So no chance of a 2-4 blade install? It would be awesome if your blade interface is ASIC agnostic and we could start with a couple AVALON blades, add some BitFury blades, and finish off with some KNC blades.
We may do smaller versions but the real benefits of this is that it is designed for size. There are so many offerings in the small and middle size already we didn't want to really to drop in there in the early rigs. There is also the issue of managing builds and with the Avalon chip supply coming potentially from customers we simply need to cut down the number of people we are working with in this early phase of production. As with everything we have done so far in the bitcoin world we are aiming to maintain our indicated timelines.
Our vision for this system is that it might have different processing elements and it is kind of modular albeit on a large scale. So to be crystal clear on that we will work with other vendor silicon as much as our own offerings FPGA or ASIC. There is merit in this model as some types of board many not be available to all markets due to export restrictions. CM4 is very likely to be in the export restricted bracket and we are taking advice on that. CM3 with a very fixed function is less likely to be an issue export wise. Beyond CM3/CM4 we have other technology options in progress and being evaluated. No timeline for these before you ask but certainly not soon. It is not a secret that we have been very busy with non-bitcoin side of our business and that is expected to be the case for some months to come yet.
We have been asked what it will look like and basically the answer at the moment is a 19" rack. So not pretty but functional. We have some plans for exterior casings that will allow the unit to go into places that it could be used for water or home heating but those are far from finalised as yet and those are a secondary task. The inital units will be traditional fan driven air cooling.
Control wise some of you may have seen the dimm socket holes on the CM3 CAD picture. That is where our board/rack controllers are going to sit. Initially this will be a combo of a small FPGA and standard processor module that will allow us to run a Linux + MPBM combination but we have other options planned as well. There is an Ethernet and USB interface routed from the socket so each board potentially could run stand alone but we have also made provision to run up to 8 boards from one controller module so a master/slave type operation is possible. This approach gives some advantages in doing other technology options as we get to reuse the same controllers for new technologies.
Power wise we are aiming for a maximum of 6KW with a CM4 solution but we are hoping to get 500GH/s at that. There is still more work to be done on this to confirm that so that number may be a little off. A CM3 solution is likely to be somewhat better at maybe 3KW at 300MH/s based on published figures. For some customer/countries the 6KW may be an issue and we may alter the unit spec a little to work in with what customer sites can support. In the UK many domestic ring mains are 30A at a nominal 230V so you can see why we are aiming at the 6KW. As most houses here only have mains feeds of 60-100A it is also something of a practical rig limit in taking half or one third of an available supply. Obviously if we are looking at industrial units or data centre operation that is maybe a different situation.
CM4 does not use Avalon chips. Only CM3 uses Avalon chips.
