Taiwanese tech giants adopt the world’s first NVIDIA Grace CPU architectures

NVIDIA has announced that leading computer manufacturers in Taiwan will release the first wave of systems powered by the NVIDIA Grace™ CPU Superchip and Grace Hopper Superchip technology for a wide range of workloads including digital twins, artificial intelligence, high-performance computing, cloud graphics, and gaming.

Dozens of server models from ASUS, Foxconn Industrial Internet, GIGABYTE, QCT, Supermicro and Wiwynn are expected in the first half of 2023. Grace-powered systems will join x86 and other Arm-based servers to provide customers with a variety of options to achieve high performance and efficiency in your data centers.

said Ian Buck, vice president of Hyperscale and HPC at NVIDIA. “Our partners’ new systems, powered by Grace Superchips technology, will bring the power of computational acceleration to new markets and industries around the world.”

The upcoming servers are based on four new system architectures featuring the Grace CPU Superchip and the Grace Hopper Superchip, which NVIDIA announced at its two most recent GTC conferences. 2U form factor designs provide server and board layouts to OEMs and OEMs to quickly bring systems to market for NVIDIA CGX™ cloud gaming, NVIDIA OVX digital platforms and platforms. AI and HPC NVIDIA HGX™.

Running modern workloads

Two NVIDIA Grace Superchip technologies enable a wide range of compute-intensive workloads across multiple system architectures:

  • Grace CPU Superchip features two CPU chips, cohesively connected via an NVLink®-C2C link, with up to 144 high-performance Arm V9 cores with scalable vector extensions and a 1TB/sec memory subsystem. Innovative design delivers the highest performance, twice the memory bandwidth, and power efficiency of today’s leading server processors to handle the most demanding HPC, data analytics, digital twin, cloud gaming and high-bandwidth computing applications;
  • The Grace Hopper Superchip combines the NVIDIA Hopper™ GPU and the NVIDIA Grace CPU via NVLink-C2C into an integrated unit designed to handle large-scale HPC and AI applications. Using NVLink-C2C interconnection, the Grace CPU transmits data to the Hopper GPU 15 times faster than conventional CPUs.

Grace wide server suite for AI, HPC, digital twins and cloud gaming

The Grace CPU Superchip and Grace Hopper Superchip server design suite includes systems available on single motherboards with single, dual, and four-way configurations available in four workload designs that can be customized by server manufacturers according to customer needs:

  • NVIDIA HGX Grace Hopper AI Training and Heuristics and HPC Systems are available with Grace Hopper Superchip Modules and NVIDIA BlueField®-3 DPU;
  • NVIDIA HGX Grace HPC and Supercomputing systems feature a CPU-only design with Grace CPU Superchip and BlueField-3;
  • NVIDIA OVX systems for digital twins and collaborative workloads feature Grace CPU Superchip, BlueField-3, and NVIDIA GPUs;
  • NVIDIA CGX Cloud Graphics and Gaming Systems Featuring Grace CPU Superchip, BlueField-3 and NVIDIA A16

NVIDIA is expanding its NVIDIA™ Certified Systems Program to include servers using the NVIDIA Grace CPU Superchip and Grace Hopper Superchip, as well as x86 CPUs. The first OEM server certifications are expected soon after the delivery of partner systems.

Compatible Software

Grace’s family of servers is optimized for NVIDIA’s advanced computing software packages, including NVIDIA HPC, NVIDIA AI, Omniverse™, and NVIDIA RTX™.

To learn more about NVIDIA Grace, watch the presentation on Computex.

Leave a Comment