Intel Teases an "Up to" Doubling of AI Performance for Its New Xeon 6 with Performance Core Chips
The Xeon 6700P and 6500P launch today, with up to 80 cores and a focus on accelerating machine learning and artificial intelligence.
Intel has announced the launch of its latest data center chips in the Xeon 6 family, now featuring performance cores (P-cores) rather than efficiency cores (E-cores) — claiming an "up to" doubling in performance for machine learning and artificial intelligence (ML and AI) workloads plus a special "edge" system-on-chip (SoC) variant.
"We are intensely focused on bringing cutting-edge leadership products to market that solve our customers' greatest challenges and help drive the growth of their business," claims Intel's Michelle Johnston Holthaus, chief executive officer of Intel Products and current co-CEO of Intel itself following the ousting of Pat Gelsinger. "The Xeon 6 family delivers the industry's best CPU for AI and groundbreaking features for networking, while simultaneously driving efficiency and bringing down the total cost of ownership."
According to Intel's internal testing, the Intel Xeon 6700P and Xeon 6500P processor families — not to be confused with the Xeon 6700E and 6500E, launched last year and using lower-performance "efficiency cores" in place of the new models' "performance cores" — offer "up to" twice the performance for artificial intelligence workloads compared to their previous-generation equivalents — alongside "an average" of 1.4× performance for "a wide range of enterprise workloads." The company is particularly interested in stemming a shift to AMD and Arm host processors for large-scale AI accelerator installations, claiming that the new Xeon cores deliver up to 1.5× more inference performance on two-thirds the cores of AMD's fifth-generation EPYC processors.
While traditional a CPU-only family targeting the data center, Intel's launch comes with the announcement of the Intel Xeon 6 "for network and edge." An all-in-one system-on-chip design, the new edge chip comes with integrated accelerators for AI workloads, network security, media, and virtualized radio access networks (vRANs) — suggesting Intel sees a future in which cellular base stations rely on machine learning models to process ever-increasing amounts of data in a finite radio spectrum.
The launch comes as Intel struggles to compete in a market that has moved compute from the CPU onto GPU and dedicated accelerators and in which it has run into a number of speed bumps in transitioning its processors to next-generation single-digit-nanometer process nodes. Earlier this month rumors spread that Broadcom and TSMC were having "preliminary and largely informal" investigations into potentially acquiring parts of Intel — a move that would see the company split in half.
More information on the new chips is available on Intel's website; pricing starts at $815 for the Xeon 6511P with 16 cores and rises to $8,960 for the Xeon 6781P with 80 cores, while Intel says that there are "more than 500 designs available now or in progress" from partners including Dell, Samsung, Hewlett Packard Enterprise, Lenovo, and Supermicro.