Chipset maker MediaTek on Tuesday unveiled the ‘Dimensity 5G Open Resource Architecture’ that provides brands with more flexibility to customise key 5G mobile device features to address different market segments.

The open resource architecture gives smartphone brands closer-to-metal access to customise features for cameras, displays, graphics, artificial intelligence (AI) processing units (APUs), sensors and connectivity sub-systems within the Dimensity 1200 chipset.

“MediaTek is collaborating with the world’s largest smartphone brands to unlock customized consumer experiences that differentiate flagship 5G smartphones,” Yenchi Lee, Deputy General Manager of MediaTek’s Wireless Communications Business Unit.

“Whether it’s novel multimedia features, unmatched performance, brilliant imaging or more synergy between smartphones and services, our architecture device makers can tailor their devices to complement a variety of consumer lifestyles,” Lee added.

With access to the in-chip, multi-core AI and display processors, brands can tailor multimedia experiences and unlock more synergy between the chipset and the smartphone’s display and audio hardware.

Brands can utilise MediaTek’s AI picture quality (AI-PQ) and AI Super Resolution (AI-SR) or can develop their algorithms with customized video parameters and scenario detection backed by their deep learning data or dynamically adjust display elements.

The open resource architecture gives brands the freedom to fine-tune a device’s performance and power efficiency through workload assignments customisation across the chipset’s processing resources.

This includes the central processing unit (CPU), graphics processing unit (GPU), deep learning accelerators and visual processors.

Using the foundation of MediaTek’s NeuroPilot, brands have better access to MediaTek’s deep learning accelerator (DLA) within the MediaTek APU. This allows brands to apply customisations to the multi-threaded scheduler and customised algorithms.

They can also choose to access the DLA’s capabilities to integrate INT8, INT16 and FP16 data types within a single AI hardware, which improves the device’s precision, performance and power efficiency.