NVIDIA has developed a new CPU, ‘Grace,’ that will power the world’s most powerful Artificial Intelligence-capable supercomputer. Grace will be used by the Swiss National Computing Center’s (CSCS) new system. It is an extensive Arm-based data center CPU developed by NVIDIA.
According to NVIDIA CEO Jensen Huang, the Swiss National Supercomputing Center will build a supercomputer, Alps, powered by Grace and their next-generation GPU. It will be made by Hewlett Packard Enterprise using the new HPE Cray EX supercomputer product line and the NVIDIA HGX supercomputing platform, including NVIDIA GPUs and the NVIDIA HPC SDK, as well as the new Grace CPU. The Alps is expected to play a crucial role in advancing science worldwide. It will replace CSCS’s existing Piz Daint supercomputer.
The Alps belongs to a new generation of machines expanding supercomputing beyond conventional modeling and simulation by taking advantage of GPU-accelerated deep learning. With the help of tight coupling between NVIDIA CPUs and GPUs, Alps is expected to train GPT-3, the world’s largest natural language processing model, in mere two days. It is almost 7x faster than NVIDIA’s 2.8-AI exaflops Selene supercomputer.
The Alps uses a novel software-defined infrastructure that can support a wide range of projects enabling different teams to use one or more partitions on a single, unified infrastructure rather than other machines.
CSCS users can apply this fantastic AI performance to a wide range of emerging scientific research ranging from climate and weather to astrophysics, life sciences, computational fluid dynamics, particle physics, as well as areas like economics and social sciences.
Grace will deliver 10x the fastest servers’ performance on the most complex AI and high-performance computing workloads. Grace is based on the hyper-efficient Arm microarchitecture found in billions of smartphones and other edge computing devices.
It is believed that Grace can support the next generation of NVIDIA’s coherent NVLink interconnect technology, thereby allowing data to move more quickly between system memory, CPUs, and GPUs.
CSCS has always supported scientists working at the cutting edge, especially in materials science, weather forecasting and climate modeling, and understanding data streaming in from a new generation of scientific instruments.
CSCS designs and operates a dedicated system for numerical weather predictions (NWP) on MeteoSwiss’s Swiss meteorological service. Its experience with NWP on GPUs will be critical to future climate simulations. It is essential not only to model long-term climate changes but also to build models to predict extreme weather events more accurately, saving lives. The team aims to run global climate models with a spatial resolution of 1 km to map convective clouds such as thunderclouds. Swiss scientists also use the CSCS supercomputer to analyze data from the Large Hadron Collider (LHC) at CERN.