Intel and Lenovo Team Up on HPC and AI Hardware

Two days before the most potent challenger of its data center market dominance AMD launches the highly anticipated second-generation Epyc server chips – a launch that’s sure to be accompanied by all the big hardware vendors’ announcements of new Epyc-powered systems – Intel announced an expansion of its partnership with one of those hardware vendors: Lenovo.

Together, the companies will build systems and software for high-performance computing and AI, according to Intel. Playing a big role in the announcement is the convergence between those two types of workloads that’s taken place in recent years.

Traditionally discrete, the architecture of supercomputers for standard HPC workloads like weather modeling, genomics, and seismic imaging is increasingly the same as the architecture of server clusters used to train and deploy machine learning workloads.

“We see a lot of convergence” between the two, Charlie Boyle, senior director of marketing for Nvidia’s DGX AI supercomputer product line, told DCK in a recent interview. “Traditional supercomputing centers are all doing AI now. They may have built a classic supercomputer, and they’re all running classic supercomputer tasks and AI on the same systems. Same architecture for both.”

This convergence presents a big growth opportunity for vendors in the supercomputer ecosystem. Instead of just the narrow HPC field, their addressable market now also includes traditional large enterprises, many of whom are in various early stages of leveraging AI technologies for business.

Related:Intel Offers DL Boost as Alternative to Accelerators in AI Servers

Intel and Lenovo will collaborate on products that use some recently launched technologies but also technologies that Intel has announced but hasn’t yet brought to market. For example, Intel’s Xe discrete GPU accelerators that are part of the Monday announcement are due sometime next year.

Already available however are DL Boost, the feature for accelerating deep-learning applications on the latest 2nd Gen Intel Xeon Scalable processors, and Intel’s new persistent memory technology Optane DC.

Part of the plan is to build systems that combine 2nd Gen Scalable chips with Lenovo’s Neptune liquid cooling technology portfolio.

Lenovo plans to make products of the collaboration available through Lenovo’s TrueScale Infrastructure services, through which the vendor “rents” its hardware out to customers, charging them on a pay-as-you-go basis. Rolled out recently by all major data center hardware vendors, these services try to bring the experience of using on-premises IT infrastructure as close as possible to the experience of consuming public cloud services.

Related:MLPerf Is Changing the AI Hardware Performance Conversation. Here’s how

Another major area of focus is aligning Lenovo’s LiCO cluster management software for AI and HPC with Intel’s upcoming Open API. Announced this June, One API is Intel’s effort to create a single programming model for all its chip architectures, including CPUs, FPGAs, GPUs, and its AI accelerators designed specifically for matrix operations in neural networks.

Finally, Intel and Lenovo said they will build joint “HPC and AI centers of excellence” around the world to help researchers and universities work on solving humanity’s big problems, such as cancer and climate change.