Topic: Bitfury: "16nm... sales to public start shortly" - page 36. (Read 108588 times)

Because people really like being right, and being right matters more than the stuff they're right about.

You're stating the ticker sale price, not the cost price. If the containers cost $8m for Bitfury, what would be the point selling them at $8m?


Why is everyone so hostile tonight

I said the container costs $8 million - I know they own Allied but that doesn't change the fact it costs that much.  Show me an actual picture of one of them, you can't because there is not one to take pictures of yet (at least not an "in use" one. I am sure they do not own the Shipping Container Manufacturers, the parts manufacturers for a 1.5 - 2 megawatt outfit that fits into the container, the Novec etc etc.  I was not guessing at the cost of the container, I was stating what it cost. 

Dogie, I know it makes you upset that someone has info that you do not have and you feel the need to try and shame that person. Sorry you feel the need to do that.

Since they bought Allied control I doubt they're still charging themselves $8m The thing is, they have the data and they've decided its the best choice so its relatively safe to assume that is the best choice. And if not then it doesn't affect us anyway. Also worth considering that its a lot easier to ship bare chips ready for OEM integration than it is to resell weird form factor, second hand boards.


Are you seriously that unable to have an adult debate? Calm down already.

Look I don't hate Bitfury  but bitmaintech has hurt them with the flood the network with hash move.

I think Bitfury has assessed they got fucked real hard by what bitmaintech did.

Looking at this from my viewpoint bitfury is scrambling to survive the move put on them by bitmaintech.

Remember none of us know just what bitmaintech can do in terms of hash rate.

If they can do 12% growth to July .  I simply don't know if bitfury has an answer to this.

As  small guy staying in the 10-15k usd self imposed limits for playing with these coins it has become very interesting to see just what bitmaintech has done

Yes, this is EXACTLY what most do to optimize in a variety of ways, including their current situation, but arguing with a mutt just drags you to lay in the floor, get fleas, and who knows what diseases.
There are still places buying S3s for decent prices in bulk.

The simple fact remains they are not mining with any significant chips (28 or 16nm) in the immersion data tanks.  Those tanks are in the $8 million range and without high density 16nm chips those Immersion Data Centers are worthless.  Filling up an $8 million immersion data tank with 28nm chips that are not packed densely is ludicrous, even filling it with 16nm chips it will be tough to turn a profit when you need to overcome $8 million

That only applies when the newer generation absolutely demolishes the old one in efficiency and when electricity costs are tangible. They are not and neither is Bitfury's 28nm obsolete.

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.

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.

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.

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.

By what punin told us apparently it isn't )



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

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?

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?

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.

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

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.

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.
 

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

They will sell the used 28nm miners?