VP8 vs VP9 Encoding Differences in libvpx

This article explores the key technical and performance differences between VP8 and VP9 video encoding using the libvpx library. It covers compression efficiency, computational complexity, features, and encoder configuration parameters, helping you choose the right codec and optimize your video processing workflows.

Compression Efficiency and Quality

The most significant difference between VP8 and VP9 is compression efficiency. VP9 is designed to deliver the same visual quality as VP8 while using approximately 30% to 50% less bitrate. This improvement is achieved through more advanced coding tools. While VP8 is limited to 16x16 macroblocks, VP9 introduces “superblocks” of up to 64x64 pixels. These larger block sizes allow VP9 to compress flat, low-detail areas (such as skies or walls) much more efficiently, making it highly effective for high-definition (1080p) and ultra-high-definition (4K) video.

Computational Complexity and Encoding Speed

The superior compression of VP9 comes at the cost of significantly higher computational complexity. Encoding VP9 using the libvpx-vp9 library requires substantially more CPU resources and time than encoding VP8 with the standard libvpx encoder.

• VP8 has lower complexity, making it highly suitable for real-time communication (like WebRTC) on older or low-power hardware.
• VP9 is computationally heavy. While it excels at video-on-demand (VOD) and offline encoding where processing time is secondary to file size, real-time VP9 encoding requires modern hardware and aggressive speed presets to prevent frame drops.

Color Depth and HDR Support

VP8 is strictly limited to the 8-bit YUV 4:2:0 color format. VP9, however, introduces profile support for advanced color spaces. Specifically, VP9 Profile 2 and Profile 3 support 10-bit and 12-bit color depth, wider color gamuts (such as BT.2020), and High Dynamic Range (HDR) content. This makes VP9 suitable for modern, high-fidelity streaming, whereas VP8 is restricted to legacy standard dynamic range (SDR) formats.

Threading and Parallel Processing

The libvpx implementations handle multi-threading differently across the two formats:

• VP8 supports basic multi-threading, but its parallel processing capabilities are limited, often scaling poorly on CPUs with many cores.
• VP9 introduces tile-based encoding and row-based multi-threading (-row-mt 1 in FFmpeg). By dividing video frames into independent columns (tiles) and rows, libvpx-vp9 can distribute the workload across multiple CPU threads much more efficiently, drastically reducing encoding times on multi-core systems.

Command-Line Configuration Differences

When using tools like FFmpeg to encode with libvpx, the recommended rate control methods differ:

• VP8 is commonly encoded using standard variable bitrate (VBR) or constant bitrate (CBR) using the -b:v flag.
• VP9 performs best using Constrained Quality (CQ) mode. This is achieved by setting the target bitrate to zero (-b:v 0) and defining a Constant Rate Factor equivalent quality level using the -crf flag (typically between 15 and 35), alongside the -cap-level or -g parameters to manage GOP size.