GigaIO FabreXT AI Memory Fabric Platform

The FabreX™ AI Memory Fabric Platform by GigaIO is a next-generation, memory-centric fabric designed to revolutionize data center architectures in response to the exponential growth of data and the rapid adoption of advanced analytics and Artificial Intelligence (AI). By disaggregating traditional server components and enabling dynamic composition of resources, FabreX addresses the challenges posed by modern compute and storage clusters, offering unparalleled flexibility, performance, and efficiency. Key Features and Functionality: - Memory-Centric Fabric: FabreX connects memory, storage, and a wide array of accelerators—including GPUs, FPGAs, and custom ASICs—either directly or via configurations like NVMe-oF, delivering industry-leading low latency and high bandwidth. - High Performance: With latency from system memory of one server to another being less than 200 nanoseconds and bandwidth scaling up to 512 Gbits/sec in its Gen4 implementation, FabreX ensures true PCIe performance across entire clusters. - Unmatched Flexibility: The platform enables the composition of diverse resources, such as GPUs, DPUs, TPUs, FPGAs, SoCs, NVMe storage, and other I/O devices, across multiple servers and racks. It supports device-to-node, node-to-node, and device-to-device communication within the same high-performance PCIe memory fabric. - Open Standards Compliance: FabreX is 100% PCI-SIG compliant, ensuring seamless integration with heterogeneous computing, storage, and accelerator components into a unified system-area cluster fabric. Primary Value and User Solutions: FabreX addresses the critical need for scalable, flexible, and efficient data center architectures capable of handling the demands of AI, Machine Learning (ML), and Deep Learning (DL) applications. By disaggregating server components and enabling dynamic resource composition, it eliminates bottlenecks and configuration challenges inherent in traditional interconnect systems. This approach not only enhances performance but also optimizes resource utilization, reducing the total cost of ownership and allowing data centers to scale both up and out seamlessly.