Flexible Accelerator Card Configurations Optimized for Both Compute and Graphic Intensive Workloads
S5N | D43N-3U provides different GPU topologies for diverse use cases:
- Four 165W–300W GPUs to accelerate HPC, AI training, AI inference, cloud gaming and rendering
- Eight 70W–150W GPUs to provide a TCO-optimized virtual workstation solution for multi-stream medium spec cloud gaming, AI inferencing, and video analytics
- Eight 150W GPUs to support cost-effective AI training and mixed workloads involving AI inference
128 CPU Cores with up to 8TB Memory to Feed High Throughput GPUs
S5N | D43N-3U can support two AMD EPYC™ 7003 top bin CPUs up to 280W and 32x DIMM slots, which is powerful enough to feed four high throughput GPU cards such as NVIDIA A100 Tensor Core GPU.
It can deliver up to 45.6 Tera FLOPS HPL performance, which makes D43N-3U a TCO-optimized solution for data centers running HPC applications in molecular dynamics, life sciences and more fields.
Supports 4x A100/A30/A6000 with NVIDIA NVLink™ Bridges
Data Analytics - Big Database
(4) NVIDIA A100 Tensor Core GPU
Each pair of A100 – a combined 80GB of graphic memory
- 600 GB/s.
(4) NVIDIA A30 Tensor Core GPU
Each pair of A30 - a combined 48GB of graphic memory
- 200 GB/s
High Resolution Rendering
(4) NVIDIA RTX A6000
Each pair of A6000 - a combined 96GB of graphic memory
- 112 GB/s
