NVIDIA
NVIDIA GH200 Grace Hopper Superchip
Grace Hopper platform combining an Arm-based NVIDIA Grace CPU with a Hopper GPU over NVLink-C2C, with high-bandwidth coherent CPU-GPU memory access.
- Architecture
- Grace Hopper
- Memory
- 96 GB HBM3 or 144 GB HBM3e GPU memory
- Memory bandwidth
- Up to 4.9 TB/s GPU memory bandwidth
- Release
- Introduced 2023
Main use cases
Large-model AI, memory-heavy inference, HPC and tightly coupled CPU-GPU workloads
Cloud pricing
Full tracker →No tracked rates for this accelerator yet.
Usually offered as full systems or specialized cloud instances rather than as a standalone GPU rental.
NVIDIA GH200 Grace Hopper Superchip in the news
GPU news →No coverage yet.