Arm aims latest processors at high-end phone gaming

Author: EIS Release Date: Jul 6, 2022

Arm’s is claiming “up to 28% more performance and up to 16% power reduction across a range of workloads, such as gaming”, for the phone GPUs and CPUs announced in its ‘2022 total compute solutions’ (TCS22).

 

Arm TCS22 CPUsThe CPUs are the second generation with the Armv9 architecture, and include Cortex-X3 and Cortex-A715. There are also updates to Cortex-A510 and the DSU-110 (DynamIQ shared unit).

 

The four blocks above are intended to form the application processing heart of high-end phones, with A715, A510 and the DSU working in big.LITTLE clusters in the engine room, and X3 flying over the top.

 

Arm TCS22 GPUsAlongside these, there is a new flagship GPU, called Immortalis, with the Immortalis-G715 being the first Arm GPU to offer hardware-based ray tracing support on mobile, said the company – ray tracing is used to improve the look of virtual lighting and shadows in 3D images.

 

At the same time, it announced the Mali-G715 GPU, which includes variable rate shading to trade power consumption against gaming performance, and runs in clusters with 7 – 9 cores. Then there is the Mali-G615 GPU (six cores or fewer) – following on from last year’s Mali-G610. “The new GPUs build upon the Mali-G710 GPU, with 15% energy efficiency improvements,” said the company.

 

“2x architectural machine learning improvements”, is claimed for the new GPUs.

 

The X3 CPU is honed for single-threaded performance, and it is claimed to deliver “a 25% performance improvement compared to the latest Android flagship smartphone and a 34% performance improvement compared to the latest mainstream laptops”. There are SPECRate2017_int_base figures with differing caveats.

 

Cortex-A715 is said to have 20% better energy efficiency and 5% more performance  (SPECint_base2006) compared to Cortex-A710 “matching the performance of Cortex-X1”, said Arm.

 

Cortex-A510 is last year’s ‘LITTLE’ Armv9 CPU, which has been updated with a 5% power reduction, and the updated DSU-110 now supports clusters of up to 12 cores – big.LITTLE is a power-performance improvement technique that slides an application between low-performance low-energy cores and high-performance high-energy cores as its needs change.