FPGA/GPU optimized but ASIC proof SHA256 coin? Is it possible?

[TheSwede75]

Completely hypothetical question. Would it be possible/feasible to design a coin that uses SHA256 but with variables that are explicitly ASIC hostile (memory requirement, cycles or other factor) towards the current BFl/Avalon architecture and optimized for current gen GPU/FPGA?

I think it's if nothing else a fun proposition.

[markm]

Sure, just do sha256(sha256(sha256))) instead of sha256(sha256)

-MarkM-

[TheSwede75]

Sure, just do sha256(sha256(sha256))) instead of sha256(sha256)

-MarkM-

Stop making me feel stupid! 

[mercSuey]

Sure, just do sha256(sha256(sha256))) instead of sha256(sha256)

-MarkM-

Begs the question, what are the benefits of deep nested SHA256 versus Scrypt algo?  This is interesting, both from a hardware perspective as well as programming/software.  I'm almost tempted to get something running on testnet and analyze performance differences.