Rubin CPX: The Next Generation AI Compute Platform by Nvidia

Key Points:

• Rubin CPX offers 30 petaflops of compute power with NVFP4 precision and 128GB of GDDR7 memory.
• AI models processing video content can benefit from the high token generation capabilities of Rubin CPX.
• Nvidia introduces the Vera Rubin NVL 144 CPX rack, enabling AI service providers to increase profitability significantly.

Article:

Nvidia is revolutionizing the AI computing landscape with the introduction of the Rubin CPX, a cutting-edge platform that offers unparalleled performance and efficiency. The Rubin CPX features two dies with 25 petaflops per die, NVLink interconnect, and 288GB of HBM4 high-speed memory. In contrast, the Rubin CPX boasts one die with 30 petaflops of performance, 128GB of GDDR7 memory, and no NVLink, making it ideal for specific high-context applications that do not require extensive memory resources.

When it comes to processing video content, AI models can consume up to one million tokens per hour, translating to hours or even days of processing time. The Rubin CPX’s impressive token generation capabilities enable it to handle large-scale processing tasks efficiently, ensuring optimal performance and speed. With its 128GB of GDDR7 memory and NVFP4 precision, the Rubin CPX delivers three times faster attention capabilities compared to previous systems, making it a game-changer in the AI computing domain.

Nvidia is also unveiling the Vera Rubin NVL 144 CPX rack, a powerful solution designed to help AI service providers enhance their profitability drastically. This innovative rack configuration includes 144 Rubin CPX GPUs, 144 Rubin GPUs, and 36 Vera CPUs, offering 8 exaFLOPs of NVFP4 compute power, 100TB of fast memory, and 1.7 PB/s of memory bandwidth. According to Nvidia, the NVL 144 CPX rack is 7.5 times faster than the current top-of-the-line systems, providing unparalleled performance and scalability for AI applications.

In conclusion, the Rubin CPX and Vera Rubin NVL 144 CPX rack represent the next generation of AI compute platforms, offering unmatched performance, efficiency, and scalability for AI service providers. With their advanced capabilities and cutting-edge technology, these solutions are set to redefine the AI computing landscape and drive innovation in the industry.