Accelerator Pooling Appliance – MI300x

Fully Managed Disaggregated Compute Accelerator Appliance for AI/DL, HPC, Visualization and Data Science Applications

Talk to an Expert

Highlights

Capacity

8x AMD Instinct MI300X 750W accelerators

High Performance

Ultra low latency 512Gb/s uplinks

High Bandwidth Memory Support

1.54TB Total (8x 192GB per MI300X)

Simplified Deployment

RESTful APIs and WebGUI

Ideal For

AI/ML training and inferencing clusters
High-performance computing environments
Data analytics acceleration
Composable infrastructure deployments

The GigaIO™ Accelerator Pooling Appliance – MI300X is the industry’s highest performing PCIe accelerator appliance, which fully supports PCIe Gen5 with up to 2.048Tb/s total bandwidth dedicated to host server upstream connections.

It delivers advanced provisioning and monitoring of accelerators and next-gen AMD Instinct socketed accelerators. Eight AMD Instinct MI300X 192GB 750W OAM GPUs are provided on the datacenter-grade universal baseboard (UBB) platform with a total of 1.54TB HBM memory.

This powerful expansion platform enables users to accelerate AI/ML training and inferencing clusters, high-performance computing environments, and data analytics applications.

Composable Disaggregated Infrastructure by dynamically provisioning and scaling PCIe devices

GigaIO’s AI fabric creates shared pools of accelerators in disaggregated GPU chassis connected by robust PCIe cables. Users can dynamically provision GPUs when needed and return them to the common pool when finished.

Intelligence for resource optimal utilization and failure prediction

GigaIO’s advanced RESTful API interface provides centralized management of GPU resources, as well as continuous monitoring for any faults for faster problem resolution and improved reliability.