eastwest connect

Tencent discloses the latest progress of self-developed chips to solve codec problems


It is learned on April 17 from Digital Intelligence Frontline that On April 15, the results of the MSU Hardware Video Coding Competition held by Moscow State University were announced. After several months of rigorous testing, Tencent’s self-developed codec chip “Canghai” took the first place in all 8 scores of the two tracks it participated in. It is understood that tens of thousands of Canghai chips have been put into use in business scenarios, serving Tencent’s self-developed business and public cloud customers.

MSU is the most influential top-level event in the field of video compression. It has been held for 17 consecutive sessions, attracting well-known domestic and foreign companies including Amazon, Nvidia, Intel, AMD, Byte, Kuaishou, Alibaba and Tencent.

Specifically, with multiple evaluation indicators such as SSIM (Structural Similarity), PSNR (Peak Signal-to-Noise Ratio) and VMAF (Video Multi-Method Evaluation Fusion), Canghai chips are significantly ahead of standard hardware such as GPUs in the industry. It is worth mentioning that at the high frame rate gear of 120fps, Canghai can save another 10% bit rate compared with the industry-leading ASIC level under the same image quality.

As an interactive mode and key technology in the era of true Internet, the encoding and decoding capabilities behind audio and video are crucial. Based on the joint optimization of self-developed chips and software, Tencent Cloud’s audio and video encoding and decoding capabilities have reached the industry-leading level.

The chip is the core part of the hardware. Facing scenarios with strong business needs, Tencent has a long-term chip R&D plan and investment, combined with computing power scenarios, to promote the scale of self-developed chips. At present, Tencent has achieved full coverage of chip end-to-end design and verification.

In addition to the Canghai chip, Tencent’s self-developed AI reasoning chip “Zixiao” adopts a self-developed storage and computing architecture and a self-developed acceleration module, which can provide up to 3 times the computing acceleration performance and an overall cost savings of more than 45%. It has also been put into use in internal business; The high-performance network chip “Xuanling” helps cloud computing scenarios to achieve zero-loss virtualization.

Based on continuous breakthroughs in underlying software and hardware technologies, Tencent is continuing to explore optimal performance and best cost-effectiveness to help more enterprises and developers innovate on the cloud.