Tencent's Canghai V2 Chip Enters Mass Production: What the New Capabilities Mean for Video Encoding

Tencent announced that its second‑generation self‑developed video codec ASIC “Canghai” V2 has been successfully powered on and will begin full online service in the second half of 2026, bringing comprehensive upgrades in codec, ultra‑HD rendering and cost efficiency.

V1, which was lit in March 2022 and entered mass production in 2023, has already been deployed in more than 100,000 chips across live streaming, on‑demand and cloud‑gaming scenarios. V2 is described as a “polygon‑all‑round warrior” that simultaneously supports full‑spec codec enhancement, end‑to‑end ultra‑HD, and GPU rendering.

In the MSU hardware video‑encoding competition held on 26 May, Canghai achieved first place in all speed tiers (30–240 fps) for SSIM, PSNR and VMAF, surpassing competitors by over 30 % on several metrics, thereby pushing the compression‑rate ceiling of hardware encoders.

The chip supports both H.265 and the newer H.266 standards with a unified hardware architecture. Compared with V1, compression efficiency improves by more than 10 % for H.265 and over 30 % for H.266, while per‑chip decoding/encoding throughput roughly doubles, improving compute‑price ratio.

V2 also integrates CPU, GPU and computer‑vision (CV) blocks, enabling on‑chip intelligent analysis and picture‑enhancement functions such as super‑resolution, sharpening, de‑haze, denoising, block‑artifact removal and color enhancement.

Its industry‑leading GPU core allows ultra‑low‑cost rendering for cloud gaming; encoding and rendering share memory on the same die, eliminating data copies, reducing latency and further lowering the compression rate by more than 5 %.

Through Tencent Cloud Media Processing Service (MPS), the V2 solution will be offered via public cloud, private cloud and SDK/LIB integrations, delivering higher‑definition, lower‑bitrate media processing to enterprises and developers.