Designed for the performance and efficiency required for modern AI data centers
Designed for the performance and efficiency required for modern AI data centers
The ever-increasing complexity of AI models and their requirements has greatly increased the importance of accelerated computing and energy efficiency in the data center. In this context, the NVIDIA Grace™ CPU marks a milestone. As a groundbreaking Arm® CPU, it sets new standards by delivering unrivaled performance and efficiency without compromise. The flexibility of the Grace CPU is remarkable: it can seamlessly interface with a GPU to optimize accelerated computing, or stand alone as a powerful and efficient CPU.
The versatility of the NVIDIA Grace CPU spans multiple configurations to meet the diverse needs of today's data centers. From high-performance servers to compute-intensive applications, it provides a solid foundation for next-generation data centers. With its ability to adapt to different deployment scenarios, the Grace CPU enables optimal utilization of resources, significantly increasing efficiency and improving performance.
Take a look at the Grace lineup
Take a look at the Grace lineup
NVIDIA GB200 NVL72
NVIDIA GB200 NVL72
The NVIDIA GB200 Grace Blackwell Superchip combines two NVIDIA Blackwell Tensor Core GPUs and a Grace CPU and can be scaled to the GB200 NVL72, a massive 72-GPU system connected via NVIDIA® NVLink®, to deliver 30x faster real-time inference for large-scale language models.
Read more
Read press release
NVIDIA Grace Superchip-CPU
The NVIDIA Grace Superchip CPU utilizes NVLink C2C technology to provide 144 Arm® Neoverse V2 compute units and 1 TB/s of memory bandwidth.
Learn more
NVIDIA Grace Hopper Superchip
NVIDIA Grace Hopper Superchip
NVIDIA Grace Hopper™ Superchip combines the Grace and Hopper architectures using NVIDIA® NVLink®-C2C to provide a coherent CPU and GPU memory model for accelerated AI and High-Performance Computing (HPC) applications.
Learn more
Discover Grace reference designs for
modern data center workloads
Discover Grace reference designs for
modern data center workloads
System designs for digital twins, artificial intelligence and high-performance computers.
System designs for digital twins, artificial intelligence and high-performance computers.
NVIDIA OVX™
for digital twins and NVIDIA Omniverse™.
NVIDIA Grace CPU Superchip
NVIDIA GPUs
NVIDIA BlueField®-3
NVIDIA HGX™
for HPC.
NVIDIA Grace CPU Superchip
NVIDIA BlueField-3
OEM-defined input/output (IO)
NVIDIA HGX
for AI training, inference, and HPC.
NVIDIA Grace Hopper Superchip CPU + GPU
NVIDIA BlueField-3
OEM-defined IO / fourth-generation NVLink