From: Implementation of a high-throughput low-latency polyphase channelizer on GPUs
With transfer
Without transfer
Old PCZ
52.591 ms / 22.8 MSps
50.808 ms / 23.6 MSps
New PCZ
2.868 ms / 418.5 MSps
0.734 ms / 1,633.6 MSps
Speedup
18.3 ×
69.2 ×