Nvidia introduces the GB200 NVL4, featuring two Grace CPUs and four Blackwell GPUs, designed for today’s data center tasks

Nvidia has introduced the GB200 NVL4, a cutting-edge platform designed to tackle modern data center demands and computational challenges. This launch is part of the company’s ongoing strategy to refine its product lineup, shifting focus from older NVL platforms to new, advanced solutions like the NVL4.

The GB200 NVL4 features a powerful combination of two Arm-based Grace CPUs and four Blackwell GPUs. Each Grace CPU includes 72 Arm Neoverse V2 cores, offering a total of 144 cores per system.

This setup delivers the processing strength required for intensive tasks like artificial intelligence (AI), high-performance computing (HPC), and other data-heavy applications. The four Blackwell GPUs further enhance the platform’s ability to handle demanding workloads efficiently.

Connectivity is another highlight of the GB200 NVL4. Each CPU comes equipped with six MCIO connectors, which ensure fast PCIe data transfer. These connectors support additional components like network interface cards (NICs) and solid-state drives (SSDs), providing the flexibility needed for scalable operations.

The platform also boasts up to 1.3 TB of combined memory, enabling smooth performance for memory-intensive applications.

Energy efficiency is a key focus of this design. Nvidia estimates that a fully configured GB200 NVL4 server consumes just over 6 kW of power. While this is a significant draw, it represents a notable improvement compared to earlier systems like the DGX-1 and HGX-1, which used around 3.5 kW.

This balance of performance and energy consumption positions the NVL4 as a practical choice for businesses aiming to optimize power usage without sacrificing capability.

As part of Nvidia’s Grace Blackwell family, the GB200 NVL4 serves as a mid-range option, sitting below the higher-performance NVL72 and NVL36. The NVL72, for example, doubles the computing power of the NVL4 and includes edge connectors for larger-scale deployments.

In contrast, the NVL4 offers a more compact and energy-conscious alternative, ideal for medium-sized operations or organizations focusing on efficiency.

The introduction of the GB200 NVL4 highlights Nvidia’s commitment to delivering advanced, flexible solutions for a variety of deployment scenarios. Whether for AI research, HPC, or other complex workloads, the platform is designed to meet the growing demands of modern data centers.

With its blend of robust performance and thoughtful energy management, the NVL4 is a strong contender for businesses looking to stay ahead in an increasingly competitive technological landscape.