Data center News
Intel Aims High With New Xeon Scalable ‘Sapphire Rapids’ Processors
Joseph F. Kovar
‘This launch is very real. OEMs, our cloud providers, our developers, are already embracing and delivering the fourth-gen Xeon and Max family of processors. And the fourth gen and the Max family delivers extraordinary performance gains, efficiency, security capabilities, breakthrough new capacities, and AI and cloud and networking, and delivering the world‘s most powerful supercomputers that have ever been built,’ says Intel CEO Pat Gelsinger.
Intel Tuesday officially launched its fourth-generation Xeon Scalable processors, code-named Sapphire Rapids, calling it the highest-performing data center processor family, and introduced a wide range of OEM partners already taking advantage of the new technology.
The company, in a live product launch, also unveiled the Intel Xeon CPU Max series GPUs, code-named Sapphire Rapids HBM, and the Intel Data Center GPU Max series high-density processors.
The introduction of the fourth-generation Intel Xeon Scalable processors comes just a week after arch-rival AMD unveiled wide range of new processors, GPUs, and AI accelerators.
The launch of Intel’s fourth generation Xeon processor and its Max family isn’t just about announcing a product, it’s about showing how customers are already taking advantage of this amazing technology, said Intel CEO Pat Gelsinger.
“This launch is very real,” Gelsinger said. “OEMs, our cloud providers, our developers, are already embracing and delivering the fourth-gen Xeon and Max family of processors. And the fourth gen and the Max family delivers extraordinary performance gains, efficiency, security capabilities, breakthrough new capacities, and AI and cloud and networking, and delivering the world‘s most powerful supercomputers that have ever been built.”
Sandra Rivera, Intel’s executive vice president and general manager of the Data Center and AI Group, said her company has over two decades of data center, networking, and cloud innovation, and that the new fourth-generation Xeon Scalable processors build on that foundation.
The new fourth-generation Xeon Scalable processors represent a paradigm shift in the way businesses run their workloads and solve computing challenges, Rivera said.
They feature the most accelerators built into a data center processor in the industry, giving an average of 2.9 times increase in performance and performance per watt efficiency across a group of the most common workloads, compared to the prior generation. The also provide customers with total cost of ownership improvements ranging from of up to 66 percent, depending on workload, she said.
The fourth-generation Xeon Scalable processors have features to improve CPU utilization and power efficiency, leading to such advances as a 10-times increase in PyTorch real-time inference and training performance using Intel Advanced Matrix extensions versus the prior generation, Rivera said. Customers can encrypt data using Intel Quick Assist technology that uses up to 47 percent fewer cores to get the same performance as before, she said.
“And we can only achieve these gains with close collaboration with the world‘s leading software providers, OEMs, and ODMs,” she said. “There are thousands of software engineers at Intel working with our ecosystem partners to optimize and validate applications to run best on Intel hardware. And Xeon-based infrastructure is available virtually everywhere in the world. Today, there are over 100 million Xeons installed in the market, from on prem-data centers, to 5G networking equipment, to edge infrastructure, and of course, the cloud service provider services.”
Xeon processors are the most ubiquitous processors in the cloud world, with Xeon-based services available from virtually every cloud service provider in every global region, Rivera said.
“With 400 Intel-based instances, Xeon is the most widely deployed platform in AWS’ infrastructure, allowing customers to access the unique features in our hardware from every region in the world,” she said. “And in addition to AWS, many of the world‘s leading cloud service providers already have fourth-gen Xeon services in public preview and general availability. These include Alibaba, Baidu, ByteDance, Google Cloud, IBM Cloud, Microsoft Azure, Oracle Cloud. and Tencent.”
Lisa Spelman, corporate vice president and general manager of Intel Xeon products, introduced Xeon Max, which she said called the first and only x86-based processor with high bandwidth memory. The Xeon CPU Max Series offers 64 GB of high bandwidth memory (HBM2e) on the package to target high data throughput for high performance computing and AI workloads, Spelman said.
“This innovative product addresses the number one constraint in high performance computing solutions,” she said. “And it dramatically increases the memory bandwidth that‘s available and accelerates key high performance computing workloads like life and material sciences, manufacturing, [and] high-energy physics.”
Also new is the Data Center GPU Max series of GPUs, which pack over 100 billion transistors into a 47-tile package to target challenging workloads like physics, financial services, and life sciences.
Intel also plans to launch a workstation version of its Xeon Scalable processors, the Intel Xeon W, in mid-February to power next-generation high performance workstations, Spelman said.
A Little Help From Intel’s Friends
To help illustrate the industry-wide acceptance of the new fourth-generation Xeon Scalable processors, Intel used its presentation to show short videos of some of its key OEM partners who discussed how they plan to use the new processors.
Dell Technologies Chairman and CEO Michael Dell said the Intel Xeon Scalable processors will power his company’s new 16th-generation PowerEdge servers targeted for multi-cloud, edge, and the incredible amount of data that needs to be stored and secure, said.
China-based Inspur Information CEO Peter Peng said the new Intel Xeon Scalable processors will be used in the company’s server like for a wide range of data center applications including green computing.
Hewlett Packard Enterprise CEO Antonio Neri said his company just launched new HPE ProLiant Gen 11 servers and HPE Cray supercomputers built with the fourth generation Intel Xeon scalable processors to push the boundaries of performance enabling, AI at scale and fueling innovation.
Lenovo Chairman and CEO Yuanqing Yang said the fourth generation Intel Xeon Scalable processors will power Lenovo‘s new infrastructure technologies including Lenovo ThinkSystem servers, ThinkAgile hyperconverged infrastructure systems, and ThinkEdge edge servers.
John Davidson, executive vice president and general manager of Cisco’s networking team, said his company will bring Intel‘s fourth generation Xeon Scalable processors to Cisco’s newest UCS M7 servers, including the UCS X modular servers managed by Cisco Intersight.
Supermicro CEO Charles Liang said his company is supporting the new Xeon Scalable and Max processors with over 20 new products to deliver servers with the best performance per Watt, “green computing,” and rack-scale solutions incorporating server, storage, networking, security, software, and services.
Nvidia President and CEO Jensen Huang said that data centers use 4 percent of the world‘s electricity, up from 1 percent five years ago, which isn’t sustainable, and so Nvidia is looking forward to pairing Intel’s fourth generation Xeon CPUs with Nvidia H100 GPUs and CX7 InfiniBand networking for a new generation of lower-power cloud super computing and AI systems.
Other data center technology providers also voiced support for the new processors.
Memory manufacturer Micron Technology Tuesday said its DDR5 server memory portfolio for the data center has already been fully validated on the fourth-generation Xeon Scalable processor line.
System builder Unicom Engineering Tuesday introduced plans to launch new server products based on fourth-generation Intel Xeon Scalable processors.
Lanner Electronics Tuesday unveiled its NCA-6530, a 2U, 19-inch rackmount network security appliance, and its HMB-6130, a compute blade for Lanner HTCA-series. Both support dual-socket fourth-generation Intel Xeon Scalable processors, and feature a built-in accelerator for higher throughputs, hardware-enabled security, and NIC module expansion.