UPGRADE YOUR BROWSER
We have detected your current browser version is not the latest one. Xilinx.com uses the latest web technologies to bring you the best online experience possible. Please upgrade to a Xilinx.com supported browser:Chrome, Firefox, Internet Explorer 11, Safari. Thank you!
12-14-2017 11:13 PM
Given an architecture A1 and a program P : running P on A1 takes T1 seconds while consuming w1 watts (E1 = w1.T1).
On the other hand, given an architecture A2 and a program P (same as above) : running P on A2 takes T2 seconds while consuming w2 watts (E2 = w2.T2)
what is the best way (formulas) to conclude, if possible, the below metrics :
Performance per watt for A1 and A2? is it reasonable to assume that performance is the execution time of P.
performance per watt PPW ratio i.e. by how much the PPW of A1 is better/worse than the PPW of A2.
how to deduce conclude the energy efficiency ratio e.g. by how much A1 is more/less energy efficient than A2.
Thanks.
12-17-2017 03:26 PM
(1) Normally performance would be inverse of execution time (ie lower time = more performance). As such, PPW = 1/E.
(2) The normal formula would be PPW1/PPW2. Suppose w1 = 2W, T1 = 0.1s (FPGA implmentation) , T2 = 80W, T2 = 0.01s (CPU implementation). The E1 = 0.2J, E2 = 0.8J, PPW1 = 5/J, PPW2 = 1.25/J, and PPW ratio = 4.
(3) This is exactly the same as 2; for the same power A1 does four times as much work.
12-17-2017 03:26 PM
(1) Normally performance would be inverse of execution time (ie lower time = more performance). As such, PPW = 1/E.
(2) The normal formula would be PPW1/PPW2. Suppose w1 = 2W, T1 = 0.1s (FPGA implmentation) , T2 = 80W, T2 = 0.01s (CPU implementation). The E1 = 0.2J, E2 = 0.8J, PPW1 = 5/J, PPW2 = 1.25/J, and PPW ratio = 4.
(3) This is exactly the same as 2; for the same power A1 does four times as much work.