Planck
System Overview
Planck is hybrid GPU-CPU supercomputer similar in architecture to the top hybrid GPU computers in the current Top 500 supercomputer list. The cluster was configured and purchased through USC NSF EPSCOR Track I funds.
- The Planck cluster combines a mixture of 264 CPU cores and 57 M2070/M1060 Nvidia GPGPU accelerator boards to form a hybrid supercomputer with a theoretical peak performance of 59 Teraflops.
- Planck has both shared and distributed types of memory: shared among cores on each node and on each GPU, and distributed among the nodes.
- There are in total 19 compute nodes (228 CPU cores), 9 NVIDIA M2070 (448 core, double precision) and 48 NVIDIA M1060 (240 core, single precision) GPU accelerators.
Node | Name | Count | Model | Processor | CPU Cores | Cache | Memory DDR 1333 | GPU | GPU Cores |
Login | c01-c02 | 2 | SL 390 G7 | 2.8 GHz Intel Xeon X5660 | 12 | 12+12 MB | 24 GB | 0 | 0 |
Storage | sn01 | 1 | DL 380 G7 | 2.8 GHz Intel Xeon X5660 | 12 | 12+12 MB | 24 GB | 0 | 0 |
Tesla Compute | ccrcis01-16 | 16 | SL 390 G7 | 2.8 GHz Intel Xeon X5660 | 12 | 12+12 MB | 24 GB | (3) M1060 | 240 |
Fermi Compute | ccrcis17-19 | 3 | SL 390 G7 | 2.8 GHz Intel Xeon X5660 | 12 | 12+12 MB | 24 GB | (3) M2070 | 448 |
- All the nodes are connected with fast communication, i.e. Infiniband 40 Gb/sec network.
- Data storage is based on 2 MSA 2312 SAS devices.
- 4 TB of scratch space (not backed up) and 11 TB of backed up storage.
Please contact rci@sc.edu for planck usage.