Intel’s 72-Core Xeon Phi Processor Boasts AI And Machine Learning Mastery, Now Commercially Available

Xeon Phi Die
Intel Xeon Phi Processor Die
It’s been nearly two years since we first heard about Intel’s next generation Xeon Phi “Knights Landing” processors, which are geared towards the High Performance Computing (HPC) segment. The processors are a big part of Intel’s Scalable System Framework (SSF) and are built on using general-purpose x86 architecture using open standards. Intel says that this offers customers greater flexibility with respect to programming languages and tools for software development — basically anything that can run on traditional Xeon processors is available to the Xeon Phi family.

Today, Intel announced that its Xeon Phi processors are finally available to customers. This comes nearly a year after the company’s originally-quoted launch date, and seven months after Intel announced that pre-production chips were already in use by select partners.

xeon phi partners

The Xeon Phi processors feature double-precision performance in excess of 3 teraflops along with 8+ teraflops of single-precision performance. All Xeon Phi processors incorporate 16GB of on-package MCDRAM memory, which Intel says is five times more power efficient as GDDR5 and offers 500GB/s of sustained memory bandwidth. The MCDRAM can effectively be used as a high-speed cache or as a complimentary addition to the system DDR4 memory.

Intel is targeting its Xeon Phi as a more suitable solution than NVIDIA’s dedicated Tesla GPU accelerators, citing up to a 5.2x performance advantage in visualization (Embree), up to a 2.7x in mathematical modeling (Monte Carlo DP) and up to a 5x increase in life sciences (LAMMPS). The company is also touting an 8x advantage in performance-per-watt and a 9x advantage in performance-per-dollars along with its suitability in engineering, oil and gas, defense, and weather forecasting.

xeon phi skus

The Xeon Phi is available in four basic configurations with 64 to 72 cores, and with processor frequencies ranging from 1.3GHz to 1.5GHz. All four support up to 384GB of DDR4 memory, but the base Xeon Phi 7210 is limited to the 2133MHz variety.

Limited availability of the new family of Xeon Phi processors starts today, while widespread availability shall commence in September. According to Intel, it has shipped “tens of thousands” of its previous generation “Knight’s Corner” Xeon Phi processors to date, and expects the number of Xeon Phi processors sold to increase dramatically this year to over 100,000.

Intel, however, had better watch its back. NVIDIA is quickly ramping up its efforts in the HPC arena. The company witnessed a 63 percent increase in its datacenter business alone during the last quarter and is sure to give competitors like Intel a few heart palpitations. The company just announced details on its Tesla P100 PCIe accelerators, which include High Bandwidth Memory (HBM2).