Fastvideo MJPEG codec hits 5500 FPS on NVIDIA RTX 4090
Fastvideo has announced a new CUDA-based Motion JPEG codec capable of achieving 5500 fps encoding for Full HD video on NVIDIA RTX 4090 GPUs. The codec is designed for high-speed imaging pipelines that require individual frame accessibility, offering a high-throughput alternative to inter-frame codecs like H.264.
Key Takeaways
- Encoding throughput reaches 5500 fps for Full HD (1920x1080) at 8-bit color using single-threaded GPU processing.
- Decodes Motion JPEG streams at up to 1500 fps on the NVIDIA GeForce RTX 4090 under single-threaded conditions.
- Supports 12-bit MJPEG alongside standard 8-bit grayscale and 24-bit color formats with variable subsampling (4:4:4, 4:2:2, 4:2:0).
- Integrates with FFmpeg for reading and writing streams and is compatible with Windows 10/11 and Linux via CUDA 12.6.
- Available as part of the Fastvideo SDK, enabling full GPU pipelines from raw acquisition to final compression.
Why It Matters
High-speed MJPEG encoding on CUDA hardware fills a critical gap for industrial and medical imaging where inter-frame codecs like H.264 fail to sustain ultra-high frame rates. By processing frames independently, this codec avoids the motion estimation overhead that typically bottlenecks high-speed pipelines, ensuring real-time performance at 1000+ fps. For the broader streaming ecosystem, this provides a pathway for ultra-low latency internal production and diagnostic monitoring before final distribution transcoding. Watch for Fastvideo's upcoming real-time MJPEG streaming module, which could shift how high-bandwidth raw diagnostic video is moved across local networks.
Additional Context
The demand for high-speed intra-frame compression is intensifying as industrial vision and medical imaging move toward higher bit depths and frame rates that exceed the capabilities of standard hardware encoders. Per Technexion (May 2026), while inter-frame codecs like H.265 and AV1 offer superior storage efficiency, they introduce significant latency and computational overhead due to motion estimation between frames. This makes them less suitable for mission-critical applications like surgical robotics or high-speed industrial inspection, where every frame must be individually seekable and extractable without temporal artifacts.
In the broader market, NVIDIA has continued to iterate on its proprietary libraries to meet these needs. According to NVIDIA's April 2026 documentation, the latest nvJPEG library updates for the Ada Lovelace and upcoming Blackwell architectures emphasize hybrid decoding and hardware acceleration on data-center-grade A100 and H100 units. However, for B2B developers using consumer-grade hardware like the RTX 4090, external SDKs like Fastvideo often provide more granular control over the full ISP pipeline, including demosaicing and denoising stages, before the final compression step.
The shift toward royalty-free and high-efficiency codecs like AV1 is also influencing the high-speed sector. Per forasoft (August 2026), while AV1 hardware encoders on the RTX 40-series can hit roughly 500 fps for Full HD, they still lag significantly behind the raw throughput possible with optimized MJPEG implementations. For precision engineers—a segment that Netint (May 2026) identifies as controlling 13% of the market but holding disproportionate influence—the priority remains maximizing throughput and minimizing latency over minimizing total file size, which maintains MJPEG's relevance despite the maturity of modern inter-frame alternatives.
Read full article at fastcompression.com
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source