Summary:

We achieved a peak performance of 76 GFLOPS with HPL(Linpack) while solving a solving a 70K x 70K dense matrix. which comprises 78% of the theoretical peak.

High Performance Linpack(HPL) Benchmark Results:

Number of Processors Problem Size(N) GFLOPS 1 14600 03.4699 2 20400 07.1108 4 29000 13.6359 8 41400 26.6700 12 50400 39.1497 16 58600 52.1630 20 65536 61.3741 22 68700 68.3405 22(optimized) 70000 75.8789

MATLAB Distributed Computing Benchmarks:

Performance of several other benchmarks like High Performance Computing Challenge(HPCC) benchmarks and NAS benchmarks has also been analysed on Neptune.

Resources:

• HPL (High Performance Linpack)
• HPC Challenge Benchmark
• NAS Parallel Benchmarks(NPB)