Computing is spreading across heterogeneous fabrics of CPUs, GPUs, application accelerators, interconnect processors, edge-computing devices and FPGAs – all of which require persistent memory and software to bind these elements into a complete solution. The race is on to generate, store and analyze data at zettascale levels. It took over 12 years to get from petascale to exascale computing. Intel has challenged itself to make it to zetta in five years: zetta 2027. Central to this goal is Intel’s work with the open ecosystem to ensure developers have optimized tools and software environments to accelerate their deployments.
Intel and SiPearl are collaborating to provide a joint platform for the first European exascale supercomputers. SiPearl is designing the microprocessor used in European exascale supercomputers and has selected Intel’s Ponte Vecchio graphics processing units (GPU) as the high performance computing (HPC) accelerator within the system’s HPC node. This partnership will offer European customers the possibility to combine SiPearl’s high-performance and low-power central processing unit (“Rhea”) with Intel’s family of general-purpose GPUs to make a high-performance compute node fostering European exascale deployments.
To enable this powerful combination, SiPearl plans to port and optimize oneAPI for the Rhea processor. As an open, standards-based programming model, oneAPI increases developer productivity and workload performance by providing a single programming solution across the heterogeneous compute node. The paired solution will also underline the value of Compute Express Link standardization in connecting compute elements, providing lower latency, coherent connectivity relative to PCIe connections.