Arm continued to flex its chip design performance muscles by introducing on Tuesday the Neoverse V1 and N2 platforms (also known as Zeus and Perseus) with up to 50% more performance over the Neoverse N1 designs introduced in early 2019.
Arm said its existing N1 chip designs have been used all the way from high performance computing down to edge devices, with applications in between for hyperscale/cloud and 5G.
Back when Neoverse N1 and E1 were delivered, they offered 60% the performance of Arm’s Cortex-A72 CPU. Arm has a global base of customers for its CPU designs that run in data centers as well as nearly every smartphone. Nvidia recently said it will pay $40 billion to buy Arm, a process that could last 18 months.
RELATED: Nvidia faces low odds of winning regulatory love for Arm deal
Chris Bergey, general manager of Arm’s infrastructure line, described V1 (code-named Zeus) as appearing in 2020 with a single-threaded performance improvement of 50% over N1, important for apps reliant on CPU performance and bandwidth. It supports Scalable Vector Extensions deemed important in machine learning and cloud and HPC. “With SVE, we are ensuring portability and longevity of software code, along with efficient execution,” he said.
Neoverse N2 (Perseus) is planned for 2021 on the Arm roadmap and will be important for power-constrained edge devices, but can also be used in cloud, SmartNICs and enterprise networking. It has a 40% improvement in single-threaded performance over Neoverse N1 with the same level of power and area as Neoverse N1.
Also on the Arm roadmap is a set of Poseidon Generation Platforms for 2022 and beyond that will be built on a 5 nm or 3 nm process with a 30% workload performance improvement. N2 is designated for 5 nm and V1 for a 7 nm or 5 nm process.