How to optimize FFmpeg performance? What are the hardware acceleration solutions?

FFmpeg performance optimization is critical for processing large-scale audio/video tasks. Proper use of hardware acceleration and encoding parameters can significantly improve processing speed.

Hardware Acceleration

NVIDIA GPU Acceleration (NVENC/NVDEC)

bash
# Use NVDEC for decoding
ffmpeg -hwaccel cuda -i input.mp4 -c:v h264_nvenc output.mp4

# Use NVENC for encoding
ffmpeg -i input.mp4 -c:v h264_nvenc -preset fast -b:v 5M output.mp4

# Complete GPU acceleration workflow
ffmpeg -hwaccel cuda -hwaccel_output_format cuda -i input.mp4 \
  -c:v h264_nvenc -preset fast -b:v 5M output.mp4

# Specify GPU device
ffmpeg -hwaccel_device 0 -i input.mp4 -c:v h264_nvenc output.mp4

Intel QSV Acceleration

bash
# Use QSV for decoding and encoding
ffmpeg -hwaccel qsv -i input.mp4 -c:v h264_qsv output.mp4

# QSV encoding parameters
ffmpeg -i input.mp4 -c:v h264_qsv -preset fast -b:v 5M output.mp4

AMD VCE Acceleration

bash
# Use VCE for encoding
ffmpeg -i input.mp4 -c:v h264_amf -quality speed -b:v 5M output.mp4

VideoToolbox Acceleration (macOS)

bash
# Use VideoToolbox for encoding
ffmpeg -i input.mp4 -c:v h264_videotoolbox -b:v 5M output.mp4

# Use ProRes encoding
ffmpeg -i input.mp4 -c:v prores_videotoolbox -profile:v 3 output.mov

Encoding Parameter Optimization

Preset Selection

bash
# Ultra fast speed (lower quality)
ffmpeg -i input.mp4 -c:v libx264 -preset ultrafast output.mp4

# Very fast speed (medium quality)
ffmpeg -i input.mp4 -c:v libx264 -preset veryfast output.mp4

# Balance speed and quality
ffmpeg -i input.mp4 -c:v libx264 -preset medium output.mp4

# Best quality (slower speed)
ffmpeg -i input.mp4 -c:v libx264 -preset slow output.mp4

CRF Quality Control

bash
# High quality (larger file)
ffmpeg -i input.mp4 -c:v libx264 -crf 18 output.mp4

# Default quality
ffmpeg -i input.mp4 -c:v libx264 -crf 23 output.mp4

# Low quality (smaller file)
ffmpeg -i input.mp4 -c:v libx264 -crf 28 output.mp4

Thread Optimization

bash
# Set thread count
ffmpeg -i input.mp4 -threads 4 -c:v libx264 output.mp4

# Auto thread count
ffmpeg -i input.mp4 -threads 0 -c:v libx264 output.mp4

Multi-threaded Processing

Parallel Processing Multiple Files

bash
# Use GNU parallel
find input_dir -name "*.mp4" | parallel -j 4 ffmpeg -i {} -c:v libx264 output_dir/{/.}.mp4

# Use xargs
find input_dir -name "*.mp4" | xargs -P 4 -I {} ffmpeg -i {} -c:v libx264 output_dir/{/.}.mp4

Segmented Processing

bash
# Segment and transcode then merge
ffmpeg -i input.mp4 -c copy -f segment -segment_time 60 segment_%03d.mp4
for f in segment_*.mp4; do ffmpeg -i "$f" -c:v libx264 transcoded_"$f"; done
ffmpeg -f concat -safe 0 -i <(for f in transcoded_segment_*.mp4; do echo "file '$PWD/$f'"; done) -c copy output.mp4

Memory Optimization

Reduce Memory Usage

bash
# Use streaming processing
ffmpeg -i input.mp4 -c:v libx264 -f null -

# Limit buffer size
ffmpeg -i input.mp4 -c:v libx264 -bufsize 1M output.mp4

Performance Analysis

View Encoding Information

bash
# Show detailed encoding information
ffmpeg -i input.mp4 -c:v libx264 -v verbose output.mp4

# Show performance statistics
ffmpeg -i input.mp4 -c:v libx264 -stats output.mp4

Benchmark Testing

bash
# Test decoding performance
ffmpeg -benchmark -i input.mp4 -f null -

# Test encoding performance
ffmpeg -benchmark -i input.mp4 -c:v libx264 -f null -

Common Performance Issues and Solutions

High CPU Usage

bash
# Reduce encoding complexity
ffmpeg -i input.mp4 -c:v libx264 -preset ultrafast -tune fastdecode output.mp4

# Use hardware acceleration
ffmpeg -hwaccel cuda -i input.mp4 -c:v h264_nvenc output.mp4

High Memory Usage

bash
# Reduce thread count
ffmpeg -i input.mp4 -threads 2 -c:v libx264 output.mp4

# Use streaming processing
ffmpeg -i input.mp4 -c:v libx264 -f null -

Slow Processing Speed

bash
# Use faster preset
ffmpeg -i input.mp4 -c:v libx264 -preset ultrafast output.mp4

# Enable hardware acceleration
ffmpeg -hwaccel cuda -i input.mp4 -c:v h264_nvenc output.mp4

Best Practices

1. Choose appropriate acceleration based on hardware: Prioritize hardware encoding when GPU is available
2. Balance speed and quality: Choose appropriate preset and CRF based on application scenario
3. Set reasonable thread count: Typically set to 1-2 times the number of CPU cores
4. Use streaming processing: Consider segmented processing for large files
5. Monitor resource usage: Use system monitoring tools to observe CPU, GPU, and memory usage

Performance optimization needs to be adjusted based on specific hardware configuration and application scenarios. It is recommended to conduct multiple tests to find the best parameter combination.