Incorrect pools behavior when using multi-socket windows machine with > 64 HW threads

On a dual-Xeon E5-2699 v4 which has 44 threads per socket (88 overall), when running

x265.exe sample.y4m --preset ultrafast -o sample.hevc

x265 only detects 44 threads in socket 0 and no threads in socket 1! The reason for this strange behavior is that windows API returns incorrect # total CPUs when the total # CPUs exceeds 64 on a multi-socket machine.