Comparative study of energy efficiency of VVC and AV1 software decoders on mobile platforms using the VCAT framework
TL;DR
It is demonstrated that a highly optimized VVC software decoder can approach the energy efficiency of the leading software AV1 decoder, supporting the viability of software VVC decoder while dedicated hardware decoding support on mobile SoCs currently remains limited.
Abstract
The deployment of VVC/H.266 and AV1 in streaming services has outpaced the integration of dedicated hardware decoding units in mobile SoCs, making software decoders’ energy efficiency a critical concern for battery life. This paper presents a comparative energy efficiency study of three software decoders—DAMO266 (Alibaba DAMO Academy’s proprietary VVC decoder), dav1d v1.5.1 (open-source AV1 decoder), and VVdeC v3.0.0 (open-source VVC decoder)— using the Video Codec Acid Test (VCAT) framework. Test bitstreams were generated with the DAMO266 encoder and SVT-AV1 v2.3.0-rc1, with quality aligned at PSNR ≈ 39 dB. The evaluation dataset comprises Netflix open-source El Fuente and Chimera sequences covering 1080p30/60 and 4K30 at both 8-bit and 10-bit depth. Experiments were conducted on a Xiaomi 14 Pro (Snapdragon 8 Gen 3) under controlled conditions: 50% screen brightness, continuous decoding from full charge until 50% battery remained. Results show that DAMO266 sustains 77%–91% of the decoding duration of dav1d, while VVdeC sustains only 58%–69%. These findings demonstrate that a highly optimized VVC software decoder can approach the energy efficiency of the leading software AV1 decoder, supporting the viability of software VVC decoder while dedicated hardware decoding support on mobile SoCs currently remains limited.
