Arm Ltd. today pulled back the curtains on two new seven-nanometer chip designs, including what the company describes as its first server-grade product capable of competing with Intel Corp.’s market-leading data center processors.

Arm already has a significant presence in the data center. The company’s chip designs, which it licenses to other semiconductor makers, are mostly found in specialized systems optimized for tasks such as cybersecurity. Arm has made notable inroads into general-purpose data center computing recently, but Intel continues to dominate the category.

The newly unveiled Neoverse N1 chip is aimed at loosening Intel’s grip on the market. Arm claims that the design provides as much as 2.5 times more processing power for certain server workloads than the previous Cortex-A72 architecture. Moreover, the N1 is at least 60 percent speedier when assessed by how fast it processes integers, a basic unit of data.

“The Neoverse N1 platform is the first compute platform from Arm capable of servicing the wider range of data center workloads with performance levels competitive with the legacy architectures used in the public cloud,” Kevin Ryan, Arm’s senior director of software ecosystem development, wrote in a blog post. Although primarily optimized for high performance, he said, it’s also designed for efficiency, achieving 30 percent more power efficiency than the Cortex-A72 in the same chipmaking process.

Another area where the chip improves upon its predecessor is artificial intelligence support. Companies almost always run AI workloads on graphics processing units, which are much better suited for the task than general-purpose chips, but there are certain scenarios where Arm does see the N1 being used to this end.

Large cloud providers such as Amazon Web Services Inc. sometimes offload AI workloads from GPUs to other processors as a way of improving operational efficiency. That’s a highly specialized use case, but it’s worth addressing given how cloud providers are some of the world’s biggest chip buyers. Arm claims that the N1 can run machine learning algorithms up to six times faster than the Cortex-A7.

In practice, speeds will likely vary significantly. Arm designed the N1 so semiconductor makers can use the architecture to build products with anywhere from four to 128 processing cores.

The Neoverse E1, the second new product the company unveiled today, has a more narrow focus. It’s designed to power systems such as cellular base stations that are specifically built to process network traffic.

Arm is looking to capitalize on the massive hardware investments carriers are making as part of the global shift to 5G, the fifth-generation wireless standard that promises to speed up mobile connections dramatically. The firm claims that the E1 transmits data up to 2.7 times faster than earlier chips with up to 2.4 times higher energy efficiency. It also provides up to 2.1 more processing power for carrying out actions on that data.

“Neoverse E1 and Neoverse N1 processors can be combined in a heterogenous design for high-performance systems,” Brian Jeff, Arm’s senior director of product marketing, detailed in a post. “Some of the example devices utilizing this kind of system are firewall appliances with deep packet inspection and intrusion detection capabilities.”

Photo: Arm