Advanced Micro Devices CEO Lisa Su announced that the code-named “Rome” 2nd Gen Epyc processors are the most powerful x86 processors in the world. The company said that its new Epyc 2 processors for the data center have as many as 64 cores and deliver twice the price performance of Intel’s fastest available chips.
AMD trotted out customer after customer at the event — Dell, Lenovo, Microsoft, HP Enterprise, Google, and Cray — to show that the company’s Zen 2 processor technology and its partners’ 7-nanometer manufacturing have resulted in stellar products for the data center, where Intel has been dominant for many years.
Bart Sano, vice president of engineering at Google, said onstage at the AMD event at the Palace of Fine Arts in San Francisco, “We have to optimize through the entire stack. That’s the reason that we chose Epyc.”
Su said that the customers have produced more than 80 world records for data center performance with the Epyc 2 chips, which are launching today. At the same time, she said they are lowering the total cost of ownership for customers by 25% to 50%.
“We’ve told you everything we have to tell you,” said Su. “I hope it is absolutely clear that 2nd Gen Epyc is the best in the industry. Google has already deployed within our data centers the 2nd Gen Epyc technology. We are already seeing great performance on a variety of workloads.”
Google will make Epyc 2 available to its customers on the Google Cloud.
Patrick Moorhead, analyst at Moor Insights & Strategy, said in an interview that he was surprised that AMD made so many architectural changes from one generation to the next.
“AMD took a big step forward today in the data center with its launch of the 2nd Gen EPYC processor and platform. It is a bigger leap forward than I had expected,” he said. “AMD improved most of its Gen 1 shortcomings like single-thread performance (+15%) and core scaling and added new RAS (uncorrectable DRAM error entry) and security (Secure Memory Encryption, Secure Encrypted Virtualization, 509 keys) capabilities in addition to substantial, multi-core performance gains.”
He noted AMD had the ecosystem in its support, from end customers like Google and Twitter to software providers like VMware, Canonical, Red Hat, and Suse, as well as manufacturers such as Gigabyte and QCT.
“Google is an interesting end customer who has exhibited it is willing to go big if it sees better performance and price,” Moorhead said. “Google was AMD’s largest Opteron customer back in the day. I will be keeping my eye on this one.”
He said that early indications show AMD is likely to do well in comparisons to Intel chips on a variety of workloads, though perhaps not all of them.
“AMD looks strong in Hadoop RT analytics (AMD says world record), Java throughput (AMD says 83% better), fluid dynamics (AMD says 2X better), and virtualization (AMD says up to 50% lower TCO),” he said. “Intel will likely have advantages on low latency ML inference workloads that take advantage of Intel’s DLBoost instructions. Intel will also look very good in in-memory database workloads utilizing Optane DC.”
But he said the industry will likely provide independent third-party benchmarks soon that will be more definitive.
Forrest Norrod, senior vice president of data center and embedded at AMD, said that AMD’s top 64-core Epyc 2 has twice the performance of Intel’s top chip at half the price.
“Rome kicks ass,” he said. “The new standard for the data center is Epyc.”
Peter Ungaro, CEO of Cray, said his company’s supercomputers will use the Epyc 2 chips in machines that will ship to the likes of the U.S. Air Force and Indiana University.
AMD gained an edge on Intel, which is still dominant in the PC processor market, a couple of years ago with its Zen design, which offered 52% better performance per clock cycle than the previous generation. Zen 2 is used in Epyc 2, while Zen 3 designs are complete and Zen 4 designs are underway.
Compared to a top Intel chip, the Zen 2-based 2nd Gen Epyc uses 61% less power than Intel’s top dual-socket Xeon product, with 75% lower software licensing costs, 50% fewer servers, and 54% lower cost of ownership, Norrod said.
“We could not be more committed to this space,” said Su.
The 2nd Gen Epyc chips have 32 billion transistors and as many as 64 cores. A spokesman for Intel said that the company has a long history of leadership in the server space.
In a statement, Intel said:
Intel has over 20 years of delivering uninterrupted data center leadership. In that time, we have built a broad ecosystem of partners who optimize their business applications around Intel platforms.
Intel’s focus is on delivering platform innovations that offer customers real-world application performance that help them solve their most critical business challenges.
Intel is taking an outside-in view on hearing what our customer’s need and delivering the silicon platforms they require – which include CPUs, accelerators, FPGAs, NNPs, memory and storage technologies, etc. Our ambitions have never been greater as a company, allowing us to target a >$200B total addressable market in the data center.
Some recent examples of Intel’s work with data center customers and partners who are leveraging Intel’s portfolio of processors, memory and AI acceleration technologies include SAP, Baidu and Lenovo.