Reply to thread

Message

<blockquote data-quote="magao" data-source="post: 626844" data-attributes="member: 86026">There are definitely limitations on what can be played back using DxVA (especially on AMD cards) - numbers of reference frames, etc. Many fansubbed anime releases for example exceed these limitations.CoreAVC 2.0 removes or reduces most of these limitations, so there are a lot more files that can be offloaded. In addition, CoreAVC will determine if it is capable of playing back using hardware offloading, and if not will automatically use software, so quality will not suffer unlike if you pass an incompatible encoding to DxVA (which will produce green blocks, etc).CoreAVC does use a little more CPU than DxVA even when hardware offloading is working - the decoded video is passed back from the video card for post-processing, then passed back to the video card by the renderer.BTW, CoreAVC 2.1 will apparently include DxVA as well, so hopefully its algorithm will be:1. Can we use DxVA (compatible encoding for video card, connected directly to renderer)?2. Can we use CUDA (compatible nVidia card, compatible encoding)?3. Use software decoding.I'm guessing that the reason you're getting blocky decoding with Matroska is that the default MP4 and MKV profiles with Handbrake have different encoding settings (I know they do for audio - it won't let you put a DD or DTS stream into MP4 for example). Make sure you confirm that the encoding settings are identical for comparison. MKV and MP4 are just containers.In general, MKV is the preferred container these days, and looks likely to become the dominant container for the next while (has significant hardware support, can hold any kind of data e.g. subtitles, embedded fonts, etc).</blockquote>

[QUOTE="magao, post: 626844, member: 86026"] There are definitely limitations on what can be played back using DxVA (especially on AMD cards) - numbers of reference frames, etc. Many fansubbed anime releases for example exceed these limitations. CoreAVC 2.0 removes or reduces most of these limitations, so there are a lot more files that can be offloaded. In addition, CoreAVC will determine if it is capable of playing back using hardware offloading, and if not will automatically use software, so quality will not suffer unlike if you pass an incompatible encoding to DxVA (which will produce green blocks, etc). CoreAVC does use a little more CPU than DxVA even when hardware offloading is working - the decoded video is passed back from the video card for post-processing, then passed back to the video card by the renderer. BTW, CoreAVC 2.1 will apparently include DxVA as well, so hopefully its algorithm will be: 1. Can we use DxVA (compatible encoding for video card, connected directly to renderer)? 2. Can we use CUDA (compatible nVidia card, compatible encoding)? 3. Use software decoding. I'm guessing that the reason you're getting blocky decoding with Matroska is that the default MP4 and MKV profiles with Handbrake have different encoding settings (I know they do for audio - it won't let you put a DD or DTS stream into MP4 for example). Make sure you confirm that the encoding settings are identical for comparison. MKV and MP4 are just containers. In general, MKV is the preferred container these days, and looks likely to become the dominant container for the next while (has significant hardware support, can hold any kind of data e.g. subtitles, embedded fonts, etc). [/QUOTE]

Verification