With Ken releasing the official specs of the 55nm chip it's time to again assess our potential position in the mining race. I'd like to aim this post specifically at AsicMiner who announced their chip specs at the end of last week.
Exhibit A:
https://bitcointalk.org/index.php?topic=438359.msg4816701#msg4816701Specification
Process node: 40nm
Package type: QFN64 8mmx8mm (with another option of QFN64 7mmx7mm possible)
I/O: Standard SPI protocol with clk, in, out and cs.
Rated Hashrate: 12.8GHash/s per chip, with a wide range of overclock/downclock options
Rated Voltage: 0.72V, recommended voltage range is 0.55V-1V
Power Consumption: 0.2J/GHash low voltage, 0.35J/GHash rated voltage
Price range
0.49$/G-0.99$/G, depending on order size and delivery speed of choice.
Exhibit B:
http://virtualminingcorp.com/shop1/index.php?id_product=34&controller=productPre-Order Fast-Hash-One 55nm Reel (3,500 Chips) +/- 1%.
BitLadder-55 Specifications:
Technology: UMC Technology: 55 nm
Die size: LPDie size: 3.126 x 3.558 mm
Substrate package: 11 x 11 mm
Package type: QFN-132 with thermal pad
Design type: Routed for density and performance
Performance: 1.8 GH/s (active cooling), 1.7 GH/s (passive cooling)
Power Consumption (Est.) 4.8 W @ 455 MHz (*active cooling) 4.5 W @ 425 MHz (**passive cooling)
Performance design: 4 engines @ 455mhz nominal
Digitally programmable between 200 MHz and 480 MHz Hash performance
Price
$20,055 per 3500 chip reel, or $20,055 for 6.3TH
$3.18/Gh
Well,
What do you guys think? I'd love to hear from the EEs among us regarding both sides of this issue.
Those specs as posted are very different than what Ken had previously posted.
For one, the QFN-132 package is going to be a bit of a PITA.
The active vs passive cooling thing makes no sense whatsoever. He is not using a standard definition of "active cooling" at all.
* "Active cooling" means a two Peltier elements (one on top of the chips, one on the underside of the PCB).
**"Passive cooling" means two fan-cooled heat sinks (one on top of the chips,one on the underside of the PCB).
Peltier cooling is ridiculously inefficient, on the order of 50%. If you're trying to cool 4.8W with a peltier, you'll have about a 5W draw from the peltier element in addition to the 4.8W draw from the chip itself. I would s
Using the specs for "passive" cooling, you're looking at 2.65J/GH at the chip level. That's even higher than Avalon at 2.05J/GH at nominal 1.3GH/s speeds. Right now you'd pay $8710 for 2500pcs of 1.3GH/s Avalon Gen2 chips, or $2.68/GH/s.
So, to sum up, at current prices if those specs are accurate you're looking at 19% higher cost than Avalon, 29% higher power use, untested chips vs proven design, no reference design vs many existing designs, and shipping optimistically April 17th versus immediately.
I know many people have issues with Avalon, but in this EE's opinion you'd be crazy to buy those chips at those prices. They need to be $1/GH/s to be even remotely attractive.