Intel debuts Ponte Vecchio data centre GPUs

Data centre GPUs to compete with existing offerings served up by Nvidia and AMD

Intel has finally unveiled the company’s long-awaited GPU processor architecture, designed to tackle the onslaught of large data and AI workloads entering the data centre.

At AI Supercomputing 2019, taking place in Denver, the chipmaker debuted its new Ponte Vecchio GPUs, which will compete with existing offerings served up by Nvidia and AMD.

Although the accelerators will initially target the data centre market, Intel said the new architecture will eventually form the basis of its consumer chips. The data centre acceleration marketed is expected to be worth $35 billion by the end of 2025, according to Market Study Report.

The Ponte Vecchios will be manufactured using Intel’s 7nm fabrication process (set for full phase roll out in 2021) and based on the company’s Xe architecture. Going forward, Xe will be the standard architecture for all Intel products and divided into microarchitectures for specific use cases.

Ponte Vecchio’s microarchitecture for analysing the massive data sets in HPC/AI workloads comprises a flexible data-parallel vector matrix engine, high double precision floating point throughput and ultra-high cache and memory bandwidth, Ari Rauch, vice president of Intel Architecture, Graphics and Software, told CRN.

“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,” added Raja Koduri, senior vice president, chief architect, and general manager of architecture, graphics and software at Intel.

In addition to the accelerators, Intel unveiled its next-generation 10nm Xeon Scalable processors, Sapphire Rapids, and a new software initiative, oneAPI, designed to make it easier for developers to code applications that function across a variety of processing architectures. oneAPI’s beta release supports Intel Xeon Scalable processors, Intel Core processors with integrated graphics, and Intel FPGAs, with further hardware support to follow.

Intel revealed the U.S Department of Energy’s Aurora supercomputer will leverage two 10nm-based Intel Xeon Scalable processors, six Ponte Vecchio GPUs and use oneAPI for the programming layer.

“The oneAPI initiative Intel launched today will define programming for an increasingly AI-infused, multi-architecture world,” Intel said in a press release. “oneAPI delivers a unified and open programming experience to developers on the architecture of their choice without compromising performance and eliminating the complexity of separate code bases, multiple-programming languages, and different tools and workflows.”