Sadly not really true. A massive chunk of space on an FPGA is devoted not to logic elements but all possible reroutable interconnections between these logic elements, and due to the complexity and manufacturing yields of very big FPGAs, one that might barely match the complexity of the lowest-end GPUs will cost more than your entire Alienware gaming rig. IMHO, using an FPGA as a computation device is far more interesting than so-called many-core CPUs.
Performance is NOT the number of cores that you can count but the number of logic gates that you can effectively utilize, and 'core' implies a lot of wasted logic that an FPGAs can devote purely to any task. This is the real strength of programmable logic that will be realized once FPGAs makers offer better tools for partial reprogrammability. (Xilinx only briefly did with JBits)