Research Article

The Potential for a GPU-Like Overlay Architecture for FPGAs

Table 3

A breakdown of how each stage of an HT memory request contributes to overall access latency.

Action StageLatency (ns)

Request (1) FPGA HT IP core 70
(2) host HT controller 32

SDRAM (1) access and data fetch 51

(1) host builds response packet 12
Response (2) host HT controller 30
(3) FPGA HT IP core 110

Total Latency 305