Unlocking High Performance Computing with Intel's Omni Path Architecture

Unlocking High Performance Computing with Intel's Omni Path Architecture

Table of Contents

1. Introduction
2. What is Omni Path Architecture?
3. Advantages of Integration with Knights Landing
4. Details of the Discrete Solution
5. Live Hardware in the Booth
6. Building the Intel Fabric Builders Community
7. Shipping and Future Plans

Introduction

In this article, we will explore Intel's Omni Path Architecture (OPA) and its significance in high-performance computing (HPC) and data center workloads. We will discuss the advantages of integrating OPA with Knights Landing CPUs and delve into the details of the discrete solutions offered by Intel. Additionally, we will highlight the live hardware showcased by Intel and the growth of their Intel Fabric Builders community. Finally, we will touch upon the shipping plans and future prospects of Intel's OPA.

What is Omni Path Architecture?

Omni Path Architecture (OPA) is Intel's next-generation high-performance fabric designed specifically for high-performance computing and data center workloads. This architecture was first introduced at ISC 2014 in Leipzig and has since evolved to provide enhanced performance and scalability. OPA offers a high-speed interconnection between nodes, allowing for efficient and reliable data transfer. With a focus on density and power efficiency, OPA aims to meet the growing demands of HPC and data center environments.

Advantages of Integration with Knights Landing

Integration of OPA with Knights Landing CPUs comes with several advantages. Firstly, it addresses the long-standing demand from the community for Intel to incorporate a high-performance fabric within their CPUs. By integrating the fabric, Intel enables higher density and power efficiency, which are crucial considerations for HPC environments. Deeper levels of integration across Intel's roadmap further improve these benefits. Integration with Knights Landing CPUs also reduces latency, providing a 23% improvement over competing solutions. The advantages will only grow as Intel continues to innovate in this field.

Details of the Discrete Solution

While integration with Knights Landing CPUs is a prominent aspect of OPA, Intel also offers a discrete solution. This solution includes a 16-lane PCI Express card with a 100-gigabit per Second link. Additionally, Intel provides a 48-port radix switch, a 24-port switch, and director switches. These hardware components offer high message rates, with the 48 radix switch achieving an impressive 195 million messages per second per port. Furthermore, Intel's OPA exhibits lower latency compared to its closest competition, even at an early stage of development.

Live Hardware in the Booth

At events like ISC 2015, Intel showcases live hardware to demonstrate the capabilities of OPA. This live hardware enables visitors to witness the performance and functionality of Intel's OPA solutions firsthand. Seeing the hardware in action provides a tangible and immersive experience that reflects the real-world usability and potential of OPA.

Building the Intel Fabric Builders Community

Recognizing the importance of community engagement, Intel has initiated the Intel Fabric Builders program. Through this program, Intel aims to build a strong community of experts, professionals, and end-users working collaboratively in the field of fabric technologies. Since its launch, the Intel Fabric Builders community has more than tripled in size, with participation from 31 individuals. Intel's commitment to fostering this community ensures the availability of comprehensive solutions and promotes knowledge sharing among industry experts.

Shipping and Future Plans

Intel is committed to delivering on their promises and has announced plans to ship OPA products by the end of the year. They have already accumulated over a hundred thousand nodes in their pipeline, further demonstrating their dedication to the success of OPA. With a firm foothold in the market, Intel's OPA is set to pave the way for higher performance and efficient fabric solutions in high-performance computing and data center environments.

Highlights

• Omni Path Architecture (OPA) is Intel's next-generation high-performance fabric for HPC and data center workloads.
• Integration of OPA with Knights Landing CPUs offers advantages such as higher power efficiency and reduced latency.
• Intel's discrete solution includes a 16-lane PCI Express card, radix switches, and director switches.
• Live hardware demonstrations provide a firsthand experience of OPA's capabilities.
• The Intel Fabric Builders program fosters a community of experts and professionals in the fabric technologies field.
• Shipping of OPA products is scheduled for the end of the year, with over a hundred thousand nodes in the pipeline.

FAQ

Q: What is the purpose of Omni Path Architecture (OPA)? A: OPA is designed to provide a high-performance fabric solution for HPC and data center workloads, enabling efficient data transfer and scalability.

Q: What are the advantages of integrating OPA with Knights Landing CPUs? A: Integration with Knights Landing CPUs allows for higher density, power efficiency, and reduced latency, meeting the demands of HPC environments.

Q: What hardware components are offered in Intel's discrete solution? A: Intel's discrete solution includes a 16-lane PCI Express card, radix switches, and director switches to support OPA implementation.

Q: How does Intel engage with the fabric technologies community? A: Intel has initiated the Intel Fabric Builders program, which aims to build a collaborative community of fabric technology experts and end-users.

Q: When can we expect Intel's OPA products to be available? A: Intel plans to ship OPA products by the end of the year, with a significant number of nodes already in the pipeline.