Louisa Crawford
2 Aug 2024 15:21
NVIDIA’s Grace CPU family aims to meet the growing demands of data processing with high efficiency by leveraging Arm Neoverse V2 cores and a new architecture.
According to the NVIDIA Technical Blog, the exponential growth in data processing demand is expected to reach 175 zettabytes by 2025. This surge stands in stark contrast to the slowing pace of CPU performance improvements, underscoring the need for more efficient computing solutions.
Solving Efficiency with NVIDIA Grace CPU
NVIDIA’s Grace CPU family is designed to solve these challenges. The first CPU developed by NVIDIA to drive the AI era, Grace CPU features 72 high-performance, power-efficient Arm Neoverse V2 cores, NVIDIA Scalable Coherency Fabric (SCF), and high-bandwidth, low-power LPDDR5X memory. It also boasts a 900 GB/s coherent NVLink chip-to-chip (C2C) connection to NVIDIA GPUs or other CPUs.
Grace CPUs support multiple NVIDIA products and can be paired with NVIDIA Hopper or Blackwell GPUs to form a new type of processor that tightly combines CPU and GPU capabilities. This architecture aims to power generative AI, data processing, and accelerated computing.
Next-Generation Data Center CPU Performance
Data centers face power and space constraints and require infrastructure that delivers maximum performance with minimum power consumption. The NVIDIA Grace CPU Superchip is designed to meet these requirements, delivering outstanding performance, memory bandwidth, and data movement capabilities. This innovation promises significant gains in energy-efficient CPU computing for data centers, supporting essential workloads such as microservices, data analytics, and simulations.
Customer Adoption and Momentum
Customers are rapidly adopting the NVIDIA Grace family for a variety of applications, including generative AI, hyperscale deployments, enterprise computing infrastructure, high-performance computing (HPC), and scientific computing. For example, NVIDIA Grace Hopper-based systems deliver 200 exaflops of energy-efficient AI processing power in HPC.
Organizations such as Murex, Gurobi, and Petrobras are experiencing outstanding performance results in financial services, analytics, and energy, demonstrating the benefits of NVIDIA Grace CPUs and NVIDIA GH200 solutions.
High-performance CPU architecture
NVIDIA Grace CPUs are designed to deliver exceptional single-threaded performance, ample memory bandwidth, and superior data movement capabilities, while significantly improving energy efficiency over traditional x86 solutions.
This architecture integrates several innovations, including NVIDIA Scalable Coherency Fabric, server-grade LPDDR5X with ECC, Arm Neoverse V2 cores, and NVLink-C2C. These features ensure that the CPU can efficiently handle demanding workloads.
NVIDIA Grace Hopper and Blackwell
The NVIDIA Grace Hopper architecture combines the performance of NVIDIA Hopper GPUs with the versatility of NVIDIA Grace CPUs on a single superchip. This combination is connected by a high-bandwidth, memory-coherent, 900 GB/s NVIDIA NVLink Chip-2-Chip (C2C) interconnect, delivering 7x the bandwidth of PCIe Gen 5.
The NVIDIA GB200 NVL72 combines 36 NVIDIA Grace CPUs and 72 NVIDIA Blackwell GPUs in a rack-scale design to deliver exceptional acceleration for generative AI, data processing, and high-performance computing.
Software Ecosystem and Porting
NVIDIA Grace CPUs are fully compatible with the extensive Arm software ecosystem, allowing most software to run without modification. NVIDIA also extends the software ecosystem for Arm CPUs, providing high-performance math libraries and optimized containers for a wide range of applications.
For more information, see the NVIDIA Technology Blog.
Image source: Shutterstock