Intel's Ponte Vecchio GPU will power the Aurora exascale supercomputer under a major U.S. government contract.

Intel today confirmed the identity of the GPU-based math-accelerator chip it will offer to supercomputer builders at the Supercomputing 2019 conference in Colorado, USA. The chip, named Ponte Vecchio after that medieval bridge in Italy, will be fabricated on Intel's forthcoming 7nm node rather than its troubled 10nm process. It will use Chipzilla's exascale Xe architecture and Foveros and EMIB packaging, and support high-bandwidth cache and memory and CXL-based interconnects.

Ponte Vecchio is described as a general-purpose GPU albeit one tuned for AI and applications found running on top-end supercomputers – think physics simulations and weather modeling and the like. It will not be available as a gaming graphics processor, though it could probably, perhaps, possibly run Crysis. According to Intel, it will offer high throughput of double-precision floating-point calculations and sport vector matrix math engines.

The chip is likely to arrive in 2021 as it will form the brains of the US government's $500m 1,000-petaFLOPS Aurora supercomputer, which will glue Ponte Vecchio accelerators to Intel Sapphire Rapids Xeon processors along with storage and other technology. Each Aurora compute node will sport dual 10nm Sapphire Rapids processors and six Ponte Vecchio chips, all using a unified memory architecture between the CPU and GPU cores. The Cray-built system will span more than 200 racks and have at least 10PB of RAM and 230PB of storage. This represents Intel's second attempt at Aurora: its earlier effort using many-core Xeon Phi chips caused it to miss the government's deadlines, forcing the company back to the drawing board.

Intel will also discuss its oneAPI programming interface for controlling systems that use a mix of CPU and GPU cores, FPGAs, and other silicon, which will be used in Aurora and serve as Intel's answer to Nvidia's CUDA. The company will use a CXL-based interconnect called Xe Link to connect Ponte Vecchio accelerators in Aurora and other systems, functioning similarly to Nvidia's NVSwitch. Additionally, Intel previously promised to ship 7nm data-center-grade GPUs in 2021, and users can request remote access to Intel hardware through Chipzilla's Devcloud to test and evaluate projects including oneAPI gear.