Little back of the envelope calculation:
Let's assume the CPU was able to do one SHA per clock cycle (absolute maximum, with specialized instruction that doesn't exist). Let's assume a constant all-core (usually only 1 core can boost, but let's assume so) boost clock of the highest boosting CPU of roughly 5GHz (high core count CPUs clock lower, but let's give it its best shot). 128 cores x 5GHz x 1 hash/s = 640GH/s.
At a power draw of probably 300-500W and probably costing around $2,000 USD.
Well if what they saying is true that it power consumption is 10 times less than what exist now, if you assume 15 watts per core for standard hardware, then this thing 1.5 watts per core. so under 200 watts total at full load. $2000 for that cpu is probably about in the right ballpark. they said the cost is 3x lower than other things.
You can get a Compac F [1] for around $200 and 15W of power draw that pushes 300GH/s, so two of those would match this hypothetical CPU at a price of $400 and 30W. That's 5 times cheaper and at least 10 times more power efficient, so you could say 50 times better.
the problem with doing that is, it can't mine anything else except sha-256 coins and even at 300GH/s I don't think it makes any money. which is why i later said that the only way this cpu could make sense in mining is to mine some other algo. where it could probably make some money.
but there is an upfront cost kind of like there is for asic mining. no getting around that. but i realize now that asics are way more advanced in bitcoin for cpus to be competitive.
