AI Training Platform
ADLINK’s new AI Training Platform offers a carrier-grade, multi-faceted platform ideal for edge deployments. Combining a Server node, plus PCIe expansion box with PCIe switching, it provides capability to support up to eight NVIDIA GPGPUs. This configurable edge platform is ideal for machine/deep learning. ADLINK AI Training Platform supports multiple topologies and bandwidths between GPUs and CPUs implemented via cable routing options. Infiniband support allows scaled multi GPU clusters.
Range features
A high level overview of what this range offers
- Dual-socket Intel® Xeon® Scalable Processors with 2 PCIen switches
- 24x DDR4 DIMM slots
- 14x hot-swappable 2.5” SATA 3.0 6Gb/s drive bays
- 10x double-wide PCIe slots for GPUs
- 1x PCIe x16 slot for NIC up to 100Gb/s
- 1x PCIe x16 storage mezzanine slot
- AST2500 BMC with Redfish support
- Supports the latest GPU technologies, up to 8 NVIDIA Tesla or 3rd party PCIe x16 acceleration cards
- 10GbE onboard + OCP 2.0 to support 100Gb networking options
- Supports single and dual root complex design
- Hot-swappable fans, drives and power supply units
- Flexible system built for Edge High Performance Computing (EHPC)
What’s in this range?
All the variants in the range and a comparison of what they offer
| Model | CPU | RAM | Key features |
|---|---|---|---|
ALPS-4800 | Dual Intel® Xeon® Scalable processors | 24x DDR4-2666 RDIMM | 8x PCIe x16 Gen3 GPGPU slots (300W/slot) Validated with NVIDIA® Tesla® P100/V100 accelerators Dual Intel® Xeon® Scalable processors 24x DDR4-2666 RDIMM 8x SATA 6Gb/s hot-swappable 2.5" drives 1x FHFL and 2x Low Profile PCIe x16 Gen3 add-on slots 1600W AC/DC Platinum PSU, 3+1 redundancy Separate airflow for CPU and GPU |
