Moscow State University has been running thorough benchmarks of hardware and software video encoders for years, including testing different presets: <a href="https://www.compression.ru/compression.ru/video/codec_comparison/2022/" rel="nofollow">https://www.compression.ru/compression.ru/video/codec_compar...</a><p>Note: this is in no way an endorsement of Russia's invasion of Ukraine. In the past, MSU has accepted payment for some of their reports. I suggest you do not pay for these (assuming it's even possible now).
> <i>Please be aware that some images may not load on this page unless your browser supports JPEG-XL</i><p>The site could provide a WebAssembly decoder to make the JPEG-XL images work for everyone.<p>For example, here's a WebAssembly decoder: <a href="https://github.com/niutech/jxl.js">https://github.com/niutech/jxl.js</a><p>Demo: <a href="https://niutech.github.io/jxl.js/" rel="nofollow">https://niutech.github.io/jxl.js/</a>
Thats a great comparison. ~x265 medium is quite impressive for a hardware encoder, and it makes more sense than some questionable claims I saw from YouTubers.<p>I'm sure you have already read this, but AMD claims their new encoding ASIC can do even better: <a href="https://www.anandtech.com/show/18805/amd-announces-alveo-ma35d-media-accelerator-av1-video-encode-at-1w-per-stream" rel="nofollow">https://www.anandtech.com/show/18805/amd-announces-alveo-ma3...</a><p>> All told, AMD is making some fairly aggressive image quality claims with the Alveo MA35D; H.264 and H.265 image quality should be similar to x264 Medium and x265 Medium presets respectively, while the card’s AV1 encoding quality should be comparable to x265 slow.<p>Since it can do so many parallel streams, I can already see some mad hatter setting up av1an with one.
It was interesting to see the hardware VP9 encoder included. I’ve found it to be the best streaming format for compatibility (ignoring h264) - a single stream can be watched on all devices and browsers
I dont know how I feel about testing of video encoder quality by someone showing me graphs with gray text on gray background. Its almost like he doesnt want me to read those and just trust his conclusions :)
<i>> These tests are more thorough than mine, but aren't accompanied by much explanation or any VAAPI results for Linux</i><p>Isn't VAAPI AV1 encoding not yet implemented for AMD?<p>May be Vulkan video will become available for it before it happens.
Not sure about their videos but I could not achieve the same quality at a limited bandwidth using NVENC vs x264. x264 quality was clearly superior. One does not have to do any pixel-peeping to notice.
Using SSIMULACRA2, I've compared Intel's QuickSync Video to Nvidia's NVENC encoder across Linux & Windows. They compete against other dedicated encoding hardware for different codecs across different GPUs, including an Intel iGPU.
I was recently asked why so few chips support decoding AV1 in hardware.<p>There were multiple mutually incompatible AV1 versions early in development, and some companies chose to not to implement either until the situation settles down.<p>But by that time X265 was already deployed en masse in the wild. And now X266 is about to take over.<p>Many will chose to leapfrom AV1 straight to X266.
> Please be aware that some images may not load on this page unless your browser supports JPEG-XL.<p>lol really? Who writes a website that only works on Firefox Nightly with a feature flag enabled?<p>I would have been interested in reading this otherwise.
The GPU architecture of Arc A770 is Xe, while the iGPU of Core i7 13700K is UHD, so they are clearly different. On the other hand, the iGPU of Ryzen 7000 is RDNA2, so it can be assumed that it has the same quality as RADEON 6000, although with a very large scale difference. Therefore, the conclusion is that the iGPU of Ryzen 7000 is the best option for having an AV1 hardware encoder that can save money for eating a lot of nice ramen.