The ARM Cortex-X1C is intended to run alongside the recently launched A78C big.LITTLE core that is used in laptops and adds key security technology.
The first Cortex-X CPU, the Cortex-X1 was launched in 2020 and has been used in Samsung’s Exynos 2100 SoC and Qualcomm's Snapdragon 888 5G mobile platform that both power the Samsung Galaxy S21 range of flagship smartphones. More recently it has been used in the Google Pixel 6 smartphone in the Google Tensor chip.
“Across aspects of performance, power and area (PPA), the Cortex-X1C is very similar to Cortex-X1, but also adopts features from Cortex-A78C to address the needs of specific applications and devices,” said Saurabh Pradhan, senior product marketing manager at ARM. “However, compared to Cortex-A78C, it delivers 22 percent better performance. Cortex-X1C also adopts features to enable ISA-compatible CPU cluster configurations of up to 8 big cores using an updated version of the DynamIQ Shared Unit (DSU).
This combination of the X1C and A78C allows chip designers to build CPU clusters that scale from high performance desktop to those that balance maximum performance and efficiency. This would see a cluster with eight large X1C cores run alongside an A78C cluster with a mix of large cores and other optimised for energy efficiency and battery life.
The Cortex-X1C adopts the advanced Pointer Authentication (PAC) in ARMv8.6, which improves data and device security through minimizing the attack surface. This ensures that any data on the device is secure. PAC also significantly reduces exploits and prevents attackers from taking control of the software control flow in Return-Orientated-Programming (ROP), which reduces exploits by over 60 percent, and Jump-Orientated-Programming (JOP), which reduces exploits by over 50 percent. Reducing potential exploits is essential to ensure that the data of users are kept safe when installing third-party applications.
Cortex-X1C is designed for more demanding workloads, providing an increased L3 cache memory of 8MB and eight big core support so it can run eight high-performance threads in parallel. Multi-OS support