NVIDIA DGX System
Nvidia DGX servers and workstations use GPGPU to accelerate deep learning applications. They have a rackmount chassis with high-performance x86 server CPUs, except for DGX A100 and DGX Station A100, which use AMD EPYC CPUs. The main component is a set of 4 to 16 Nvidia Tesla GPU modules on an independent system board, integrated using a version of the SXM socket. DGX systems have powerful cooling systems to cool thousands of watts of thermal output.
Backend.AI is validated as the first NVIDIA DGX-Ready Software in Asia Pacific region.
Backend.AI on DGX-family
Complements to NVIDIA Container Runtime
- GPU sharing for multi-user support
- Features for machine learning pipeline
- Scheduling with CPU/GPU topology