Regarding computational cost, we measured separately training/parameter estimation time and base-calling time for the four packages that we could run on our benchmark linux computer: Ibis, BayesCall, naiveBayesCall and Rolexa.
We estimate our approach with a SimpleScalar based simulator run on the Mediabench benchmark suite and compare to the trace cache performance on the same benchmarks.
The performance of this new version of the algorithm is demonstrated using the SSCA#1 (Bioinformatics) benchmark running on one GPU and on up to four GPUs executing in parallel.
Recently, a benchmark run on the K computer has achieved 10% of the peak performance using five nodes (40 cores), and usage has been verified with 81,920 nodes (655,360 cores).
We also give some performance results from a mini-benchmark run on the V2 prototype.
We ran our benchmark on a Ubuntu-Server 10.04 on a Dell PE T610 2x Intel Quad core X5570 Xeon Processor, 2.93 GHz 8 M Cache, 64 GB Memory (8×8 GB), 1333 MHz Dual Ranked RDIMMs for 15 Processors, Advanced ECC.
The benchmark was run on Java SE 8, CentOS 7, Intel Core i7-4790 CPU @ 3.60GHz with 16 GB of RAM.
This performance benchmark was run on a laptop equipped with a processor Intel Core i7-4810MQ 2.80 GHz, with 16 GB of RAM and running Windows 7 Enterprise (64 bits).
This benchmark was run on a standard desktop computer (4 GHz Pentium intel with 2 GB of memory running Windows XP).
This benchmark was run on a system with dual six-core AMD Opteron 2427 processors and 32 GB random access memory, equipped with a RAID-10 disk array.
Ian Douglas has run some benchmarks on his Froyo-powered Nexus One, and it's scoring almost 5 times better.
