Report identifies leading software-defined data center chip vendors serving cloud infrastructure and platform operators.
The world's chip supply may be in disarray, but the companies making them are raking in revenue as customers scramble to bring processing power to bear on giant data center workloads. The classic battle between Intel and AMD for market share in data center has been joined by Nvidia, Arm and Marvell as each manufacturer races — sometimes together, sometimes separately — towards a future where their product dominates the data center landscape.
AMD's innovation engine has been leading Fortune 500 companies and cutting-edge scientific research for more than 50 years. The company recently partnered with Microsoft to create Azure HBv3 virtual machines enhanced by 3rd Gen AMD EPYC processors. The units come with AMD 3D V-cache, codenamed "Milan-X," which provides 80 percent more performance for certain workloads. AMD's partnership with HPE and the Oak Ridge National Laboratory resulted in the Frontier supercomputer, which set a record for performing a quintillion calculations in a second. This partnership demonstrates AMD's favored position over Intel's Ice Lake chip in the competitive data center market.
In late May, Nvidia deployed its Arm-based Hopper GPUs and Grace CPU "superchips," a move that could prove to be a massive compute shift in the data center market. Among the first computer makers to release the new chips will be industry leaders such as Asus, Foxconn Industrial Internet, Gigabyte, QCT, Supermicro and Wiwynn. Those chips can be used for a wide range of workloads spanning digital twins, AI, high performance computing, cloud graphics and gaming, with systems expected to begin shipping from those companies in the first half of 2023. Nvidia, based in Santa Clara, Calif., has been at the forefront of modern computer graphics since 1993, delivering high performance computing power to demanding applications in artificial intelligence, as well as in industries such as manufacturing, health care, and transportation. Jensen Huang serves as President and CEO.
Intel launched a second-generation Habana AI deep learning processor to churn massive AI workloads at the data center — a result of Intel buying Habana Labs for $2 billion in 2019. The new chips include the Habana Gaudi2 and Habana Greco, built on a seven-nanometer architecture. Gaudi2 was designed to bring improved deep learning performance and efficiency to cloud and on-premises systems. The Xeon Scalable Processor serves as a foundation for data center scalability and agility, while its FPGA-acceleration chip Stratix 10 devices deliver up to 2X performance gains over the previous generation. Matthew Murphy serves as President and CEO.
Arm designs energy efficient processors and software platforms that enable advanced computing. In the data center, the company's Neoverse N1 CPU is optimized for cloud-native workloads and gives customers a 30 percent gain in compute efficiency over previous generations of the product. The N1 can handle sub-35W systems for networking and specific tasks, all the way to hyperscale environments. The company's partnership with Nvidia has been a boon for the technical reputation of both companies. Marvell, based in Santa Clara, Calif., has for the last 25 years stored, processed and secured the world's data with its semiconductor line up. The company's data center switches deliver bandwidth and scale, along with advanced packet processing and analytics. A surge in demand among data center customers spurred a 71-percent year over year growth last quarter, with the company predicting 41-percent year over year growth next quarter.