Intel's Xe HPC (high performance compute) architecture powers the company's most powerful vector compute device to date, codenamed "Ponte Vecchio." The processor is designed for massive HPC and AI compute applications, but also features raster graphics and real-time raytracing hardware, giving it a dual-use as a cloud gaming GPU. Our Xe HPG architecture article covers the basics of how Intel is laying its client discrete GPUs out. The Xe HPC architecture both scales-up and scales-out from that. The Xe-core, the basic indivisible sub-unit, of the Xe HPC architecture is different from that of Xe HPG. While Xe HPG cores contain sixteen 256-bit vector engines alongside sixteen 1024-bit matrix engines; the Xe HPC cores features eight 512-bit vector engines, besides eight 4096-bit matrix engines. It also features higher load/store throughput, and a larger 512 KB L1 cache.