Bitfury: "16nm... sales to public start shortly"

[RoadStress]

Just my take on it but: Because one has to start *somewhere* and get the ball rolling. The older chips were on-hand and they have working boards for them. As a bonus they get to test the DC at full load.

Newer more efficient chips/boards are in the pipeline but not on-hand so you either have a DC sitting there empty/partly used or ya plug in what you have right now and upgrade when the new designs get into full production.

Sure, but doing that needs money and the difficulty is raising like crazy. They will never make ROI on the "miners" so that means that they have money to burn just like I said. I understand that they need to get the ball rolling and that they need to test the DC, but if the 3rd party manufacturers can get miners shipped in March then they can get 16nm miners in the DC too. There is simply no time to make ROI on this 28nm miners.

[NotFuzzyWarm]

Then it becomes a write-off. Hell, stuffing the DC with 28nm chips could be part of their R&D budget...

[HyperMega]

Just my take on it but: Because one has to start *somewhere* and get the ball rolling. The older chips were on-hand and they have working boards for them. As a bonus they get to test the DC at full load.

Newer more efficient chips/boards are in the pipeline but not on-hand so you either have a DC sitting there empty/partly used or ya plug in what you have right now and upgrade when the new designs get into full production.

Sure, but doing that needs money and the difficulty is raising like crazy. They will never make ROI on the "miners" so that means that they have money to burn just like I said. I understand that they need to get the ball rolling and that they need to test the DC, but if the 3rd party manufacturers can get miners shipped in March then they can get 16nm miners in the DC too. There is simply no time to make ROI on this 28nm miners.

Please correct me, if I'm wrong.

Let's assume they
1) produce a 28nm miner for $0.06/GH
2) operate it at 0.36 J/GH (conservativ guess)
3) while having electricity costs of $0.05/kWh (or less)

they should make $0.03/GH per month at the current difficulty, which means ROI in about 2 months.

There will be probably first miners based on Bitfury's 16nm ASIC in wild and their DC end of March, but I guess the real volume ramp up will happen in April.
So it should be enough time, to make some profit with the new 28nm hardware.

[baazju]

Just my take on it but: Because one has to start *somewhere* and get the ball rolling. The older chips were on-hand and they have working boards for them. As a bonus they get to test the DC at full load.

Newer more efficient chips/boards are in the pipeline but not on-hand so you either have a DC sitting there empty/partly used or ya plug in what you have right now and upgrade when the new designs get into full production.

Sure, but doing that needs money and the difficulty is raising like crazy. They will never make ROI on the "miners" so that means that they have money to burn just like I said. I understand that they need to get the ball rolling and that they need to test the DC, but if the 3rd party manufacturers can get miners shipped in March then they can get 16nm miners in the DC too. There is simply no time to make ROI on this 28nm miners.

Please correct me, if I'm wrong.

Let's assume they
1) produce a 28nm miner for $0.06/GH
2) operate it at 0.36 J/GH (conservativ guess)
3) while having electricity costs of $0.05/kWh (or less)

they should make $0.03/GH per month at the current difficulty, which means ROI in about 2 months.

There will be probably first miners based on Bitfury's 16nm ASIC in wild and their DC end of March, but I guess the real volume ramp up will happen in April.
So it should be enough time, to make some profit with the new 28nm hardware.

If you think there will be a 5TH/s miner being sold for $300 within the next week then maybe you can hit ROI with those numbers. Doubt that will ever happen. It needs to be ready to hit the market at half the price of an S7... today.

[RoadStress]

Please correct me, if I'm wrong.

Let's assume they
1) produce a 28nm miner for $0.06/GH
2) operate it at 0.36 J/GH (conservativ guess)
3) while having electricity costs of $0.05/kWh (or less)

they should make $0.03/GH per month at the current difficulty, which means ROI in about 2 months.

There will be probably first miners based on Bitfury's 16nm ASIC in wild and their DC end of March, but I guess the real volume ramp up will happen in April.
So it should be enough time, to make some profit with the new 28nm hardware.

I don't see how it would be possible to produce full working miners meaning PCB+chips+components at $0.06/GH. Even if this were true, which I find very unlikely, the big private buyers are being ripped off big time while we are being lied that BitFury is trying to help the decentralization.

[goxed]

at least pin out ?

Did you look at the schematics we just posted? 2 pins for data (SCK&SDATA) and all the rest for power.

Actually that is bullshit. Bitfury_POWER has around 40 pins, most of them non-obvious.

Will you actually dare to disclose what the PARITY pin does and what is the function of the two associated LEDs?

Actually you're wrong. The 28nm chip has VCC, GND and two I/O pins as you can clearly see from images like the 1block_short.jpg for example. The 250nm is another story. It's a comm chip with 3 dedicated lines for each 28nm chip, reference voltage, clock and data + bunch of other stuff. Calling bullshit when you didn't even comprehend the question that I answered is a bit harsh.

How do the COMM chips connect to the external world? Will the 16nm Bitfury chip be available in QFN package?

[NotFuzzyWarm]

^^ The possibility of the package size and same limited coms setup makes for an interesting thought: Perhaps aside from some changes to the support chips to handle higher throoughput, the new chips could basically use the existing hashboards or at least most of their already validated portions perhaps? Hmmm..

That would certainly speed deployment in BitFury's almost mobile mini-pharms http://bitfury.com/products#container-datacenter and the new DC... Once more than just engineering samples are produced, tweaked hashboards tested and the foundry is given the the green light to start a full scale run(s) it's then just a matter of changing tankfulls of boards one at a time. If the data tanks are removable as a unit, all the better. prep/change over tanks to the new 16nm chips & boards  off line and swap out the old ones with the new. Rinse & repeat as desired & production (and money) allows for.

Should give their selected integrators a leg up as well I'd think.

A last thought on why deploy 28nm in the new DC: Time to produce enough boards/chips to fill it. The building & all infra itself was completed how long ago? (seriously, when?) With 28nm chips in production at the time or left over from the final runs, why not? Is a good CYA to start filling/using the DC until the 16nm chips start getting made en-mass which is, um, now if I've read things right.

Query to BitFury: If I may ask, how many boards & chips are in each tank? Based on the stated 100GH/s a 12Ph mini-farm takes 120k chips. What size wafer is used? What is the actual silicon die size (along with full package size?

[dogie]

Its then just a matter of changing tankfulls of boards one at a time.

Why would they remove 28nm stuff when its already paid for and still profitable? The answer I guess is capacity and just how many data centers you want to build, but that's part of the reason they want to sell these to others.

[carlosmnk]

Its then just a matter of changing tankfulls of boards one at a time.

Why would they remove 28nm stuff when its already paid for and still profitable? The answer I guess is capacity and just how many data centers you want to build, but that's part of the reason they want to sell these to others.

They will sell the used 28nm miners?

[dogie]

Its then just a matter of changing tankfulls of boards one at a time.

Why would they remove 28nm stuff when its already paid for and still profitable? The answer I guess is capacity and just how many data centers you want to build, but that's part of the reason they want to sell these to others.

They will sell the used 28nm miners?

If they're in immersion cooling board form, they're certainly not going to be selling them.

[2112]

If they're in immersion cooling board form, they're certainly not going to be selling them.

Why? The whole point of immersion cooling is that it allows using the same boards design as for air cooling. The difference is only in how far the board can be overvolted/overclocked for optimum GH/$. Theoretically the buyer would air cool them to optimize GH/J, which means undervolting/underclocking.
 

[HyperMega]

Please correct me, if I'm wrong.

Let's assume they
1) produce a 28nm miner for $0.06/GH
2) operate it at 0.36 J/GH (conservativ guess)
3) while having electricity costs of $0.05/kWh (or less)

they should make $0.03/GH per month at the current difficulty, which means ROI in about 2 months.

There will be probably first miners based on Bitfury's 16nm ASIC in wild and their DC end of March, but I guess the real volume ramp up will happen in April.
So it should be enough time, to make some profit with the new 28nm hardware.

I don't see how it would be possible to produce full working miners meaning PCB+chips+components at $0.06/GH. Even if this were true, which I find very unlikely, the big private buyers are being ripped off big time while we are being lied that BitFury is trying to help the decentralization.

Sorry, but I think this is the sad truth.

Bronto estimated earlier the pure costs of their packaged and tested 16nm ASIC are about $0.03/GH (at 100 GH/chip) and I bet the costs of their 28nm generation ASIC are not much higher, $0.05/GH should be a good guess.
This means, if they produce a 10 TH/s miner they spend $500 for ASICs and would have $100 left for the PCB and remaining components to end up at $0.06/GH.
It's a string design probably with a 2 layer PCB! You even don't have to negogiate hard with the suppliers to stay below $100.

I have to admit, I ignored the PSU. Including that their production costs should be more in the range of $0.08/GH.

[dogie]

If they're in immersion cooling board form, they're certainly not going to be selling them.

Why? The whole point of immersion cooling is that it allows using the same boards design as for air cooling. The difference is only in how far the board can be overvolted/overclocked for optimum GH/$. Theoretically the buyer would air cool them to optimize GH/J, which means undervolting/underclocking.
 

Because that's even more cost to repurpose hardware that is perfectly fine, already paid for, already mining and still profitable.

[2112]

Because that's even more cost to repurpose hardware that is perfectly fine, already paid for, already mining and still profitable.

Oh, c'mon. It would be a waste of immersion bath to run old underclocked miners.

And what it means "paid for"? Presumably Bitfury used either its own or borrowed capital to populate their immersion cooled facility. Selling them would actually release working capital to spend it on manufacturing new generation miners.

[dogie]

Because that's even more cost to repurpose hardware that is perfectly fine, already paid for, already mining and still profitable.

Oh, c'mon. It would be a waste of immersion bath to run old underclocked miners.

And what it means "paid for"? Presumably Bitfury used either its own or borrowed capital to populate their immersion cooled facility. Selling them would actually release working capital to spend it on manufacturing new generation miners.

So they're going to spend money to buy 16nm, spend money to retrofit 28nm, spend money to sell 28nm, spend money to install 16nm..... and that's all going to cost them less than just selling 16nm?

[jstefanop]

If they're in immersion cooling board form, they're certainly not going to be selling them.

Why? The whole point of immersion cooling is that it allows using the same boards design as for air cooling. The difference is only in how far the board can be overvolted/overclocked for optimum GH/$. Theoretically the buyer would air cool them to optimize GH/J, which means undervolting/underclocking.
 

Not quite...board density and ASIC spacing for Air cooled boards are big limiting factors. Sure you could slap a huge heatsink on an immersion board, but whats the point when you can only run the ASICS at 50% because of thermal constraints?

[RoadStress]

Oh, c'mon. It would be a waste of immersion bath to run old underclocked miners.

By what punin told us apparently it isn't )

And what it means "paid for"? Presumably Bitfury used either its own or borrowed capital to populate their immersion cooled facility. Selling them would actually release working capital to spend it on manufacturing new generation miners.

I think that by "paid for" he means that they have so much money that they can afford to burn some...

[2112]

Not quite...board density and ASIC spacing for Air cooled boards are big limiting factors. Sure you could slap a huge heatsink on an immersion board, but whats the point when you can only run the ASICS at 50% because of thermal constraints?

Not really. This is well known since the days of GPU mining. Each board has two optimum operating points:

1) max GH per board, overvolted and overclocked, thermally limited

2) max GH/J per board, undervolted and underclocked, simultaneous-switching-noise limited

One quick look at the photo of the 28nm board shows that there's plenty of open space and wide copper traces to allow air cooling without the need to install heathsinks. When compared to rather dense Spondoolies' boards this is quite obvious that Bitfury's design has much wider thermal margins.

With those board it looks like point (1) is achieved in the immersion bath and point (2) with forced air cooling but without heathsinks.

So they're going to spend money to buy 16nm, spend money to retrofit 28nm, spend money to sell 28nm, spend money to install 16nm..... and that's all going to cost them less than just selling 16nm?

This is just nonsense. In particular "retrofit of 28nm boards". Those boards obviously don't need retrofitting (immersion to forced air) because of rather low power density. The opposite would be true for air to immersion move: remove heathsinks and fans, extraordinary deep cleaning of flux residue and lacquer layers, possible need to replace the components that leach something in the Novec bath. The only retrofitting required for immersion to air move is a trivial reprogramming of the voltage and clock operating points in the controller.

I'm positive that nearly every MBA-level course has some examples of how to optimize the utilization of the working capital. Usually it is done with some historical examples like a move from steam power to diesel power in railroad locomotives. I'm not going to reproduce it here.

[dogie]

stuff

No, throwing out or selling 28nm stuff to replace it with 16nm simply doesn't make sense. They have a limited amount of hosting capacity so its 16nm OR 28nm and the other has to be sold. While they can get a better capital $ / GH with replacing the 28nm with 16nm, minimizing capital expenditure isn't always the sensible or most profitable thing to do - especially in a company as cash rich as Bitfury.

[2112]

No, throwing out or selling 28nm stuff to replace it with 16nm simply doesn't make sense. They have a limited amount of hosting capacity so its 16nm OR 28nm and the other has to be sold. While they can get a better capital $ / GH with replacing the 28nm with 16nm, minimizing capital expenditure isn't always the sensible or most profitable thing to do - especially in a company as cash rich as Bitfury.

Really, it isn't my problem that you can't understand this. Even garage miners understand that selling the old miners on eBay (or to Venezuela) and replacing them with new ones is part of the ROI optimization strategy. I think ASICMINER was the first who did that on industrial scale.