Hardware Benchmarking Topic HPC CPU Scaling

Page Under Construction

Introduction

TUFLOW HPC can run on either CPU or GPU hardware. When running CPU hardware multiple CPU cores can be utilised as the calculations are performed in parallel. When using multiple cores, at each timestep there needs to be a transfer of information between CPU cores. This adds an overhead and means that the scaling is never linear. For example, running on two cores is not twice as fast as running on a single CPU core.
This page discusses the speed benefits when running a range of models across different numbers of CPU cores. For a discussion on scalability of TUFLOW HPC on different numbers of GPU devices please refer to HPC (GPU) Scaling.

A computer with Intel(R) Core(TM) i7-5960X CPU @ 3.00GHz processor was used for the test. This provids access to 8 physical cores. With hyper-threading enabled this equates to 16 threads (or logical processors). With hyper-threading turned off it equates to 8 threads (or logical processors).

Results

???? Charts of scalability or runtime with varying numbers of cores are required. It would be good to have this for a couple of cell sizes / models. E.g. FMA 2 at 20m and 10m at 1,2,4, etc cores.

Hyper-threading /

Intel Hyper-Threading (tm) is designed to improve parallelisation of computations. For each physical processor core on the CPU two virtual / logical cores are utilised.
???? It would be good to have runtimes for say the FMA 2 model at 20m for a range of computers on the maximum numbers of cores and maximum number of threads.