Designing and deploying high-performance computing systems that meet the demands of today’s data- and compute-intensive workloads is a complex undertaking. Scaling performance through sheer server count is no longer practical—systems requiring hundreds of thousands of nodes are financially, physically, and environmentally unsustainable. Instead, modern HPC design increasingly relies on fewer, more powerful compute units built using technologies like Intel® Xe-HPC GPUs and other accelerators. While this enables significant performance gains, it also introduces new challenges in software development. Applications must now run efficiently across heterogeneous architectures, requiring new approaches to programming, performance portability, and workload optimisation.
The Cambridge Open Zettascale Lab (COZL) serves as one of Dell Technologies’ global HPC & AI Centers of Excellence. In collaboration with the wider HPC community, these Centers foster a connected network of expertise—drawing on the collective knowledge of technology developers, service providers, and industry specialists to accelerate innovation in high-performance computing and AI.
A core focus at the Lab is Intel’s oneAPI, an open, unified programming model that supports development across multiple compute architectures—including CPUs, GPUs, and FPGAs. oneAPI aims to simplify development while delivering high performance and eliminating the constraints of proprietary environments. It also supports the integration of existing codebases, making it an attractive solution for evolving HPC workloads.
Another key toolset is the Intel® oneAPI Rendering Toolkit, a suite of open-source libraries for building high-performance, high-fidelity visualisation applications. It includes components such as Intel® Embree, Intel® Open Image Denoise, Intel® OSPRay, and others—each designed to provide rendering capabilities optimised for Intel CPUs and upcoming Xe GPU architectures. These tools are widely used in scientific visualisation, simulation outputs, and digital twin environments.
Aims
At COZL, we are actively engaged in porting a range of major HPC applications to the oneAPI framework, including CASTEP, FEniCS, and AREPO—tools used in materials science, computational physics, and cosmology. The Lab also serves as a hub for oneAPI training, offering workshops, short courses, and expert-led instruction delivered by certified Zettascale Software Engineers. We are committed to growing a skilled community around open, cross-architecture development for the next generation of high-performance computing.
This workshop will focus on the use of DPC++/OpenMP for offloading on to GPUs.
On 10-11th May 2023, our oneAPI Centre of Excellence instructors will be running a practical introduction to Intel’s oneAPI in Daresbury for STFC.
A hands-on, practical workshop where you will learn the fundamentals of programming in a heterogeneous environment.
This workshop will focus on the use of DPC++/OpenMP for offloading on to GPUs.
Learn the fundamentals of programming in a heterogeneous environment.
This two-part workshop will focus on the use of DPC++/OpenMP for offloading on to GPUs.
This two-part workshop will focus on the use of DPC++/OpenMP for offloading on to GPUs.
Learn the fundamentals of programming in a heterogeneous environment.
Learn the fundamentals of programming in a heterogeneous environment.
Arjen Tamerus, senior research software engineer at COZL presents remotely his paper on oneAPI and CASTEP in the SYCL session at the Durham HPC/AI Days conference.
The Durham HPC/AI Days began this week. Dawn pioneer users will speak about their positive experiences on the UK's fastest AI supercomputer. Later in the week, oneAPI Centre of Excellence trainer and COZL research software engineer Arjen Tamerus will give a talk on SYCL.
Zettascale Software Engineer Arjen Tamerus will give a talk at the Durham HPC/AI Days at Durham University in May. His paper will cover the work of our oneAPI Centre of Excellence.