architecture - Parallela FPGA- 64 cores performance compared with GPUs and expensive FPGAs? -

this parallela:

http://anycpu.org/forum/viewtopic.php?f=13&t=66

it has 64 cores, 1gb ram, runs linux, ethernet- shouting it....

my question is, performance/capability perspective how parallela compare more expensive fpgas? have wider buses/more memory/faster processor clocks/more processors on chip?

i understand gpus massively parallel simple operations , cpus better more complicated single-threaded computation- expensive fpgas , parallela fit on curve?

the parallela runs linux- yet under impression fpgas have logic flashed on them writing verilog or vhdl?

a partial answer : fpgas tend not have processors on chip (there exceptions) - if think processing fetching instructions , executing them 1 after other, haven't grasped fpgas. if can see how execute 1 complete iteration of inner loop in single clock cycle, you're getting there.

there tasks easy, , fpga can wipe floor other solution. there tasks impossible, , parallela contender. don't see 1 high performance solution overall winner; there impressive things being done gpus (low power isn't 1 of them!), , many-core xmos or parallela solutions have place too.