Vivo held a dual-core × image technology communication meeting, the meeting introduced the depth of cooperation between Vivo and MediaTek and released the second generation of self-research image chip V2.
Since 2021, Vivo has launched two generations of self-research chips, V1 and V1+. This time, the self-research chip V2 adopts a new iteration of AI-ISP architecture, which brings a comprehensive improvement of compatibility and functionality and significantly upgrades the on-chip memory unit, AI computing unit and image processing unit. The FIT dual-core interconnection technology is proposed to establish a new high-speed communication mechanism between the self-research chip V2 and the flagship platform of Dimensity 9200, so that the two chips with completely different architectures and instruction sets can complete the dual-core interconnection synchronization within 1/100th of a second, realizing optimized coordination and high-speed collaboration of data and arithmetic power.
According to the introduction, thanks to the near memory DLA (Vivo’s own AI deep learning gas pedal) module and large-capacity dedicated on-chip SRAM (high-speed low-consumption cache unit), the self-research chip V2 has re-matched the arithmetic capacity, arithmetic density and data density to significantly increase the capacity and computing speed of on-chip cache. Compared with the usual DDR external memory design of NPU, SRAM data throughput power consumption can be theoretically reduced by up to 99.2%, and the energy efficiency ratio is improved by 200% compared with traditional NPU.
The FIT dual-core interconnect and near-storage DLA structure design enables the AI-ISP architecture of the self-research chip V2 to be established and complement the ISP algorithm and arithmetic power of the platform chip NPU to achieve the ultimate image processing effect and the ultimate energy efficiency ratio.
At this communication meeting, Vivo also brought advanced self-research image algorithms such as telephoto image, motion capture and dark light capture, etc. With the solidified algorithm capability of self-research chip V2 and the support of new HDR, NR and ProMEMC, the mobile image technology of Vivo will be pushed to a new height.
The Ultra Zoom EIS technology combines IMU, OIS and EIS modules to effectively counteract shake during high zoom shooting and ensure the stability of the preview image, even when handheld. It can steadily transport the mirror even when handheld.
To solve the problem of shutter delay and shorten the gap between cell phone shooting and professional cameras, Vivo has developed “zero latency” capture and a new generation of motion detection algorithms. By optimizing the image processing process to enhance the sensor start-up speed, the shutter delay time is reduced to 30ms, reaching the level of professional cameras.
To meet the full-scene image needs of professional users, Vivo also realized the dark-light capture function through algorithm overlay and full-link optimization of photography, which significantly improves the sensor’s light sensitivity in dark-light scenes; through multi-frame fusion technology and self-developed RawEnhance 2.0 algorithm to realize motion picture overlay and eliminate trailing shadows, enabling users to easily capture high-quality dynamic images in dark-light environments.
Vivo will launch the world’s first Dimensity 9200 processor, which is based on TSMC’s second-generation 4nm process technology. Vivo said it was involved in the development of the Dimensity 9200 flagship platform at an extremely early stage, and through 20 months of close cooperation between the two sides, it brought five joint R&D features, including MCQ multi-loop queue, King of Glory adaptive image quality mode, chip eye protection, APU framework fusion and AI airport mode. The five jointly developed features are the core of the two companies.
MCQ Multi-Cycle Queue is a new processing engine redefined to bring out the ultimate performance of CPU, supporting up to 8 channels of data transfer between CPU and UFS, making application switching and background download wake-up faster and smoother.
The Adaptive Image Quality Mode of King of Glory is a black technology jointly developed by Vivo, MediaTek and King of Glory, based on the MAGT game adaptive cycle. When the adaptive function is turned on, the game frame rate is close to the full frame with an average variance of 0.92, which is the top level in the industry, when it runs for 1 hour at 26℃ under the King of Glory 120+ extreme configuration.
The chip’s eye protection technology reduces blue light in real-time by monitoring the percentage of blue light on the screen and reducing it in real-time through innovative algorithms and hardening it into IP, so that the percentage of high-energy visible blue light is less than 5% and the degree of color shift is reduced by 12%. At the same time, the screen color effect is dynamically adjusted according to the monitoring results to reduce blue light while ensuring that the screen does not deviate from the color.
APU framework fusion is based on the hardware features of the sixth-generation APU of the Dimensity 9200 flagship platform and encapsulates MediaTek’s underlying general-purpose capabilities into its own VCAP heterogeneous computing acceleration platform. The platform is deeply optimized from the bottom of the platform to the framework layer, allowing algorithms to be collaboratively scheduled among multiple processors, bringing significant energy efficiency improvements, and thus introducing features such as camera ultra-clear documents and live text.
AI airport mode can save an average of 30% energy during the user’s flight; after turning off the flight mode, the network connection can be quickly restored in as little as 1.52 seconds, making the restoration of network speed a step faster.
In addition to joint research and development, Vivo and MediaTek also jointly tuned a number of details to help the flagship platform of Dimensity 9200 perform at its best, with a score of more than 1.28 million in Rabbit Labs.
In the game scenario, Vivo brings technologies such as game super score, and optimizes the game experience in multiple aspects and in all aspects through the fast start engine and network acceleration engine; in the video scenario, for the first time, advanced cyclic parallax network is realized with extremely low power consumption, which greatly optimizes energy efficiency performance and reduces energy consumption by 15%. Especially in 4K 60 fps extreme video scenarios, the power consumption of the Dimensity 9200 is 25% lower than that of the Dimensity 9000.
