CARLSBAD, Calif.: #AIFabric--GigaIO, a pioneer in scalable edge-to-core AI platforms for all accelerators that are easy to deploy and manage, will showcase its latest innovations at ISC High Performance 2025, taking place June 10-13 in Hamburg, Germany. Visitors to stand H22 can see how GigaIO’s revolutionary AI fabric technology, which seamlessly bridges from edge to core with a dynamic, open platform built for any accelerator, powers its two flagship products, SuperNODE and Gryf.

SuperNODE is the world’s most powerful and energy-efficient scale-up AI computing platform, and Gryf is the first suitcase-sized AI supercomputer that brings datacenter-class computing power directly to the edge. GigaIO’s architecture, powered by its AI fabric, effortlessly integrates GPUs and inference accelerators from NVIDIA, AMD, d-Matrix, and more, enabling organizations to slash power and cooling requirements by up to 30% without compromising performance.

GigaIO’s AI fabric implements a native PCIe Gen5 architecture that enables direct memory-semantic communication between distributed computing resources, eliminating protocol translation overhead while maintaining sub-microsecond latencies for GPU-to-GPU transfers. This enables AI workloads to achieve near-linear scaling across pooled accelerators that appear as if locally attached to the host.

GigaIO’s groundbreaking paper, “Rail Optimized PCIe Topologies for LLMs,” was selected for presentation at ISC 2025 on Thursday, 12 June 2025, from 9:00am to 9:25am in Hall F (2nd floor). This research explores optimized network architectures for large language model training and inference. Scaling LLMs efficiently requires innovative approaches to GPU interconnects, and GigaIO’s rail-optimized, PCIe-based AI fabric topologies offer up to 3.7x improved collective performance with an accelerator-agnostic design, ensuring adaptability across diverse AI workloads.

“ISC 2025 arrives at a critical juncture, as AI workloads demand unprecedented hardware resources, making optimized infrastructure essential for organizations to achieve their performance targets,” said Alan Benjamin, CEO of GigaIO. “Our expanded conference participation will demonstrate how our PCIe-based fabric technology delivers superior performance for LLM training and inference, while dramatically reducing power consumption and total cost of ownership.”

Stop by stand H22 at ISC 2025 or schedule a meeting during the event with the GigaIO team.

About GigaIO

GigaIO redefines scalable AI infrastructure, seamlessly bridging from edge to core with a dynamic, open platform built for every accelerator. Reduce power draw with GigaIO’s SuperNODE, the world’s most powerful and energy-efficient scale-up AI computing platform. Run AI jobs anywhere with Gryf, the world’s first suitcase-sized AI supercomputer that brings datacenter-class computing power directly to the edge. Both are easy to deploy and manage, utilizing GigaIO’s patented AI fabric that provides ultra-low latency and direct memory-to-memory communication between GPUs for near-perfect scaling for AI workloads. Visit www.gigaio.com, or follow on Twitter (X) and LinkedIn.

Fonte: Business Wire