Inside NVIDIA Rubin: Six Chips, One AI Supercomputer

The NVIDIA Rubin Platform is a groundbreaking development in AI infrastructure, designed to support the demanding needs of modern AI factories. Unlike traditional data centers, these AI factories require continuous, large-scale processing capabilities to handle complex reasoning and multimodal pipelines efficiently. The Rubin Platform integrates six new chips, including specialized GPUs and CPUs, into a cohesive system that operates at rack scale, optimizing for power, reliability, and cost efficiency. This architecture ensures that AI deployments can sustain high performance and efficiency, transforming how intelligence is produced and applied across various industries. Why this matters: The Rubin Platform represents a significant leap in AI infrastructure, enabling businesses to harness AI capabilities more effectively and at a lower cost, driving innovation and competitiveness in the AI-driven economy.

The evolution of AI into an industrial phase has led to the creation of AI factories, which are continuously operational systems that convert power, silicon, and data into intelligence at scale. These AI factories are not just about performing discrete AI model training or human-facing inference; they are about sustaining complex workflows and multimodal pipelines that require long-context reasoning and real-time inference. The NVIDIA Rubin platform is a response to this new reality, designed to handle the demands of AI factories by treating the data center as the unit of compute rather than a single GPU server. This approach ensures that performance and efficiency are maintained in real-world deployments, rather than just in isolated benchmarks. The Rubin platform is built on the principle of extreme co-design, where GPUs, CPUs, networking, security, software, power delivery, and cooling are architected together as a single system. This integration allows for predictable, secure, and efficient production of intelligence at scale. The platform’s flagship, the Vera Rubin NVL72 rack-scale system, operates as a coherent machine within a larger AI factory, optimized for sustained intelligence production. This means that the system is not just about peak performance but about maintaining predictable latency, high utilization, and efficient power conversion into usable intelligence. At the heart of the Rubin platform are six new chips, each engineered for a specific role in the AI factory. These chips include the NVIDIA Vera CPU, Rubin GPU, NVLink 6 switch, ConnectX-9 networking interface, BlueField-4 data processing unit, and Spectrum-6 Ethernet switch. Together, these components form a synchronized architecture that supports the execution of transformer-era workloads. The Rubin GPU, in particular, serves as the execution engine, designed for continuous training, post-training, and inference, optimizing the full execution path that turns power, bandwidth, and memory into tokens at scale. The significance of the Rubin platform lies in its ability to sustain performance and lower costs per token at scale, which is crucial for the next generation of AI factories. By integrating compute, networking, and infrastructure directly into the architecture, the Rubin platform ensures that communication, coordination, security, and efficiency are prioritized. This matters because as AI continues to evolve and expand, the ability to efficiently and securely produce intelligence at scale will determine the competitiveness and impact of AI-driven applications across various industries.

Read the original article here