Meta and Oracle adopt NVIDIA Spectrum-X Ethernet switches for AI data centers, achieving 95% data throughput vs 60% for standard Ethernet. Major boost for hyperscale AI infrastructure.
Key Highlights:
- Meta integrating Spectrum-X Ethernet into Facebook Open Switching System platform
- Oracle building giga-scale AI supercomputers with Spectrum-X technology
- Platform delivers 95% data throughput compared to 60% for standard Ethernet
- Technology enables interconnection of millions of GPUs for AI training
- Deployment accelerates training efficiency and reduces time to insights
NVIDIA Spectrum-X Powers Next Generation AI Data Centers
NVIDIA announced today that Meta and Oracle will standardize on Spectrum-X Ethernet networking switches to accelerate their AI data center infrastructure. The move represents a significant shift toward open, accelerated networking architecture designed specifically for trillion-parameter AI models.
NVIDIA CEO Jensen Huang described Spectrum-X as the nervous system of the AI factory, enabling hyperscalers to connect millions of GPUs into a single giant computer. The technology addresses growing demands from massive language models that are transforming traditional data centers into giga-scale AI factories.
Record-Breaking Network Performance for AI Workloads
Spectrum-X Ethernet has demonstrated breakthrough efficiency in large-scale deployments, achieving 95% data throughput with its congestion control technology. This represents a dramatic improvement over standard Ethernet solutions, which typically suffer from thousands of flow collisions that limit throughput to roughly 60% at scale.
The platform features advanced congestion control, adaptive routing, and AI-driven telemetry capabilities specifically optimized for massive AI training and inference clusters. These innovations deliver unprecedented efficiency and predictability for workloads requiring millions of interconnected GPUs.
Meta Deploys Spectrum Switches for FBOSS Platform
Meta will integrate Spectrum Ethernet switches into its networking infrastructure for the Facebook Open Switching System, the software platform the company developed to manage network switches at massive scale. Gaya Nagarajan, Meta’s vice president of networking engineering, stated the company’s next-generation AI infrastructure requires open and efficient networking at a scale the industry has never seen before.
The integration will speed deployment across Meta’s data centers, unlock gains in AI training efficiency, and shorten time to insights for research and product development.
Oracle Builds Giga-Scale AI Supercomputers
Oracle plans to construct giga-scale AI factories accelerated by NVIDIA’s Vera Rubin architecture and interconnected by Spectrum-X Ethernet. Mahesh Thiagarajan, Oracle Cloud Infrastructure’s executive vice president, emphasized the partnership extends Oracle’s AI leadership by allowing the company to interconnect millions of GPUs with breakthrough efficiency.
The deployment will help Oracle customers train, deploy, and benefit from the next wave of generative and reasoning AI models more quickly than previously possible.
Spectrum-XGS Enables Cross-Data Center AI Super-Factories
NVIDIA’s Spectrum-XGS Ethernet technology expands the platform’s capabilities beyond single data centers, enabling scale-across functionality to link facilities across cities, nations, and continents into vast AI super-factories. This advancement allows companies to create unified AI computing systems from geographically dispersed infrastructure, maximizing resource utilization for training the largest models ever built.
Ainewshub
Discover more from AI News Hub
Subscribe to get the latest posts sent to your email.
