Parallel Performance – Intel Xeon Phi Systems

Posted on February 2, 2017 by Rachel B

Today’s high performance computing applications are becoming as resource-intensive and complex as they are diverse. Across disciplines such as Earth Sciences, Financial Services, Physics, and Machine Learning, demands for CPU power and memory bandwidth have increased exponentially. Within this set of workloads a common component is the need for highly parallel processing to maximize performance. However, up until now there hasn’t been a good solution to run these applications.

One solution was to to use regular Intel® Xeon™ processors. However, although extremely robust, the processor architecture is designed for general IT workloads. These processors support advanced features for virtualization and security that work great for central IT and storage activities. However, these special features are not required for highly parallel workloads. This means that a lot of the core processor functionality is wasted.

On the other end of the spectrum, there are highly specialized GPU processors specifically designed for parallel processing (CUDA). Unfortunately, these processors are not bootable and require legacy code to be rewritten for an entirely new architecture. This can dramatically increase development costs and the length of time it takes to get a model up and running.

With this in mind, Intel has created a new HPC processor that lowers the barrier to entry, without sacrificing performance. The Intel Xeon Phi™ processor is a bootable host processor delivering massive parallelism and vectorization in support of the most demanding high-performance computing applications. By using an x86 processor architecture, the Intel Xeon Phi let’s a developer run legacy applications originally written for the Intel Xeon platform with minimal changes. This dramatically cuts down on development costs and time to market — or time to insight.


Optimized for Parallel Performance

The new Intel Xeon Phi bootable host processor offers an integrated architecture for powerful, highly parallel performance that will blaze the trail to greater innovation and impact for today’s most-demanding High Performance Computing applications. With up to 72 out-of-order cores, this processor can deliver over 3 teraFLOPS (floating-point operations per second) of double-precision calculations.

Reduced Power Consumption

The many core architecture of the new Xeon Phi processors enables significant improvements in power consumption. The higher degree of parallelism in these processors results in greater compute per unit of energy consumed for highly-parallel applications.

Simplified Code Optimization

In the past, parallel CPU architectures required significant code re-writes in order to reap maximum performance benefits. This processor is binary-compatible with Intel Xeon processors, which allows it to run any x86 workload, optimizing asset utilization across the data center.

Deploy Your Own Custom Intel Xeon Phi System

Pogo Linux is introducing a new line of systems featuring the Intel Xeon Phi processors. Whatever your workload, our engineers can put together the ideal system for your needs. Now you’ve got a way to make it all happen.


To learn about how Pogo can help you tackle the technological challenges your organization is facing, please call us at 888-828-POGO, or email .