Core density gives AMD a boost in the supply-constrained server CPU market.
Demand among cloud service providers along with ongoing chip shortages helped boost shipments of Arm-based servers during the third quarter.
Five percent of servers shipped to cloud providers during the July-September period contained an Arm CPU, according to market tracker Omdia. Amazon Web Service’s Arm-based Graviton processor contributed to the surge. Ampere Computing, another vendor offering Arm processors, also saw strong demand from cloud customers such as Equinix and Oracle.
Meanwhile, China’s Huawei stepped up cloud server deployments of its Kunpeng CPU based on the Arm architecture, Omdia reported.
Overall, third quarter server shipments remained flat on a quarterly basis at 3.4 million units. Quarterly revenues rose 6 percent on an annual basis to $21.6 billion, largely driven by rising server prices that reflect ongoing chip shortages.
“The data center server market continues to be supply-constrained because of the shortage of key semiconductor components like power management ICs, microcontrollers and other ASICs,” said Manoj Sukumaran, Omdia’s data center, computing and networking analyst.“The demand for servers remains very strong across market segments and vendor order backlogs are at historically higher levels.”
Chip shortages prompted Omdia to lower its server CPU market forecast to 86 million. The market tracker also does not expect component shortfalls to improve until “at least” the second half of 2022.
Along with Arm-based processors, AMD continued to make inroads in the server CPU market. Omdia reported that 18 percent of servers shipped during the third quarter incorporated an AMD processor, up to 2 percentage points over the previous quarter.
The gains were attributed to strong demand among hyperscale cloud providers steadily deploying AMD Milan and Rome CPUs.
AMD’s server gains “are a reflection of their x86 market-leading core density and cache memory per socket,” Sukumaran noted. The new Bergamo CPU optimized for cloud deployments contains up to 128 cores, making it “very compelling” for cloud service providers,” he added.
According to Synergy Research Group, the number of hyperscale data centers increased to 700 by the end of the third quarter. “Both the number and average size of hyperscale data centers continue to grow steadily,” said John Dinsdale, chief analyst at Synergy Research Group. “We also see a very healthy pipeline of hyperscale data centers being planned, developed or fitted out, supporting our strong five-year growth forecasts.”
If those forecasts pan out, demand for server CPUs will only increase, perhaps prompting cloud vendors to diversify data center infrastructure by deploying more Arm-based designs.
This article was originally published on EE Times.
George Leopold has written about science and technology from Washington, D.C., since 1986. Besides EE Times, Leopold’s work has appeared in The New York Times, New Scientist, and other publications. He resides in Reston, Va.