Nvidia, HPE, and the University of Bristol unveil Isambard 3, a supercomputer using 384 Nvidia Grace CPUs, each with 144 cores, reaching speeds of 2.7 petaflops
Nvidia Corp. today detailed three initiatives in which researchers will use its chips to support scientific discovery.
Context & Ripple Effects
Isambard 3 is an early deployment milestone for Nvidia's Grace platform, following the Grace CPU Superchip's introduction as a 144-core Arm data-center processor. It moves that CPU from product announcement into a named university-led scientific system built with HPE.
The project also extends Nvidia's established role in research computing after its work on Perlmutter, an AI-focused supercomputer at NERSC. The distinction here is the emphasis on Grace CPUs, rather than a GPU-led configuration alone.
First-order effects
- The University of Bristol gains a new high-performance computing resource for scientific workloads, while HPE gets a public deployment reference for systems based on Nvidia Grace CPUs.
- Nvidia receives a concrete validation point for Grace in a research-supercomputing environment, with 384 processors deployed in one system.
Second-order effects
- HPE and other HPC integrators can use the deployment to assess and market Arm-based CPU configurations for research customers that have traditionally bought established server architectures.
- The system gives scientific-computing buyers another data point when comparing CPU-centric and heterogeneous architectures, increasing pressure on platform vendors to demonstrate software readiness alongside raw performance.
Third-order effects
- If comparable deployments broaden, HPC procurement could shift further toward heterogeneous systems in which CPUs, accelerators, memory, and software are selected as an integrated stack rather than as interchangeable server parts.
- That shift would make ecosystem support—compilers, applications, and system integration—a more durable competitive factor for Arm-based data-center CPUs, though a single installation does not establish broad adoption.
The trend: Isambard 3 is part of the move toward heterogeneous, tightly integrated compute platforms for scientific and AI workloads.