Our latest generation of AMD-based HPC clusters, Antares, has come online adding 1152-cores to bramble’s computing capacity. We’ve been really impressed with the pace of this machine, so we wanted to share some benchmarking data with you.
We built Antares using second-generation AMD EPYC chips and FDR Infiniband, all housed in Gigabyte hyper-converged chassis. The cluster consists of four 2U chassis, each with 128 cores and 1 TB of RAM.
Cluster prices
We designed our clusters to balance CFD performance and cost. Prices start around £25,000 for 128 cores and reach up to £183,000 for a full 1152-cores. (Prices are valid as of March 3, 2021, and include a one-year warranty.)
Cluster benchmarking
Many of the CFD simulations that are run through bramble are full car models running steady-state RANS or transient DES solves. These models vary in size depending on the amount of detail, but are typically between 80 million and 150 million cells.
For these benchmarks we compared Antares with our first-generation AMD EPYC cluster, Hyperion. We’ve generated two versions of the standard OpenFOAM® motorbike test cases: one at 82 million cells and the other at 148 million cells. We created the mesh using snappyHexMesh and solved using simpleFoam (RANS) and pimpleFoam (DES), running each in both single and double precision.
Please contact us if you’d like to see some different benchmarks
Now for the interesting part… the actual benchmarking data, starting with the meshing times. snappyHexMesh often struggles with cache size and memory access. Thankfully though, this is where Antares has made big strides forward over the older Hyperion.
Antares cuts meshing times by an impressive 20% across all core counts and model sizes. The next two tables compare average iteration speeds for the simpleFoam and pimpleFoam solves in single precision (SP) and double precision (DP) respectively.
Once again Antares shows its speed with 4 to 20% reductions in solve time seen depending on core-count and precision.
Antares is already working hard for bramble’s customers, but if you’d like to find out more or are interested in purchasing your own cluster, please contact us.
