Getty Images/iStockphoto
Intel unveils more Xeon 6 chips for AI servers
Intel launches Xeon 6 processors, the 6700P and 6500P, targeting AI systems and multi-socket servers, and enhancing performance for enterprise applications while reducing prices.
Intel has introduced its latest Xeon 6 processors, aiming the hardware at AI systems and the multi-socket data center servers that perform the computational grunt work.
On Monday, the company introduced the 6700P and 6500P series of Granite Rapids Xeon 6 chips, focusing on AI systems with CPUs working alongside GPUs and mainstay data center tasks such as virtualization, storage and computing in large multi-socket servers. The latest chips come five months after Intel released its most powerful Xeon 6, the 6900P.
The P in the name refers to P-core, which means the processors are designed for high-performance tasks through higher clock speeds, hyper-threading support and other features aimed at the largest enterprise applications.
Intel needed more top-of-the-line Xeon 6 processors for servers running AI applications, according to Jack Gold, an analyst at J. Gold Associates. Before the latest release, Intel had only the 6900P series and the Xeon 6E line. E-core chips are energy efficient and utilize less power than the P-core line.
"They needed something more in the high end to go after workloads increasingly driven by AI capabilities, AI types of workflows -- and the P cores do that," Gold said.
Intel designed the 6700P and 6500P for AI systems and servers of four, eight and more sockets running, for example, large installations of SAP HANA and an Oracle in-memory database, both relational databases providing real-time data processing capabilities.
The chips have built-in accelerator engines that Intel expects will differentiate the latest products from competitor hardware. Most discussions of accelerators in the chip industry refer to GPUs running AI workloads. Intel's accelerators in Granite Rapids Xeon 6 chips focus on AI, analytics, network security and storage.
Enterprises can use the CPUs to manage and coordinate workloads in an AI system powered by GPUs. The CPUs will handle data preprocessing, memory management and communication with GPUs using interconnects such as Compute Express Link (CXL).
Intel pricing falls
Intel pricing for its CPUs is coming down, Omdia analyst Alexander Harrowell said. Prices for Intel's two previous generations of Xeon processors cost as much as an AMD EPYC CPU and an Nvidia L4 GPU, an energy-efficient AI accelerator.
"AMD had a substantial price edge," Harrowell said.
However, AMD has raised its prices lately because of the higher cost of building advancements into the latest CPUs and, possibly, an attempt to increase its profit margin, Harrowell said.
"Intel has begun to set its prices more sharply in order to defend market share," he said.
Intel also launched the Xeon 6 system-on-chip with built-in accelerators for virtualized radio access networks (vRAN). The SoC also has Intel's Media Transcoder Accelerator, which is used in video streaming, content delivery and cloud-based media services.
VRANs are virtualized versions of traditional RANs that connect smartphones and other mobile devices to a telecommunications system's core network. VRANs are typically found in 5G networks to replace conventional RANs in proprietary hardware systems.
The Xeon 6 SoC was necessary to provide more advanced capabilities and to stay ahead of competitors Nvidia and AMD in the 5G market, Gold said.
"They needed to give the companies they work with a reason to upgrade," Gold said. "It's much easier to upgrade to a new chipset than it is to switch vendors."
Antone Gonsalves is an editor at large for Informa TechTarget, reporting on industry trends critical to enterprise tech buyers. He has worked in tech journalism for 25 years and is based in San Francisco. Have a news tip? Please drop him an email.
