I helped icedrill get his up and running and fixed a few obvious cgminer bugs that only showed up with real hardware especially when it misbehaved (the changes went into git master). He had a 3 module unit. One module was completely dead and had to be disconnected or it wouldn't fire up. One module would overheat and trip the code that disables it almost immediately unless it was run at 20% speed. The final module ran stable at ~420GH. There's only so much debugging I can do remotely on the code without the hardware myself, but there are still bugs in the code preventing it running stable for more than a few hours at a time, but that does appear to be software related. The one working module appears to work quite well at that speed. I know there were a couple of board revisions and I suspect there may have been one more (revision) in the works but not really sure which boards went to icedrill.

I doubt very much these early engineering samples hand assembled are a sign of what the final product will be like, but perhaps that last module is a better indicator. I don't know anything about his power consumption.

   That's true...

HashFast's chips are about 3 times more powerful when compared on a GH per sq.mm basis.

HashFast is shipping with top-shelf quality Sea Sonic PSUs, so our customers won't have to put up with exploding/self-immolating units like KNC's did.

In addition, HashFast lined up an exclusive with CIARA, which means HashFast machines will be assembled and shipped with utmost expediency (no random guys helping load some random van), in professional/durable packaging.

That means double-walled boxes, so HashFast customers' hardware won't arrive broken like so much of KNC's did.

Finally, (again unlike KNC) HashFast has been working closely with ckolivas for a long time to ensure cgminer works flawlessly with our innovative new GN Protocol, Global Work Queue, etc.