Antminer S2 Speculation thread

[Gator-hex]

The Bitfury2 is the most efficient 55nm Bitcoin chip on the planet. http://www.bitfurystrikesback.com/product/bitfury-55nm-rev2-samples/

Technobit just showed 16x of them running today at 61GH (3.8GH per chip).
No official wattage yet but the old ones were about 1.5W/GH and I'm estimating power at 1.24W/GH.
https://bitcointalk.org/index.php?topic=307897.msg5640531#msg5640531

If you built an Antminer out of these things it would be around 243GH/301W I'd buy that, it's as good as my 28nm A1 modules!

Next questions how cheap could you make it?

€5 a chip = $7 so it's $448 in 64x chip cost. Double that for other costs/profit and it could be under $900. Again in the 28nm A1 ballpark.

So it is do-able with 55nm.

[dropt]

Seems to me regardless of what process node they use they will need to use a chip with improved power consumption in the S2.  In the original U1 the chip was always under clocked.  The U2+ just took the same chip added a bigger heat sink and changed the default clock rate.  Since the S1 was already operating at close to the maximum clock rate they will almost certainly have to use an improved chip in the S2.

Why?  Just do the reverse of what they did going from U1 -> U2 and increase the chip count.

[JoRoidx]

When the Bitmain -- Antminer 1Th and 2 Th are avaible we should do a group buy> 

[MANofthePEOPLE]

I hope it wont draw 2k. If not and if it's priced well I will probably buy at least one, hopefully more. If priced at BTC3,5 probably a small farm 

[Xer0]

no, no rebranded chinese 1T please!

i never seen a bigger waste of space

[BTCInvestorPro]

Seems to me regardless of what process node they use they will need to use a chip with improved power consumption in the S2.  In the original U1 the chip was always under clocked.  The U2+ just took the same chip added a bigger heat sink and changed the default clock rate.  Since the S1 was already operating at close to the maximum clock rate they will almost certainly have to use an improved chip in the S2.

Why?  Just do the reverse of what they did going from U1 -> U2 and increase the chip count.

Because Bitmain still needs to turn a profit and adding more chips would increase their production costs.

[dropt]

Seems to me regardless of what process node they use they will need to use a chip with improved power consumption in the S2.  In the original U1 the chip was always under clocked.  The U2+ just took the same chip added a bigger heat sink and changed the default clock rate.  Since the S1 was already operating at close to the maximum clock rate they will almost certainly have to use an improved chip in the S2.

Why?  Just do the reverse of what they did going from U1 -> U2 and increase the chip count.

Because Bitmain still needs to turn a profit and adding more chips would increase their production costs.

My understanding is that bare chip costs are quite marginal.

[klondike_bar]

Seems to me regardless of what process node they use they will need to use a chip with improved power consumption in the S2.  In the original U1 the chip was always under clocked.  The U2+ just took the same chip added a bigger heat sink and changed the default clock rate.  Since the S1 was already operating at close to the maximum clock rate they will almost certainly have to use an improved chip in the S2.

Why?  Just do the reverse of what they did going from U1 -> U2 and increase the chip count.

Because Bitmain still needs to turn a profit and adding more chips would increase their production costs.

My understanding is that bare chip costs are quite marginal.

chip costs are minimal but the initial R&D can be several million dollars

[klondike_bar]

I can envision having an antminer where a pair of big waterblocks measuring about 10"x10"x2" (about the size of two of the current heatsinks stacked vertically) could have 4 blades bolted to each it and would only require about the same amount of space as 2 antminers

      {[]} {[]}
      {[]} {[]}       {,} represent hashing boards    [] represents waterblock
        |\   /|
      [radiator]

     <-~12"->

Bitmain could probably achieve this via chinese manufactures pretty cheaply, maybe at a 5% premium considering the overall metal mass would be similar to 8 heatsinks and the rad could run with 2 fans instead of 4 fans (save $5-10 per fan).

The only issue is that the board design or orientation may need slight tweaking in order to accomodate the back-side chips near the requlator as well as allowing connections for power and ethernet still. (In theory, a single control board could probably be used, further reducing costs - but might require a slightly faster processor)

[dropt]

Seems to me regardless of what process node they use they will need to use a chip with improved power consumption in the S2.  In the original U1 the chip was always under clocked.  The U2+ just took the same chip added a bigger heat sink and changed the default clock rate.  Since the S1 was already operating at close to the maximum clock rate they will almost certainly have to use an improved chip in the S2.

Why?  Just do the reverse of what they did going from U1 -> U2 and increase the chip count.

Because Bitmain still needs to turn a profit and adding more chips would increase their production costs.

My understanding is that bare chip costs are quite marginal.

chip costs are minimal but the initial R&D can be several million dollars

That's the point.  We're operating on the assumption that S2 still utilizes their 55nm chips, so decreasing Vcc/Clk and increasing the chip count could provide a more efficient miner at greater speeds (overall) with marginal increases in production cost(s).

[klondike_bar]

^if re-designing the board like that, a water block would be particularly good. putting 500GH+ on a single block with a radiator would make cooling a lot easier and reduce the noise and costs of air cooling.

[Abraxas601]

I was checking on any S1 Firmware updates and noticed this little tidbit in README.md regarding future software updates,

https://github.com/AntMiner/AntGen1/tree/master/PCB

"Future: We are designing a type of new control board based on ARM A8. When it is ready, the chain number can be up to 16, and the chip number can be up to 256 per chain."

[yxt]

same chip, modular 19" design

[precrime3]

Comparing the gains from U1 to U2 should be reasonable. Considering that according to there website, the only difference is a extra 400 mh/s from U1 to U2 (or 20% increase) with no change to power consumption, and slightly bigger footprint, proabbly to accommodate the different heatsink, I think a reasonable speculation/assumption would be the S2 having the same power consumption (360w Im hearing from forums) and having 20% increase in hash ( or 216 gh/s) with out OC. This may be flawed, but Bitmain is selling the U1 for 18 BTC while u2 is 13.9 BTC per 500 units. This may be the preorder, but if that scales as well, we could expect a 33% discount in price or .77BTC. Just my thoughts.

[klondike_bar]

Comparing the gains from U1 to U2 should be reasonable. Considering that according to there website, the only difference is a extra 400 mh/s from U1 to U2 (or 20% increase) with no change to power consumption, and slightly bigger footprint, proabbly to accommodate the different heatsink, I think a reasonable speculation/assumption would be the S2 having the same power consumption (360w Im hearing from forums) and having 20% increase in hash ( or 216 gh/s) with out OC. This may be flawed, but Bitmain is selling the U1 for 18 BTC while u2 is 13.9 BTC per 500 units. This may be the preorder, but if that scales as well, we could expect a 33% discount in price or .77BTC. Just my thoughts.

the u2 is the same design as U1 i think, except that its now hard-coded to run at 2.0GH rather then needing the bmsc-options variable in cgminer

[Delarock]

Official announcement from Bitmain:

https://bitcointalk.org/index.php?topic=515448.0

How close was this to your guesses?

[thomas_s]

Official announcement from Bitmain:

https://bitcointalk.org/index.php?topic=515448.0

How close was this to your guesses?

Spot on, we had pics of the prototype unnamed on our facebook page since the Texas Bitcoin Conference =), well with Bitmain stuff in the background.

[daddyfatsax]

Official announcement from Bitmain:

https://bitcointalk.org/index.php?topic=515448.0

How close was this to your guesses?

Spot on, we had pics of the prototype unnamed on our facebook page since the Texas Bitcoin Conference =), well with Bitmain stuff in the background.

That thing is huge!!!

[topminingcontracts]

Speculation ends, Production Starts, Preorder available https://112bit.com

Juan