Lenovo sees strong channel opportunity in new single-socket AMD EPYC servers

The first generation of AMD’s EPYC server processors re-established the company in that market, but was mainly a hyperscaler play. The market for the next-generation will be much broader, and a significant opportunity for Lenovo’s channel.

Lenovo ThinkSystem SR655 Rack Server

Last week, AMD excited the industry with their launch in San Francisco of their second-generation AMD EPYC 7002 Series server processors, which bring enormous enhancements over the well-received first-generation, and strongly position AMD in the data centre market going forward. While AMD’s achievements will benefit all their hardware OEM partners. Lenovo has a specific strategy for taking the new processors to market that they think will position them well against their competitors. With the ThinkSystem SR635 and SR655, Lenovo is focusing solely on single-socket servers, which they think are better suited for edge use cases, and which provide the performance of a dual-socket server. In addition, while Lenovo sold AMD’s original Naples EPYC processors to hyperscalers, the second-generation is a full-blown channel product.

“This second generation of AMD EPYC processors, codenamed Rome, is designed as either a single-socket or a dual-socket system,” said Kamran Amini, Vice President and General Manager, Data Center Infrastructure and Software-Defined Solutions at Lenovo Data Center Group. “The stack is identical. You go from an 8-core CPU all the way up to a 64-core CPU, with the same functional capability across the whole stack.”

AMD traditionally competed against Intel on price, with chips that they said were as good as, or almost as good as Intel, but which offered more bang for the buck. Analysts were pretty consistent last week in believing that Rome had shattered that paradigm, and that for the moment, the new release puts EPYC significantly ahead of existing Intel Xeons.

“There are four main things that AMD is delivering here,” Amini said. “The first is that its support for PCIe Generation 4 brings much more bandwidth and IO capability.” With each processor having 128 PCIe lanes on top of AMD’s 7nm technology, it effectively doubles speed regardless of the number of cores.

“There are also tons of cores per socket, faster memory and bigger bandwidth and embedded hardware security for virtual machines,” Amini added. The denser 7 nm environment allows for the same energy to be used as was used in the first generation, despite all the additional cores.

These benefits from Rome will be shared by all Lenovo’s OEM competitors as well. Lenovo, however, has concentrated on a single CPU system, which Amini said brings them some distinct advantages.

“The unique thing we did was completely optimize our system for Rome to fully take advantage of it,” he stressed. “We are optimized for Generation 4 speeds and the wattage of this CPU. We think we are truly differentiated in the market with this. We also think that the focus on single CPU will help us increase our total market, with a channel first strategy that responds to customer demands. With Rome, we focused on the single socket because we can get 64 cores with it. It’s a very attractive high-performance technology.”

Lenovo’s Rome-based servers will really be the first introduction to EPYC for many of their enterprise customers and partners, because while the original Naples EPYC processors put AMD back in the data centre picture, Lenovo sold it entirely to hyperscale customers.

“EPYC Naples re-introduced AMD to the data centre, and started the differentiation of what they could deliver, but the early adopters were hyperscalers,” Amini said. “That’s why we didn’t have a general-purpose product with Naples at all, just a hyperscale one. Rome is now truly trying to tackle what an enterprise client would look for. We are very channel-centric, and partners will find that these platforms were truly designed to optimize everything that Rome can deliver.”

Lenovo has two new single-socket servers, the 1U rack ThinkSystem SR635 and 2U rack SR655, both of which will ship on August 30. The company is touting 19 new single socket world benchmarks on these servers, and their higher throughput, lower latency, higher core density, and ultra-dense NVMe drive capacity that is the largest of any single-socket server on the market.

Amini said that they expect they expect these products to be especially significant in Smart Cities projects, to provide more computing power in confined spaces with less energy. For instance, they have a partnership with HCI vendor Pivot3, which has a strong position in the video security space, and Lenovo’s testing states that the Lenovo ThinkSystem SR655 combined with Pivot3’s software provides an edge computing video security solution that can support up to 33 per cent more video cameras per node than similar solutions.

“Our solution with Pivot3 requires a platform that supports massive amounts of data, so this is ideal,” Amini said.

So does this mean the two socket server is dead? While Lenovo will focus on the single socket for AMD, they still sell Intel’s two-socket offerings, and expect to see continued demand for those.

“We have a broad portfolio around the Intel E3 class today, and we expect these will continue to sell, for a variety of reasons,” Amini said. “Could these new servers take over a lot of the two-socket market? Absolutely. Do I see the two-socket market crumpling? No, for the same reason we still have four and eight socket markets, because some customers want them.

“Some of it is a traditional legacy way of doing  business,” Amini continued. “There are people who are conservative in how they deploy, just as there is a segment that doesn’t care what’s inside, only the SLA. Some people also said that the large mainframes and UNIX were indispensable, but when they saw how four-socket servers could do massive mainframe jobs, many people moved. But there are still mainframes, because there are migration issues. There are customers who care very much about performance because they are in a leading-edge environment where they hit bottlenecks, but not everyone wants or needs top CPU SKUs. That will be a journey where customers will evaluate based on the right price performance for them. So we will have a range of offerings for the right price performance environment.”

“We see these new servers as augmenting the channel offerings that we have had on the market,” said Nicole Roskill, Executive Director, Global Channels, Data Center Group at Lenovo. “We are very much thinking about how customers solve business problems, and our messaging to partners is that this gives them a broader portfolio to solve customer problems, and address different types of workloads that requires these technologies.”

Roskill said that while she doesn’t have a crystal ball to forecast demand for these new servers, they clearly have a strong value proposition for partners, and Lenovo is rolling out strong enablement programs around them.

“With shipments starting on August 30, we have been planning about how we can best enable the channel around it, and we will have channel programs and incentives in market to help promote these new systems,” she stated. This will be in our second half to coordinate with shipment support. In North America, it will be part of our Lenovo eXpert Achievers Program [LEAP]. All the SKUs will be available to partners, and we are building some TopSeller models that reflect the most popular configurations.”