Compare older CPU encoding with NVENC video capture techniques.
Compare older CPU encoding with NVENC video capture techniques.
I aim to record video using a StarTech DVI USB capture box and a Blackmagic analog capture device. I’m unsure which method works best. I currently set OBS to the 1080/60 preset with a large file size and around 40 Mbps bitrate. That’s more than sufficient. I could run CPU encoding on an i5-4590, which should handle high bitrate 1080/60 easily. Alternatively, I could use NVENC with a spare Quadro M2000 card in an LGA 775 system. I’d need to install a USB 3 PCIe card since both devices require it. I haven’t used NVENC much, but it should perform well, similar to a GTX 960. I’d start with an i5 system in a SFF case, which has a quiet cooler and is easy to place anywhere without disturbing recordings. If the Quadro delivers much better quality, I’d opt for it. Power usage isn’t an issue, heat isn’t a big concern, and keeping noise low is important to me.
Generally, GPU encoding struggles to match CPU encoding quality when all other conditions are equal. However, recent Intel AV1 GPU encoders perform comparably to slower x265 presets in certain scenarios. This isn't the case for your setup; the M2000 card is quite outdated. Unless you conduct side-by-side comparisons of your own videos on both platforms and closely examine pixel data, I suggest prioritizing CPU encoding for the best results.
The CRF configuration works by storing video frames at varying quality levels, similar to saving images at different compression settings. It begins with the highest quality (100%) and adjusts based on previous frames to estimate compression efficiency and quality loss. For x264, a CRF value of 2 to 5-6 offers near-lossless results, akin to 99% JPG quality, though file sizes increase significantly. Levels around 8-12 provide bluray-like clarity, while 18-22 correspond to approximately 15-25 Mbps bitrate—similar to large movie files. In practice, starting with the "ultrafast" preset in CRF mode balances CPU usage and file size. Faster modes use more CPU for compression but yield quicker results, whereas slower modes offer better compression efficiency. Adjust settings based on your hardware: ultrafast works well with most CPUs, but superfast may be needed for optimal performance. Proper color settings and format selection are crucial, especially when using capture devices that default to specific color spaces. For SD content under 1024x768, ensure correct color space is applied, and consider using advanced options like --color-prim to fine-tune output.
you could capture using preset ultrafast settings, CRF 2, and uncompressed audio formats like FLAC to achieve 50-100 mbps bitrates for 1080p videos with near-lossless quality. afterward, process the large files with Handbrake or similar tools (such as MeGUI) to re-encode them using x264 with CRF 15-20, very slow speed, resulting in high-quality output at reasonable bitrates like 10-25 mbps. audio can remain in FLAC or be compressed to Opus at 192 kbps or AAC at 256 kbps. at very slow speed, your CPU might operate at 10-20 frames per second, making the re-compression process take 3-4 hours for an hour of video. alternatively, if your graphics card supports hardware encoders like AV1 or HEVC, use those to compress to AV1 or HEVC, which preserves more quality at the same bitrate compared to H.264.