Skip to main content

Table 2 Performance comparison between rapidGSEA and broadGSEA

From: rapidGSEA: Speeding up gene set enrichment analysis on multi-core CPUs and CUDA-enabled GPUs

H (50) 1,024   4,096   16,384   65,536   262,144   1,048,576  
broad Xeon 83.6   307.1   1,149.0   4,681.0   18,301.0   70,946.8  
rapid Xeon 3.0 (28) 7.0 (44) 22.7 (51) 85.8 (55) 339.4 (54) 1,357.8 (52)
rapid Tesla 1.8 (45) 2.7 (112) 6.5 (176) 21.9 (214) 83.7 (219) 329.9 (215)
rapid Titan 1.6 (54) 1.9 (164) 3.1 (367) 8.3 (564) 28.2 (648) 105.2 (674)
C2 (4,725) 1,024   4,096   16,384   65,536   262,144   1,048,576  
broad Xeon 1,113.0   4,600.0   17,962.0   70,213.4   [274,464.0]   [1,072,878.9]  
rapid Xeon 89.3 (12) 282.9 (16) 1,084.8 (17) 4,266.2 (16) 17,069.0 (16) 68,682.3 (16)
rapid Tesla 13.1 (85) 31.6 (146) 108.5 (166) 418.7 (168) 1,685.5 (163) 6,732.6 (159)
rapid Titan 10.3 (108) 20.8 (221) 61.9 (290) 214.0 (328) 895.9 (306) 3,447.3 (311)
  1. Runtime in seconds and speedups of rapidGSEA compared to broadGSEA (round brackets) using up to n=1,0242 permutations on the Hallmark (H: 50 gene sets) and Curated (C2: 4725 gene sets) collection. Please note that runtimes in square brackets indicate linearly extrapolated runtimes of broadGSEA in log-log space