NVIDIA GRACE HOPPER SUPERCHIP REMOTE TEST

Solving the biggest AI and HPC problems requires high capacity, high bandwidth memory (HBM). The fourth generation of NVIDIA NVLink-C2C provides 900 gigabytes per second (GB/s) of bidirectional bandwidth between the NVIDIA Grace CPU and NVIDIA GPUs. The link provides a unified, cache-coherent memory address space that combines system and HBM GPU memory for simplified programmability. This coherent, high-bandwidth connection between CPU and GPUs is the key to accelerating tomorrow's most complex problems.

NVIDIA Grace is the first server CPU to leverage LPDDR5X memory with server-class reliability through mechanisms such as error correction code (ECC) to meet the needs of the data centre, while delivering two times higher memory bandwidth and up to ten times better energy efficiency compared to currently available server memory. The NVIDIA Grace CPU integrates ARM Neoverse V2 compute units into an NVIDIA-developed Scalable Coherency Fabric to deliver high performance in an advantageous design to facilitate the work of scientists and researchers.

Even though the parallel computing capabilities of GPUs continue to advance, workloads can still be limited by serial tasks running on the CPU. A fast and efficient CPU is a critical component of the system design to enable optimal workload acceleration. The NVIDIA Grace CPU integrates Arm Neoverse V2 cores to deliver high performance in a low-power system, making work easier for scientists and researchers.

Generative AI is memory and compute intensive. The NVIDIA GB200 superchip uses 380 GB of HBM memory, providing more than 4.5 times the GPU memory bandwidth of the NVIDIA H100 Tensor Core GPU. Grace Blackwell's high-bandwidth memory is combined with CPU memory via NVLink-C2C to provide nearly 800GB of fast-access memory for the GPU. This delivers the memory capacity and bandwidth required for the world's most complex generative AI and accelerated computing workloads.