I guess FPGA division (nee Xilinx, which was always a bit sketchy, even if they had best silicon) doesn't learn much from the GPU division.