New Products and Collaborations Drive oneAPI Momentum

Sapphire Rapids will change the way developers work by providing in-package high bandwidth memory with HBM2e that leverages a multi-tile architecture.  The processor also provides enhanced performance, new accelerators, PCIe Gen 5 and other capabilities optimized for artificial intelligence (AI), data analytics and HPC workloads as well as a range of libraries and languages.  Taken together, these advancements improve the work carried out by developers and researchers working in heterogeneous compute environments.

This evolution is essential, as the oneAPI spec’s open, unified, cross-architecture programming model enables CPUs, GPUs and accelerator architectures (including FPGAs and others) to run on a single code base and software stack, and improve performance.

More Opportunities for HPC Workloads Means More Specialization

This evolution in HPC demands a mix of heterogeneous architectures.  “The x86 architecture can rapidly process large scalar workloads, but to produce orders-of magnitude performance gains, we must critically look at how HPC workloads are run within vector, matrix and spatial architectures, and we must ensure these architectures seamlessly work together,” said Jeff McVeigh, vice president and general manager of the Super Compute Group at Intel Corporation.

To achieve this goal, Intel adopted an “entire workload” strategy.  This enables workload-specific accelerators and GPUs to seamlessly work with CPUs.

Collaboration – it’s the Name of the Game

Executing Exascale in Europe.

“Ponte Vecchio” is the code name of Intel’s next-generation Xe HPC based GPU.  Behind it are industry leaders partnering to drive heterogeneous computing and to simplify software development.

A statement issued by Quanta/QCT, a global datacenter solution provider, says the company looks forward to building the next-generation HPC and AI solutions with the new Ponte Vecchio GPU and the new Sapphire Rapids CPU to empower a new era of exascale computing.

Another organization advancing the state of the industry is SiPearl (The Silicon Pearl).  SiPearl, along with 27 partners within the European Processor Initiative (EPI) project, in collaboration with Intel, is designing a high-performance, low-power microprocessor for the European exascale supercomputer that can perform 1 billion billion calculations per second.

With this new generation of microprocessors, European businesses and organizations can take advantage of HPC with a powerful solution which is close to home, energy efficient, and secure.

Among the nearly 30 partners of the EPI project are members of the scientific community and supercomputing centers, as well as corporations within the electronics and automotive sectors.  Each is not only a stakeholder, but a future client, which helps ensure all voices are heard at each phase of the project.

On the selection of Intel Xe HPC GPUs as SiPearl’s HPC accelerator to deliver Europe’s first exascale supercomputers, McVeigh had this to say.

“Intel is proud to support SiPearl in adopting oneAPI’s unified programming model to increase developer productivity and workload performance.  We see great potential in our work with SiPearl as we execute on our IDM 2.0 strategy to bring leading-edge technologies to our European partners,” said McVeigh.

The rate of compute demand demonstrates the need for higher performance, openness and scale in HPC.

HPC in Action

While much of the work performed using HPC in most industries takes place outside the view of the general public, research on COVID-19 conducted with HPC is reported on a daily basis in the media around the world. Since the pandemic began, research labs and in-cloud HPC-as-a-service providers have monitored, studied, and developed drugs used to prevent and treat the SARS-CoV-2 novel coronavirus.

Such rapid developments would be impossible without the extensive availability of HPC technologies.  And going forward into the exascale era of computing, more supercomputing will be indispensable to accelerate scientific discoveries and address critical issues.  This will require higher performance, openness and scale.  And to meet these criteria, the industry must invest in diverse technologies to provide orders-of-magnitude performance improvements and deliver greater accessibility to HPC solutions.  Therefore, we must reconsider how the components that power these systems are designed and built.

In the future, the ability to democratize HPC and deliver “supercomputing for all” will help solve problems and crises we’ve yet to confront, or those whose existence we are unaware of.