Intel unveils new GPU architecture and oneAPI software stack for High-Performance Computing and AI Acceleration

Intel has now revealed a software stack, OneAPI, which is a scalable and unified programming model designed for the high-performance Artificial Intelligence and computing-based solutions.

Intel revealed the stack at Supercomputing 2019 event held in Colorado, US. The company is also making additions to its data-centric silicon portfolio along with a new software initiative that represents a move away from the single vendor and single architectures programming models.

With the launch of OneAPI, Intel claims that the OneAPI initiative to deliver a simplified and unified programming model for application management in the heterogeneous processing architectures.

The supported architectures include the likes of the GPU’s, CPU’s, Field Programmable gate array, and some other similar accelerators.

Executive Opinion

Raja Koduri, senior vice president, chief architect, and general manager of architecture, graphics, and software at Intel, said, "HPC and AI workloads demand diverse architectures, ranging from CPUs, general-purpose GPUs and FPGAs, to more specialized deep-learning NNPs, which Intel demonstrated earlier this month. Simplifying our customers’ ability to harness the power of diverse computing environments is paramount, and Intel is committed to taking a software-first approach that delivers a unified and scalable abstraction for heterogeneous architectures."

Building the Foundation of Intel OneAPI

OneAPI platform will offer the developers the power to work on the architecture of their choice while addressing the issues of separate codebases, multiple programming languages, and different tools. Intel also stated that the OneAPI will preserve the existing software investments while still offering the flexibility to create versatile applications.

Along with the OneAPI, Intel also announced a new category of general-purpose Graphics Process Units based on the Intel X architecture.

Called by the name as “Ponte Vecchio,” the GPU is built for the high-performance simulation workloads, computing modeling, and AI training. Ponte Vecchio is built on the Intel 7nm technology and will use the Foveros 3D and EMIB packaging innovations, notes the announcement.

The GPUs also feature extra technology such as the Compute Express Link Interconnect, high bandwidth memory, and some other intellectual properties.