The exponential growth of artificial intelligence and high-performance computing demands a new class of network infrastructure designed for massive scale and deterministic performance. Addressing this need, FS has introduced its PicOS® AI Switch System, a comprehensive networking portfolio engineered specifically for large-scale AI training, inference, and HPC environments. By integrating high-performance Broadcom silicon with its robust operating system and management platform, FS provides a solution that ensures lossless networking, ultra-low latency, and high GPU utilization in demanding data center clusters.
FS has launched the PicOS® AI Switch System, a networking portfolio for AI and HPC data centers.
The system integrates Broadcom Tomahawk series chips with FS's PicOS® and AmpCon-DC management platform.
It delivers lossless RoCEv2 networking, which is critical for maintaining high GPU efficiency in clusters.
The portfolio includes a range of 400G and 800G switch models to meet diverse bandwidth and density requirements.
Built-in features like advanced congestion control and redundant architecture ensure consistent, reliable performance.
The AmpCon-DC platform simplifies the deployment and management of large-scale GPU networks.
As AI models grow larger and more complex, the network fabric connecting GPU clusters becomes a critical bottleneck. The FS PicOS® AI Switch System is designed to eliminate this bottleneck. Powered by Broadcom Tomahawk 3, 4, and 5 series chips, the switches provide the high bandwidth and deep buffers necessary to handle the intense east-west traffic patterns of distributed AI training jobs. This foundation ensures lossless performance using RoCEv2, preventing network congestion from stalling valuable GPU compute cycles.
FS offers a complete range of switches to fit various deployment scales and performance tiers. The portfolio includes models like the N9600-64OD with 64x 800G ports for the most demanding spine applications, down to 32-port 400G models for leaf-tier deployments. This flexibility allows organizations to build scalable, non-blocking fabric architectures that can grow with their computational needs, from initial clusters to exascale-level infrastructures, all managed under a unified operating system.
Beyond hardware, the solution is defined by its integrated software stack. The PicOS® network operating system provides the intelligent traffic optimization and granular control needed for a predictable network. Coupled with the AmpCon-DC management platform, it delivers much-needed operational simplicity. This combination enables data center operators to automate deployment, monitor cluster health, and manage the entire network lifecycle from a single pane of glass, significantly reducing operational overhead and complexity.
The launch of the PicOS® AI Switch System positions FS as a key enabler for the next generation of AI data centers. By providing a tightly integrated solution of hardware, software, and management, FS empowers enterprises and cloud providers to build high-performance, scalable, and efficient networking infrastructures. This allows them to fully leverage their GPU investments and accelerate innovation by ensuring their network can keep pace with the relentless demands of AI and HPC workloads.
FS is a trusted provider of ICT products and solutions to enterprise customers worldwide. Established in 2009, the company focuses on HPC, Data Center, Enterprise, Telecom, providing tailored product development and solution design based on professional customer needs. Leveraging dedicated R&D and testing teams, comprehensive technical service experts, a robust supply chain system, globalized warehousing centers, and convenient shopping platform, FS delivers a wide range of highly efficient customer-centric ICT products, solutions and services to global vertical industry and enterprise customers across ISP, telecom, retail, education, etc.