Arm Launches Next-Gen Big-Core: Cortex-A725


As part of today’s new 2024 Compute Subsystem (CSS) launch, Arm is unveiling its newest high-performance big core – the Cotex-A725, formerly codename Chaberton.


This article is part of a series of articles from Arm’s Client Tech Day:

 


The new Cortex-A725 is the successor to the Cortex-A720 which was launched last year. It is part of the 2024 Client Compute Subsystem (2024 CSS) which comprises the DSU-120, the interconnect subsystem, and the Immortalis GPU. The Arm compute power is housed in the DSU-120. This includes the Cortex-A520, and the new Cortex-A725, and the Cortex-X925 itself. All three are second-generation Armv9.2 CPUs. Armv9.1 brought GEMM and BFloat16 support while Armv9.2 brought SME (SVE2) support.

Details of the new Cortex-A725 were rather lackluster this year. Overall, the company says it has widened many of the back-end buffers. The new A725 offers a new 1 MiB L2 configuration, effectively doubling the capacity over its predecessor. All-in-all, Arm says the new A725 is the widest and deepest big core designed to date.

A large part of the effort around the new A725 revolved around the 3-nanometers process node. Beyond special optimizations specifically for the process, Arm says that various architectural changes took place to offer better PPA. One of those changes were noted on the first bullet point above which dealt with register file structure enhancement. Here Arm says that various lower-level optimizations were done to the memory buffers specifically for the 3-nm process. For example, the new 1 MiB L2 capacity option is delivered at ISO-area compared to the A720, which is beyond the area improvement offered by the process technology alone.

Overall, the Cortex-A725 offers a sizable performance-efficiency improvement. When compared to the Cortex-A720, the new A725 is said to deliver up to a 35% improvement in performance-efficiency. In other words, at ISO-performance, taking into account the new process improvements and architectural changes (e.g., 4nm vs 3nm), the new A725 is said to deliver up to a 35% reduction in power. Likewise, the A725 is said to deliver up to 25% higher power improvement at the Performance level of the Cortex-A720 peak performance for SPECint_base2017.

 
 

With the new 2024 CSS, Arm doubled the caches in its reference designs. From 32K/512K/8M L1/2/3 on the Cortex-A720 to 64K/1M/16M L1/2/3 on the new Cortex-A725. With those larger cache configurations and overall DSU cache improvements, Arm says that the A725 can see up to 20% traffic improvement compared to the A720.

The new Cortex-A725 is expected to be found in lead premium mobile devices by the end of this year.



Spotted an error? Help us fix it! Simply select the problematic text and press Ctrl+Enter to notify us.

Spelling error report

The following text will be sent to our editors: