ON5488M5
Designed for hyper-scale data center AI training applications
Key Features:
- Ultra-high GPU Density
- Energy Conservation
- Optimization Architecture for AI
Overview
This GPU resource pooling solution, PCI-e switch enabled and supporting up to 16GPU in 4OU form factor, delivers the highest density and performance available for AI training and inference scenarios. It provides flexible GPU typologies to meet different customer demands.
Ultra-high GPU Density
- 4OU 16*GPU for training or 16*FPGA for Inference
Energy Conservation
- Shared power & fans
- I/O front access for hyper-scale data center
Optimization Architecture for AI
- Accelerates training process
- Improved efficiency on computation expansion
Specifications
| Model | Aivres OCP AI Node |
|---|---|
| CPU | ATOM C3000 64bit CPU (2 cores) |
| Memory | Single channel, 1*DDR3 or DDR3L SODIMM |
| Storage | 1* SATA DOM, 1* SATA port |
| Ethernet | Integrated Ethernet GBE port |
| PCIe | 16 standard PCIe x16 slot to support 16 GPU card |
| I/O | 2* USB2.0 + 1* VGA + 1* COM |
| PSU | Bus Bar |
| Chassis | 536mm (W) x 181.9mm (H) x 800.6mm (D) |