[schoen]

Indeed, Bitcoin ASICs are not only "optimized" for the Bitcoin mining task (computing a particular hash function), they usually literally don't include the logic to perform other general-purpose computations at all! That's a big contrast with GPUs.

There have been interesting discussions about a cryptocurrency whose proof of work task would be something in some way more interesting or more useful than partial hash collisions, but I don't think many such systems have caught on. There is a prime-related one called Primecoin:

https://en.wikipedia.org/wiki/Primecoin

So, I guess that's a precedent for creating new cryptocurrency designs that do something else. I don't know if there's a way to make any of the BOINC tasks into cheap-to-verify PoW systems or if anyone's tried to do so, but that might be a cool project.