Maximum Acceleration for AI / Deep Learning and HPC
MAXIMUM ACCELERATION A+ GPU SYSTEM
With Supermicro's advanced architecture and thermal design, including liquid cooling and custom heatsinks, our 2U, 4U GPU systems drive NVIDIA's latest HGX A100 4-GPU or 8-GPU baseboards, as well as supporting up to 8 directly attached double-width or single-width PCI-E GPU cards, including the latest NVIDIA A100 or AMD Instinct MI100 GPUs in a hyper-dense system.
Supermicro's unique AIOM and a slew of PCI-E 4.0 slots of these systems enhance the multi-GPU communication and high-speed data flow between systems at a large scale.
The H12 GPU systems feature the latest technology stacks such as 200G networking, NVIDIA NVLink and NVSwitch, 1:1 GPUDirect RDMA, GPUDirect Storage, and NVMe-oF on InfiniBand.
Featured Models:
AS -2114GT-DNR
2U 2 Node Single-Processor System with 3 GPUs per Node
AS -4124GO-NART
4U Dual-Processor GPU System with NVIDIA HGX A100 8-GPU, NVLink, and NVSwitch
AS -4124GS-TNR
4U Dual-Processor, Dual-Root GPU System with 8 PCI-E GPUs
- AI/ML
- Deep Learning Training and Inference
- High Performance Computing (HPC)
- Research Laboratory/National Laboratory
- Molecular Dynamics Simulation
- Astrophysics Simulation
- Chemistry Simulation
Dense and scalable multi-GPU powerhouse supporting latest GPUs
Directly attached PCI-E 4.0 x16 CPU to GPU lanes with low latency, NVIDIA NVLink and NVSwitch
Flexible AIOM/OCP 3.0 networking for up to 200G, GPUDirect RDMA and Storage