Intel plans to jump into the single socket market with a portion of its just announced Cascade Lake microprocessor line according to one media report. This isn’t a shocking surprise as many observers speculated Intel would counter AMD’s one-socket version of its Epyc processor line with an Intel part.
Intel made no mention of single socket offering plans at its Data-Centric Innovation Day in San Francisco on April 2 although it revealed its new line of microprocessors will be implemented in a multi-chip package something that AMD did with its Epyc line introduced in June 2017. (see HPCwire article, Intel Launches Cascade Lake Xeons with Up to 56 Cores)
News of the single socket offering broke last week from news outlet STH which reported:
“When Intel briefed the press and sent STH as well as other outlets its SKU list it left a series of processors out. Intel is feeling the pressure from AMD EPYC competition and is launching a series of “U” processors that do not have UPI links and are therefore single socket only parts.
“[We] we have learned that Intel is going to be selling U-series Intel Xeon Gold CPUs for the single socket only market to address competition from AMD EPYC. We have learned of two SKUs, potentially more SKUs coming. Thus far, we have confirmed with Intel and a major OEM that at least the three SKUs we have are valid.”
The Xeon Gold U-series chips are still on Intel’s Skylake processor microarchitecture and 14nm manufacturing process. “What sets the U-series processors apart from their other siblings is the lack of the UPI (Ultra Path Interconnect) link, and therefore they will not be able to mingle with other processors of the same kind. In other words, the Xeon Gold U-series lineups are tailored specifically towards single-socket servers,” reported tom’s HARDWARE late last week.
AMD’s successful return to the server market with its Epyc line surprised many observers given the abrupt way it abandoned the market earlier. Intel clearly still dominates, but AMD’s Epyc line now has five percent of the market, according to Scott Aylor, AMD corporate vice president and general manager of Enterprise Solutions. A single socket gambit was always part of AMD’s plan and it has met with considerable interest. (See HPCwirearticle, AMD Wins Another: Baidu to Deploy EPYC on Single Socket Servers)
At the time of Epyc’s launch Aylor told HPCwire, “We can build a no compromise one-socket offering that will allow us to cover up to 50 percent of the two-socket market that is today held by the [Intel Broadwell] E5-2650 and below. In our one socket offering we have come up with a clever way to maintain all of the I/O capabilities that you would get in a two socket as well as the full complement of eight memory channels. Today people buy two socket, sometimes because they need to, but more often than not because they have to. There are many examples in which I/O rich [workloads] like storage, like GPU compute, and some vertical workloads where people don’t necessarily need two sockets from a CPU performance perspective.”
Intel’s move to multi-chip modules and now reportedly offering a single socket solution suggests AMD must be moving in the right direction. AMD has even begun to show traction in high-end HPC – Lawrence Livermore National Lab (Penguin Computing), NERSC (Cray), and High-Performance Computing Center of the University of Stuttgart (HPE) have all contracted for Epyc-based systems. (see HPCwire article, AMD Sets Up for Epyc Epoch).
It will be interesting to watch how the market responds to Intel’s next-gen Cascade Lake processor line. Despite AMD’s recent success Intel remains the goliath in the datacenter and matching AMD’s single-socket offering means it will no longer have a hole in this nascent part of the market. Stay tuned.