Skip to main content

Table 11 Summary of SW implementations on accelerators and low-power devices over the past five years

From: DNA sequences alignment in multi-GPUs: acceleration and energy payoff

Device Hardware model Power Implementation Input size GCUPS GCUPS/W Ref.
FPGA Altera Stratix V 25 W (a) OpenCL 23Mx25M 37.67 1.50 [34]
Accel. Intel Xeon Phi 3120P 270 W (a) OpenCL 23Mx25M 30.36 0.12 [34]
GPU Nvidia Tesla K20 225 W (a) SW# 23Mx25M 44.19 0.19 [48]
GPU ” Tesla K20 225 W (a) CUDAlign 3.0 23Mx25M 40.69 0.18 [49]
GPU ” GeForce GTX 980 165 W (a) SW# 23Mx25M 67.55 0.41 [48]
GPU ” GeForce GTX 980 165 W (a) CUDAlign 3.0 23Mx25M 84.84 0.51 [49]
GPU ” GeForce GTX 980 103.06 W CUDAlign 4.0 51Mx50M 112.71 1.09  
GPU ” Titan X Maxwell 189.00 W CUDAlign 4.0 51Mx50M 158.10 0.84  
GPU ” Titan X Pascal 193.96 W CUDAlign 4.0 51Mx50M 276.53 1.43  
  1. aAuthors do not measure real power consumption, but estimate it using TDP (Thermal Design Power)