Linpack benchmark results code#

All subsequent results were obtained with the binary of the C version of the benchmark ( linpack.tos) in this repository, with double precision and using the default array size when prompted (200, which results in a n=100x100 matrix to be solved). Note that to some degree the results depend on the compiler that is used and on its options. This was compiled with GNU Fortran 8.3.0. AVX512 doubles the vector width from 256 bit to 512 bit over AVX2. These Skylake-X Core i9 'desktop' processors benefit form having Intels latest vectorization hardware, AVX512. Linpack solves a dense (real8) system of linear equations (Axb), measures the amount of time it takes to factor and solve the system, converts that time into a performance rate, and tests the results for accuracy. This post is a short listing of results for the Linpack benchmark which is still my personal favorite CPU performance metric. We first started the affected container and recorded the Linpack test results and CPU utilization. The original benchmark (written in FORTRAN) is also included as linpackd.tos. Linpack Xtreme is a console front-end with the latest build of Linpack (Intel Math Kernel Library Benchmarks 2018.3.011). The Linpack problem size of the affected container and the overloaded container is set to 500, respectively.

Linpack benchmark results code#

The source code for linpack.tos was taken verbatim from this C version of the Benchmark. It solves a linear equation system and is used to evaluate the performance of computers and super-computers since 1979. It requires at least a 68020 CPU and a FPU. In addition to the performance results of LINPACK and the eight NAS parallel benchmarks, we have also included peak performance of the machine, and the LINPACK n and n(sub 1/2) values. Efficiency drops from 1 node to 2 nodes however, the scalability is mostly linear afterwards. In this study we have used the reported benchmark results and performed a number of statistical experiments using factor, cluster, and regression analyses. Efficiency for a single node is 69, while 67 for 2, 4 and 8 nodes. This is the famous LINPACK benchmark to measure the floating-point performance of Atari machines running TOS (or MiNT). As figure 3 shows, LINPACK performance for a single 8268 node is 3059 GFLOPS and 23946 GFLOPS for 8 nodes which means 7.83X scaling from 1 node to 8 nodes.