Quote from: markm on March 07, 2012, 03:42:34 PM

Of course if these are not the full blown type of ASIC but the structured ones or something hopefully there will also be R&D overhead for a full blown implementation also to take into account once the R&D that went into this one is accounted for.

This is quite true. The cost of fabrication is extensive, and much of that cost is for very very experienced engineers to validate the design before it is manufactured. Silicon wafers, and the literal fabrication process are only part of the cost. It's easy to understate the complexity of putting together an integrated circuit. We're talking about tens of millions of transistors, complex timing circuits, and little tiny wires that all have to fit together perfectly to result in a working product.

FPGAs are a great because they provide power efficiency within an order of magnitude of the ASIC, but without having to worry whether your simulation of the metal layout might not work just right with the lithography process. Unfortunately, FPGAs carry with them a couple of serious drawbacks versus ASICs:

1. FPGAs use a lot more logic to implement the same amount of logic. That logic eats up power.

2. FPGAs themselves are very complicated chips with high fabrication costs. That keeps the unit cost high even at high production volumes.

In our exploration last year of various options for increasing the efficiency of Bitcoin mining, we thoroughly evaluated the FPGA options, because of course using FPGAs instead of ASICs would have greatly sped things up and reduced our risk. But, what we discovered was that the power, space, and unit cost efficiencies of ASICs will always win out over FPGAs in the long run.

So, if you believe that Bitcoin is going to be around for many years to come, as we do, then you'll want to make a long term bet on ASICs rather than loading up on FPGAs that will be left in the dust as the difficulty factor is pushed higher by ASICs. Short term calculations based on today's difficulty factor and exchange rate are just not relevant. If your goal is short term profit and you fear the future of Bitcoin, then stick with FPGAs.

I am very pleased that there are others like Butterfly Labs who seem to also be working with ASICs. This is great news for Bitcoin - the more custom logic that gets pushed out there, the lower the risk that Bitcoin will be co-opted by botnets operated by nefarious groups. Because, last time I checked, your mom's PC didn't have a Bitcoin ASIC in it.