HIGH PERFORMING COMPUTING INFRASTRUCTURE
The OR-X features a DGX POD, a reference architecture developed and tested specifically for running AI and deep learning applications:
- DGX A100 P3687 System
- 8x NVIDIA A100 80 GB Tensor Core GPUs
- 320 GB of GPU Memory
The Fast-Storage NetApp AFF A400 high-availability cluster is designed for high-performance reloading of experiments into the DGX servers. Data generated before, during, or after an experiment can be stored securely and made available for further processing at an extremely high level of performance:
- NetApp AFF A400
- Capacity RAW / Usable: 92.16TB / 56.09 TiB
- Workload Capacity: 30TiB
- Throughput / IOPS: 10000 IOPS
Possible connections from OR and Skills Lab are: SDI, HDMI, XLR, Ethernet, 3.5 mm audio, USB (2.0, 3.0). Each of the single signal encoding hubs is connected via 10 Gbit/s SFP+ to the collection switch. The network of the HPC server supports a bandwidth of 100 GB/s.