Microsoft is turning up servers for Microsoft Azure based on Marvell’s ThunderX2 Arm server processor portfolio. The servers are the result of a collaboration that began in March 2017 between Marvell, Microsoft, and Foxconn subsidiary Ingrasys.
The chipmaker and Microsoft initially demonstrated cloud service workloads developed for Microsoft’s internal use running on ThunderX2 in 2017 and later that year released detailed specifications of the ThunderX2 server motherboard, including the block diagram, management subsystem, power management, and field-programmable gate array (FPGA) card support to the Open Compute Project (OCP).
The ThunderX2 platform is based on the second generation of the company’s Armv8-based server processors and designed to “drive high computational performance by delivering outstanding bandwidth and memory capacity,” according to Marvell. The company says the framework is tailored to run multiple workloads in scalable cloud configurations and computationally intensive applications.
The platform is also compliant with Microsoft’s Project Olympus, which is open sourced through the OCP. The project, which got underway three years ago, aims to enable a common hardware cloud design to ease deployments and accelerate the delivery of products.
“Microsoft’s deployment of Marvell’s ThunderX2 solutions in Azure further solidifies our position as one of the industry’s most deployed Arm-based server solutions in the world,” said Gopal Hedge, VP and GM of the server processor business at Marvell, in a prepared statement. “Integrating ThunderX2 into Azure infrastructure and application software environments is another major milestone that demonstrates the maturity of the ThunderX2 platform.”
‘Milestone’ for Microsoft Azure
Leendert van Doorn, a distinguished engineer at Microsoft Azure, described the deployment of Marvell’s ThunderX2 server processor as a milestone for product development on Microsoft’s public cloud. “Microsoft is committed to driving platform innovation, and our expanded support for silicon architectures is part of our commitment to industry leadership and best practices,” van Doorn said in a prepared statement.
“Collaborating with Marvell and Ingrasys has resulted in the most complete and highest performing Arm server solution possible for our internal use,” van Doorn explained.
The large-scale deployment of the ThunderX2 platform follows a busy spring for Marvell. The chipmaker doubled down on its 5G and data center chip business when it acquired Avera Semiconductor, the ASIC business of GlobalFoundries, for $650 million in cash.
The following week, Marvell announced plans to sell its WiFi and Bluetooth connectivity business to NXP for $1.76 billion. That deal is expected to close during the first quarter of 2020. Marvell also acquired Cavium for $6 billion in July 2018 to combine the companies respective strengths in infrastructure silicon. The ThunderX2 platform was designed by Cavium and reached general availability in 2018.
By moving into the ASIC market, Marvell is targeting more specialized workloads like machine learning and IoT with customizable microchips. That positions the chipmaker for opportunities in 5G networks and edge computing as service providers move data processing to more places.
Marvell’s strategy is winning over Wall Street, which appears to be bullish on the opportunity. Company stock is up 70% since the beginning of the year.