Hash Board: Bitfury 28 nM "UHB_BF28NM" (Read 2326 times)