Difference Between libvpx cpu-used 0 and 8

In the libvpx encoder (used for VP8 and VP9 video compression), the --cpu-used parameter controls the trade-off between encoding speed and video quality. Setting cpu-used 0 configures the encoder for maximum quality and compression efficiency at the cost of extremely slow encoding speeds, while cpu-used 8 prioritizes rapid encoding speed, resulting in faster processing but lower overall video quality and larger file sizes at a equivalent bitrate.

The Role of the cpu-used Parameter

The cpu-used parameter (often set via -cpu-used in FFmpeg) dictates how much computational effort the encoder exerts when compressing video frames. The encoder uses various algorithms to analyze motion, block partitions, and pixel differences. By adjusting this setting, you tell the encoder whether to perform an exhaustive, high-precision search for the best compression or to use fast heuristics to get the job done quickly.

libvpx cpu-used 0: Maximum Quality, Slowest Speed

When you set cpu-used 0, you are instructing the encoder to prioritize visual quality and compression efficiency above all else.

• Exhaustive Analysis: The encoder performs a thorough search of all possible block partitions, motion vectors, and intra-prediction modes.
• CPU Consumption: This setting utilizes maximum CPU resources and takes the longest time to encode. It is not uncommon for encodes at this level to run at a fraction of real-time speed.
• Best For: Archiving, final distribution of high-quality assets, or scenarios where encoding time is not a constraint but minimizing file size while maintaining pristine quality is critical.

libvpx cpu-used 8: Maximum Speed, Lowest Quality

Setting cpu-used 8 shifts the encoder’s focus entirely toward speed and low latency.

• Heuristics and Shortcuts: The encoder skips complex, computationally heavy analysis. It makes fast assumptions about motion and frame structures to compress the video as quickly as possible.
• Lower CPU Overhead: Because the mathematical analysis is simplified, the encoding process finishes rapidly and demands fewer system resources.
• Best For: Live streaming, real-time video conferencing, or rapid prototyping where immediate turnaround is necessary and minor compression artifacts or larger file sizes are acceptable.

Key Differences At a Glance

For most practical offline encoding tasks, a middle ground like cpu-used 1 to cpu-used 4 is recommended to achieve a reasonable balance between time spent encoding and the final quality of the video.