Taipei, Friday, Mar 29, 2024, 03:32

News

MediaTek Reveals More Details About Dimensity 1000 5G SoC

By Korbin Lan
Published: Dec 02,2019

TAIPEI, Taiwan - Three days after launching the Dimensity 1000 5G chip, MediaTek Inc. once again held a rare technical briefing on the chip. All department heads responsible for development were present at the briefing where the company provided a more in-depth explanation of the operational performance, 5G technology, and architecture of the chip.

More on This

MediaTek to Open New IC Design Center at Purdue University in Indiana

MediaTek announced their commitment to accept a state transition assistance package from the Indiana Economic Developmen...

ITRI and TSMC Announces a New SOT-MRAM Technology at VLSI 2020

TAIPEI, Taiwan – Taiwan’s ITRI(Industrial Technology Research Institute) announced a world-leading SOT-MRAM technology which is co-developed with TSMC...

The department heads in attendance on the occasion who provided explanations of the technology included Product Marketing Division, Wireless Communications Division Manager Nian Yu-cun, Wireless Communications Division General Manager Li Zong-lin, and Wireless Communications Division Director of Product Marketing Thomas Ho, all of whom are important members responsible for mobile phone product development.

With 7-nm process, 5G modems will support SA / SA in the future

The first of its kind in terms of 5G modem chips, a 7-nm process was utilized for the 5G modem for the Dimensity 1000, giving it double-digit improvement in power and performance over the previous 12-nm process.

In addition, this type of modem supports dual carrier aggregation (2CC CA) technology, which boosts coverage by 30%; furthermore, it is the first modem in the world to support 5G + 5G dual card dual standby functionality. Nian Yu-cun further emphasized that MediaTek’s dual sim dual standby fully supports voice over NR technology with improved performance in voice and data communications.

In consideration of the varying business plans of different global telecommunications providers, MediaTek’s Dimensity 1000 simultaneously supports NSA and SA specifications; however, as 5G networks are popularized and become more mature in the future, MediaTek will launch SA/SA versions.

As for the reasoning behind selecting Sub-6GHz as the first consideration for development, Li Zong-lin stated that MediaTek possesses both Sub-6GHz and mmWave technologies; however, as Sub-6GHz is currently used by 48 operators worldwide, it was definitively selected as the first specification.

The four cores are not overclocked – optimizing performance and power consumption

In terms of CPU computing performance, unlike other manufacturers’ approach of using one large core, MediaTek utilizes the quad-core Cortex-M77, which is designed with four small cores. Integrated with a balanced architecture with excellent computing power, it is clearly superior to the competition in terms of performance and power consumption.

Thomas Ho stated that the Dimensity 1000 has a primary core operating frequency of 2.6GHz; however, there is no deliberate overclocking. As a result, it does not overheat after running for long periods of time; therefore, from the standpoint of mobile phone manufacturers there is no need to add heat dissipation to the design, and this helps save overall design costs.

For the GPU, MediaTek utilizes the latest generation Mali-G77 computing core, which employs a nine-core architecture, facilitating a 40% improvement in performance over the previous generation (G76).

APU 3.0 hardware accelerates AI and facial recognition

However, the biggest highlight of the technical briefing was the artificial intelligence processor (APU 3.0), with completely new architecture, revealing the speed and depth of MediaTek's AI technology layout.

Thomas Ho explained that MediaTek was the first operator in the world to integrate AI engines into their computing chips and to invest in the long-term research and development of independent AI technology. He pointed out that the new APU 3.0 utilizes an architecture which features two large cores and three small cores along with an additional tiny core, with MediaTek’s independently-developed AI cores as the large cores for handling major AI tasks and VPUs as the small cores to assist with image computing power.

Meanwhile, the tiny core is designed to assist facial recognition tasks. Thomas Ho stated that unlike the software models utilized by other manufacturers, MediaTek employs a hardware design to handle facial recognition operations. This is not only faster but also consumes less overall power.

In addition, he also explained that MediaTek makes use of a hybrid architecture for their AI technology R&D in which the AI acceleration engine is combined with the VPU, and operations are carried out through network computing. At the same time, the company adheres to a path of independent research and development.

CTIMES loves to interact with the global technology related companies and individuals, you can deliver your products information or share industrial intelligence. Please email us to en@ctimes.com.tw

2677 viewed

Most Popular

comments powered by Disqus