Covering Scientific & Technical AI | Friday, December 13, 2024

AMD’s Latest Third-Gen Epyc ‘Milan’ CPUs Include AI Implications as Well 

Image credit: AMD

A wide range of new third-generation AMD Epyc “Milan” CPUs were unveiled March 15 (Monday) by the chipmaker, from a flagship 64-core model to 8-core versions built for a myriad of server workloads.

For AI users, the big news is that the latest generation of AMD server chips show promise in improving performance for many AI processes, according to the company.

The latest Epyc chips were announced in a virtual launch event held by AMD CEO Lisa Su, who showed off 19 Epyc 7003-series SKUs, including the flagship 64-core, 280-watt model 7763, which is aimed at HPC, enterprise and cloud workloads.

The third-gen Epyc Milan chips achieve 19 percent more instructions per clock (IPC) than their predecessor, the second-gen Rome chips, according to the company. Like Rome (released in August 2019), the Milan series processors employ up to 64 7nm cores per processor and integrate PCIe Gen 4 connectivity, but the latest versions add per socket performance and per core performance with new “Zen 3” cores and enhanced security features.

Lisa Su, AMD CEO

Although Milan actually started shipping in the fourth-quarter of 2020, today marks the official product launch.

“High performance computing has never been more important than it is now,” amid the ongoing COVID-19 pandemic, said Su. “From how we work to how our kids learn to how we connect and do business, computing has enabled [it all]. And this trend has become more clear over the past year. We're actually now in a high-performance computing megacycle and it's driven by cloud computing, digital transformation, 5G and AI.”

Of course, existing power levels won’t continue to cut it, she said. “And the fact is everyone needs and wants more compute. At AMD, we focus on pushing the envelope in high performance computing, and that's across PCs, supercomputers, servers, consoles or edge devices,” said Su. “AMD is all about high performance.”

Epyc and AI

In a virtual pre-briefing for reporters on the new products last week, Ram Peddibhotla, AMD’s vice president of product management, said that the latest chips gain dramatically improved intake bandwidth in the cores. “And that provides significant [generation to generation] benefit as we go from Rome to Milan, in terms of what AI inference on CPUs can accomplish on the Epyc product line,” he said.

Ram Peddibhotla, AMD vice president product management

AMD is working closely with a number of customers who do inline AI inference on CPUs, Peddibhotla added. “And they're seeing tremendous benefit when you look at specific AI applications, like recommendation engines, natural language processing, or computer vision. So our overall approach is to look across that focus on the various elements of that solution stack, starting with the improvements that we've made in the core, and then collaborate with our partners and customers along with the improvements in the software stack.”

More details about these capabilities will be announced as the Epyc product line is dispersed through the rest of the year, he said.

Image credit: AMD

The new Epyc processors increase transactional database processing by up to 19%, improve Hadoop big data analytic sorts by up to 60%, and provide 61% better price-to-performance than the competition, AMD claims. They also offer superior performance for flexible hyperconverged infrastructure, according to the company.

Making the new chips even more valuable for customers, said Forrest Norrod, senior vice president and general manager of the data center and embedded solutions business group at AMD, is that they are socket-compatible with the second generation, enabling customers and partners to integrate them more easily.

“All of these processors, regardless of their core count, include the complete set of Epyc features, 8 channels of high performance memory supporting up to four terabytes of DRAM per CPU, 128 or more lanes of PCIe generation 4 I/O to connect to networking, storage and accelerators, without the cost or performance impacts of external bridge chips,” said Norrod. “Crucially, this Epyc family offers an unmatched set of security features to help keep your data, your users and your customers safe.”

An Analyst Weighs In

Dan Olds, analyst

Dan Olds, the chief research officer for Intersect360 Research, told HPCwire that he sees the Epyc launch as another positive step from AMD, continuing the momentum of the last two launches.

“With Zen3, AMD is once again pushing the bar up in terms of performance, which will be of particular interest to potential HPC customers,” said Olds. “According to our latest research, AMD processors are already deployed in more than 70% of HPC data centers. Interestingly, this growth was strongest in the commercial sector.”

And AMD isn’t just increasing core counts to bolster their results, he said. “They’ve also significantly increased the single thread performance over their previous generation. This will improve performance for customers who need high per thread throughput and who are not helped as much by higher core counts.”

Partners Announcing their Epyc CPU Implementations

AMD partners are already unveiling their own plans for using the latest AMD chips in their products and services, according to the chipmaker.

Among them are:

▪ AWS reports it will add the latest Epyc series processors to its core Amazon EC2 instance families later this year.

▪ Cisco has introduced new Cisco Unified Computing System (Cisco UCS) rack server models with AMD EPYC 7003 series processors designed to support modern hybrid cloud workloads.

▪ Dell Technologies is announcing its new PowerEdge XE8545 server with AMD Epyc 7003 series CPUs, and the company will support the new processors within its PowerEdge server portfolio.

▪ Google Cloud has announced that the Epyc third-gen processors will power a new compute optimized VM, C2D, and an expansion of the existing general purpose N2D VM later this year. Google Cloud Confidential Computing will be available on both C2D and N2D.

▪ HPE has announced that it will double its lineup of AMD Epyc processor-powered products, including HPE ProLiant servers, HPE Apollo systems and HPE Cray EX supercomputers.

▪ Lenovo is adding 10 Lenovo ThinkSystem server models and ThinkAgile HCI solutions built on the new Epyc CPUs.

▪ Microsoft Azure has announced multiple new virtual machine offerings powered by the new CPUs. New Azure HBv3 virtual machines for HPC applications are generally available today, and Confidential Computing virtual machines that utilize the full security features of the new AMD chips are in private preview.

▪ Oracle Cloud Infrastructure has announced that it is extending its flexible virtual machine and bare metal compute offerings with the new E4 platform based on the new processors.

▪ Supermicro has introduced the new CPUs in its Supermicro A+ single and dual socket family of Ultra, Twin, SuperBlade, storage and GPU optimized systems.

▪ Tencent Cloud has announced that the new Tencent Cloud SA3 server instance will be powered by the latest AMD CPUs.

▪ VMware has announced that its latest release of VMware vSphere 7 is optimized to take advantage of the virtualization performance of the new CPUs, while supporting the processors’ advanced security features, including SEV-ES for both virtual machine based and containerized applications.

Todd R. Weiss contributed to this report. A version of this article first appeared on sister website HPCwire.

About the author: Tiffany Trader

With over a decade’s experience covering the HPC space, Tiffany Trader is one of the preeminent voices reporting on advanced scale computing today.

AIwire