Exploring Many-Core Design Templates for FPGAs and ASICs

Table 4

32-Node. Performance comparison between MARC, hand-optimized, and GPGPU.

Configuration	Per iteration	Relative
	Time (𝜇s)	Perf.
GPGPU scaled reference	174	0.0024
MARC-Ropt-F	2550	0.0002
MARC-C1-F	172	0.0025
MARC-C2-F	124	0.0034
MARC-C4-F	136	0.0031
Hand design FPGA	51.4	0.0082
MARC-Ropt-A	27.6	0.0152
MARC-C1-A	1.47	0.2863
MARC-C2-A	1.11	0.3808
MARC-C4-A	1.17	0.3608
Hand design ASIC	0.422	1.0000