Quote from: defxor on September 03, 2011, 07:32:35 AM

Quote from: grod on September 03, 2011, 05:34:29 AM

If the price per BTC is $50 but cost to produce is $3 there is no real reason to optimize for cost of production. You will want to optimize for volume.

Exactly. That's why I don't get Synaptic's "price must go up for ASIC mining to take off". It's the other way around.

Quote from: grod

Now you're following along as to why I think FPGA mining may be the beginning of the end.

So where I'm headed with this: a crash of difficulty because the ONLY viable way to mine is FPGA, with each board having a quite low Mhash/$ but high Mhash/joule efficiency. Further fueling a collapse in price (which makes GPUs that much more impractical for mining) until a new stability level is reached. Mining 3 coins a day at a sale price of $1/coin and power cost of $2 is not as smart as mining 1 coin at a cost of 20c.

this

Quote from: defxor on September 02, 2011, 03:38:26 PM

I think it's the other way around actually. I won't even touch mining until we're at ASICs, since I consider the price to follow production cost until we hit the uptake vs deflation knee of the curve. GPU miners need to charge $6/BTC. FPGA miners can charge $0.5. ASIC miners can ...

Quote from: defxor on September 02, 2011, 04:23:55 PM

I consider mining bitcoins to be just as mining any other form of natural resource. The price floor is at extraction cost (plus minimal profit), anything above that is speculation. Since I agree with you that speculation won't continue to drive BTC price in the near term my projection is that the floor is where the cheapest mining can be had. Currently that's FPGAs, but they will be replaced by ASICs. GPU miners, like CPU miners already have, will need to drop out.

Do note that there's a qualifier, the above is only true if speculation doesn't take over. We've already seen that happen once and it could very well happen again if there's suddenly a huge uptake which needs a capital influx the current 7200 BTC/day inflation can't satisfy.

it costs at least a million USD to develop and fab an asic in any cost effective economy of scale...

if not a million than some other large number that will never be spent on such an idiotic venture.