HPL Results
Revision as of 15:12, 5 December 2014 by David (talk | contribs) (→EPCC 4 node / 2 k40 per node / 2680 v2 cluster / liquid cooled)
Introduction
This page will be used as an archive for the results of HPL on varies system configurations
Intel Platforms
A 16 node E5-2697v2 QDR Fabric (92% Efficiency)
- Theoretical Peak 8294.4 GFlops
- Measured performance 7628.1 GFlops
- Output File
AMD Platforms
GPU Systems
2x K20 GPUs
- Measured output: 2.159 Tflops
- Full output file:
================================================================================
T/V N NB P Q Time Gflops
--------------------------------------------------------------------------------
WR03R2L2 80640 896 2 1 161.92 2.159e+03
--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)= 0.0033049 ...... PASSED
================================================================================2x K40 GPUs
- Measured output: 2.632 Tflops
- Full output file:
================================================================================
T/V N NB P Q Time Gflops
--------------------------------------------------------------------------------
WR03R2L2 80640 896 2 1 132.84 2.632e+03
--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)= 0.0032065 ...... PASSED
================================================================================EPCC 4 node / 2 k40 per node / 2680 v2 cluster / liquid cooled
- Measured output 10.14 TF
- output file
================================================================================
T/V N NB P Q Time Gflops
--------------------------------------------------------------------------------
WR03R2L2 161280 896 4 2 275.74 1.014e+04
--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)= 0.0000195 ...... PASSED
================================================================================