ARM ramps up Mali T604 with OpenCL



Posted by TG Daily Staff

ARM has submitted its flagship Mali T604 graphics processing unit (GPU) to the Khronos Group for full profile conformance with OpenCL 1.1.



An ARM spokesperson told TG Daily the UK-based processor licensing company is the first GPU IP supplier to submit conformance for full profile OpenCL - which will help pave the way for desktop GPU computing capabilities across the mobile, embedded and smart-TV markets.



ARM ramps up Mali T604 with OpenCL



Indeed, conformance with the standard is expected to create a consistent and effective platform for hardware architects, as well as provide tangible benefits for end users, both in terms of optimized performance and longer battery life.

"GPU computing is a trend where the computational performance of the GPU, historically used for graphics, is harnessed to augment the main processor (or CPU) for certain applications where the GPU architecture will be more effective," the spokesperson explained.



"The result? Improved performance, energy-efficiency and a more efficient use of the system as a whole - making computational photography, computer vision, advanced imaging, point-of-interest extraction and augmented reality possible [on a wide range of devices]."

The Mali-T604 is the first GPU based on ARM’s Midgard architecture and offers scalability from one to four cores. 

It is traditionally marketed as a complete embedded graphics and GPU Compute acceleration platform - facilitating the development of advanced user interfaces (UIs), gaming capabilities and GPU Compute apps for a wide range of devices, including smartphones, smartbooks, HDTVs and portable gaming.

Multicore scheduling and performance scaling is fully handled within the graphics system, while the job manager within Mali-T600 Series GPUs offloads task management from the CPU to the GPU - enabling seamless load balancing across active shader cores. Compute tasks span the CPU and GPU as well as other available compute resources, accessing data more efficiently during shared processing in a filly heterogeneous system.