#RAWcooked soon with a #FFV1 GPU-based encoder and decoder. Over the past few months, we’ve been collaborating on adding a GPU-based encoder and decoder for lossless #FFV1 video to #FFmpeg.
More at https://www.linkedin.com/feed/update/urn:li:share:7327717894555865088/
www.linkedin.comRAWcooked | MediaArea#RAWcooked soon with a FFV1 GPU-based encoder and decoder. It's with a lot of gratitude that we consider all the research, testing, support, and contributions that have gone into the RAWcooked project over the past (nearly) 8 years. From the beginning, RAWcooked emerged as a response to the challenges of film scanning for preservation, and it has continued to evolve thanks to your efforts and support through licensing.
We’re excited to share a significant new development with you first: Over the past few months, we’ve been collaborating on adding a GPU-based encoder and decoder for lossless FFV1 video to #FFmpeg. While GPU development is complex, it offers substantial speed improvements over traditional CPU-based encoding and decoding.
Now we have a proof of concept of the FFV1 GPU encoder and decoder up and running. On a Linux system with an nVidia RTX 4070 Ti graphics card, we're seeing performance around 40 fps for encoding and 55 fps for decoding 4K2K (4096x2160) 10-bit content. This is almost double the speed of a CPU-based setup at a comparable cost.
This GPU acceleration currently works with graphics cards from AMD, NVIDIA, Intel, and ARM. Based on our testing, the best performance-to-cost ratio is seen with the NVIDIA 4070 Ti and AMD RX 6900XT on Linux systems. Unfortunately, high-end Tensor-based GPUs like the NVIDIA 6000 ADA or H100 have not delivered good results, and Apple’s integrated GPUs are not supported due to the lack of Vulkan-compatible drivers, which our system relies on.
We’d like to refine our recommendations for optimal GPU environments, and your help would be invaluable. If you’re able, please run benchmark tests on your available platforms, drop us an email at info@mediaarea.net for a downloadable benchmark script.
This feature has been a substantial investment of time and resources. In order to sustain and further develop this capability, we’re making GPU-accelerated encoding and decoding available in RAWcooked as a licensed feature at a one-time cost of 1000 € / $1200 / £900. If you’re interested in supporting our work and accessing this feature, drop us an email at info@mediaarea.net .
Thank you again for being part of the RAWcooked journey.
https://lnkd.in/eGTWefDJ
Teresita
