It makes sense that KnC in partnership with ORSoC would use an ORSoC Linux solution.
If they use the SO-DIMM pictured:
(FYI, SO-DIMM is the form factor typically found for laptop memory)
It's NOT a CPU powerhouse, at only 166MHz.
It does not need to be!
It _IS_ a perephial powerhouse!
It has 5 USB ports.
It has 5 ethernet ports and can do GiG ethernet.
It has an Altera FPGA on it for lots of helpful miner specific things.
Lets consider each in a bit more detail.
The SO-DIMM:
Only 1 connector needed for the brain, everything needing connection to the rest of the system is handled by the PCB that it rides on.
You just snap em in, possibly with screws to retain it.
You can preconfigure/program them on existing dedicated programming stations instead of having to hook up each miner for programming. FAST!
It is an industry standard 'module' in use for many years. (unlike the Pi and BBB form factor)
Most servers have SO-DIMM sockets so you can add a PC to your server to monitor the health etc.
The FPGA:
They are like magical reconfigurable printed circuit boards. (their most basic feature)
All you have to do during design is make sure you have a trace leading from the reconfigurable I/O pins to each external device and you can worry about the exact details as needed.
The FPGA in question has many additional features that address a miners needs.
Mainly lots of configurable logic (22,000 LUT's), clocks, and high speed communication, along with coordinating that communication.
I'm not suggesting that it will(no massive heatsink), but it is in theory possible that a miner with ZERO modules could still hash on the FPGA all by it's lonesome.
They chose a Linux solution that already has an FPGA onboard that the competition has had to add to their controller boards.
An FPGA primer linked below
http://www.ni.com/white-paper/6983/en/para4 The Ethernet Transceivers and Switch Engine:
Ethernet is a good choice for interfacing with the, wait for it,,, internet!
USB has a length limitation that ethernet solves with ease.
Hosting USB units would be a pain and need lots of additional components.
USB onboard:
5ports, 4 for ASICS and one for 'users'.
Utilize existing software that primarily uses USB to communicate with mining units.
We don't know the detail of the 5th port but it is possible that it could support a WiFi connection.
166MHz CPU:
It does not take much brain power to interface with a specialized team of perephials.
Consider "seemingly" incompetent middle managers(very rare but they exist), they are successful because their job is coordination not necessarily whatever specialty they manage.
Here is a link to the info for the pictured(by KnC) solution
http://www.orsoc.se/?page_id=484In my mind KnC have demonstrated good component choices for getting an ASIC solution to market FAST! Imagine being an employed FPGA engineer with an interest in mining. You take your idea to your boss and they not only listen but say YES. The Mars FPGA evidently did well enough to convince the team to decide to throw down on the competition.
So far the execution of the 'margins on top of margins' KnC plan appears to be on target.
All ASIC vendors are at the mercy of their chosen foundry. An orginization with a long established relationship with a foundry reduces this risk. They can hold future business over their head to help insure timelines are met. They can get production scheduling and predict delivery based on past performance. This is a 'margin' the competition may not have.
We will know one way or the other shortly.
Show Posts
