First,
Thanks to fsb4000 for the post with the links to the pictures, you beat us to it
16 boards with 8 chips = 96 chips. Anyone knows the specs at which the chip will/should run according to metabank?
I have a feeling the pictures shown are for the "double" machine where they posted about it in the 65nm thread before (sorry, I hadn't mentioned it before). So by double I think they mean 240 GH/s, which kind of makes sense since I am seeing 11 blades in this one, although some of them look different (silver vs beige/black capacitors), so not sure what's going on there - probably this is an early prototype with different cards.
The picture on the Metabank.ru/asic site shows 8 ASIC chips per board, and above that it states they are rated at 3GH/sec each. So I would extrapolate that to mean 8 asics * 3GH/s = 24GH/blade. So you only need 5 blades @ 24 GH each to get 120GH/sec. So I seriously doubt they will ship a box with around 11 blades for a 120GH product.
They also mention up to 1.6W/GH (not clear at what voltage that is though). So at 120GH, in theory it'd be a max of 192 Watts just for the ASICs alone, not including power required for all the other "doodads" on the PCBs. But should be well under 300W, maybe even less than 250W, but we'll find out I suppose. If the watts per chip is on the low end (0.85W/GH) or aprox 1/2 that, then obviously it'd be closer to 100-125W's.
From our own tests a month ago (see bottom of this blog post:
https://bitcentury.io/blog/initial-testing-of-bitfury-asic ) , when we tested above 2.4 GH (last entry at the bottom), the chip temperature hit 80 Celsius and kept rising, and power consumption hit 2.52 Watts/chip but the hashrate was unknown. We didn't wait around to find out the hashing speed because we had limited alpha chips and didn't want to kill any with excess heat (we had no heat sink for these tests); but if we assume 3W per chip for 3GH/s, then with 40 chips (8 chips * 5 boards), we're looking at 120 Watts + power required for all the other PCB components etc. So highly likely under 200W. This is all educated speculation though ;-)
Yeah, but bitfury's chips can be run at different voltages. There had been talk about running them at a higher voltage to get more performance per chip, more like 2.9GHash/chip and 170W for the device. That's why I'd like to know whether this is really the metabank device or not.
There is a very high chance this is indeed the Metabank device. I base that guess on the look of the blade PCBs. They look nearly identical to the one they posted on their site:
https://metabank.ru/images/asic/asic_board.jpgYou'll note the latest pictures show a green PCB instead of white, but the huge black/beige capacitors are a dead give away.
As for whether the blades are soldered on the master board or sockets, your guess is as good as mine - I can't quite tell from the pictures. My concern looking at this box is that I don't see any screws for opening it up, so hopefully they don't glue it shut, in case we need to open them to properly secure the blades before shipping. Further, I wonder if those two fans in back of picture #73 are going to be providing sufficient airflow to run these things for prolonged periods at room temperature.
Lastly, I'm still not clear on the PC connection type. What we can see in picture #72 on the bottom right is a Raspberri Pi, and those have both USB and Ethernet connections, but not sure which one(s) the software will support. In Picture#73, I don't see any cables coming out of the Ethernet port, so not sure if this unit was even plugged in or what.
The good news is that if these box dimensions are in milimeters as per the PDF file, then the box is aprox 1/2 the size (for the double hashing unit!) or less of an avalon with twice the hashing power and 1/2 or much less the power consumption. Not sure if they just made 1 box for single and double hashing units, or whether they made a smaller box for the single (120GH) hashing units.
