MML is happy to announce the availability (currently in private preview) of our R6 cluster for scientific computing to research groups and individual researchers. Access to R6 does not require a project proposal.
R6 is optimized for memory bandwidth-driven applications, such as machine learning, molecular modelling as well as mesoscopic and continuum modelling. Compute nodes offer both interruptible (low priority) and uninterruptible (high priority) EPYC 7V12, 7V73X and CPUs and T4, A100 and H100 GPUs in multiple configurations.
For example, R6 molecular dynamics-optimized nodes feature AMD EPYC 7742 with 120 processor cores, 4 GB of RAM per CPU core with no simultaneous multithreading, providing up to 350 GB/s of memory bandwidth and FP64 compute of up to 4 teraflops. R6 VM interconnects are based on 200 Gb/sec Mellanox HDR InfiniBand, in a non-blocking fat tree for optimized RDMA performance, supporting Adaptive Routing and Dynamic Connected Transport.
Currently, each R6 cluster can provision up to 1000 nodes or 120000 cores on Linux SLURM.
To purchase an R6 consumption quota, click on the following link
https://www.paypal.com/ncp/payment/CS6L5XVRVATF6
Please address R6 onboarding queries to hpc@mmlpi.ch