Research Computing Services offers a range of advanced platforms for computationally intensive research.
Our two main high performance computing platforms are Cumulus (a CPU cluster) and Wilkes3 (a GPU cluster). These are collectively known as the Cambridge Service for Data-Driven Discovery (CSD3).
Cumulus is the workhorse platform for high performance computing. It is designed to be used by researchers who need access to scalable high performance facilities.
Cumulus consists of four partitions of compute nodes containing Intel Xeon Scalable Processors:
- Skylake – 18000 CPUs (~6 GB memory per CPU)
- Cascade Lake – 37000 CPUs (~3.5 GB memory per CPU)
- Ice Lake – 42000 CPUs (~3.5 GB memory per CPU)
- Sapphire Rapids –12544 CPUs (~4.5 GB memory per CPU)
The nodes comprising Cumulus are connected via a very high bandwidth, specialist network.
Wilkes3 offers the latest generation high performance computing and AI platform. At installation in 2021, it was the most powerful academic HPC system in the UK.
The specifics of the system include:
80x Dell PowerEdge XE8545 servers, each with
2x 3rd Generation AMD EPYC 64-Core CPUs
1 TB memory
4x NVIDIA A100-SXM-80GB GPUs
2x HDR 200 Gbit InfiniBand network.
Each server is capable of:
- 50 teraflops of FP64 performance (HPC)
1,200 teraflops of peak Tensor core FP16 performance (AI).
For further information about the service, please choose the option that best identifies your user needs:
