And the BFL chip performance is only marginally better hash/$ than Avalon's. Why would anyone take that risk?
Price/Performance
BFL: 2600 MH/s for $89 is 29.2 MH/s per $1
Avalon: 280 MH/s for $10 is 28 MH/s per $1
Power efficiency
Avalon: 66,000 MH/s @ 600W 110 MH/w
BFL: 5200 MH/s @ 35W 148MH/w (the ars technica box measured 5.2 @ 50W)
This advantage turns negative when taking into account that these chips theoretically arrive 4 to 5 weeks after avalon chips arrive when ordered at the same time. In this time the difficulty grow so much that there is no advantage anymore or even an disadvantage.
The pricing for BFL chips is extremely goofy. The better performing chips are actually cheaper per engine, so until BFL updates their prices there's no reason to purchase anything but Grade A chips.
As for why you would choose the BFL chips over the Avalon ones, I would say that BFL's present more interesting opportunities for board and cooling design. The BFL ASICs consume less power than their Avalon counterparts, and the lower device count per GH simplifies board design. If both were available immediately at the current prices I think it'd be a no-brainer.
The big question for BFL (and Avalon) is whether they can actually deliver on the timelines they've talked about. Both of them have had huge issues with getting production out the door for completed miners, but chips sales should be much simpler. It's been 7 weeks since people started paying for Avalon chip sales now. If Avalon does get their chips out in the next 3 weeks as promised, 3+ months for BFL chips will be hard to justify. If Avalon is late (and they're already 3 weeks late on providing samples), it might be closer.
?
