Originally Posted by Mfusick
I generally would want MadVR on 4k machine or projector/display system so I would not choose a Pentium chip, rather a quad core intel with a dedicated robust GPU. That was my point. You don't need 4k support on the low end. It makes no sense. 4k is a high end feature.
If 4K is a high end feature, then what is MadVR ?
MadVR is an extremely high-end feature, useless in my opinion, that I have used only for experimental reasons.
It's not high-end from a cost view, but from a "geek" view.
There are no essentials features that EVR/EVR-CP renderer is missing, compared to MadVR for the majority of sources, using H.264 video format.
For example the hardware de-interlacing of Intel's QuickSync VPP engine, are more than enough for every interlaced source.
I could say the same for scaling algorithms or other VPP operations.
Also, I would like to have the ability to decode a 4K clip and display it to a 1080p TV or PC monitor.
It's the same as of decoding a 1080p clip and display it to a 720p or lower resolution display.
Because I can and because there are clips in 4K and maybe I don't have right now a 4K display. It's that simple!
BTW, using - the last almost 2 months - a budget Pentium Haswell processor, I have to say that I will probably never go back to a fast Quad core CPU + a dedicated GPU.
Because 2D performance of this 50€ chip is equal to a 150€ dedicated GPU and the Video processor performance of QuickSync is the fastest in the world and fully featured.
The dual-core CPU of the very advanced Haswell architecture, is more than enough for anything but 3D games or professional applications of course (like Video/3D editing etc)
And this 50€ chip can decode 4K H.264 up to 80fps in HW!
Besides 3D gaming, the average user and the HTPC user is completely covered by a fused CPU-GPU-VPU processor like Pentium.
Originally Posted by RLBURNSIDE
I don't really get who in their right mind would use h.264 for 4K over HEVC / h.265, given that it's double the size and less quality. Makes no sense whatsoever.
Do you think that the hundreds or thousands of digital video broadcasters that in 2014 still using MPEG-2! have a "wrong" mind ?
It's time to bust another myth.
H.265 is not a revolution over H.264.
It's just a small, tiny evolution.
H.264 was a real revolution regarding video codecs/ formats.
The distance between H.264 and MPEG-2 is huge in terms of bandwidth/ quality, compared to the H.264 vs H.265 discussion.
Also if we say that 4K H.264 ecosystem is small, then 4K H.265 ecosystem is not existant!
4K H.264 is trying to be established since 2010 and I'm sure that 4K H.265 will come after 4K H.264 or in parallel with it and not over 4K H.264.Referring to the luck of 4K H.265 HW acceleration right now is just another excuse to avoid discussion about not delivering 4K H.264 HW acceleration yesterday
Originally Posted by RLBURNSIDE
The higher end content and resolution you get, the more you need GPU and CPU hardware acceleration in order for your entire system to not chug to a crawl. Efficiency is a good thing to pursue, even in the high end space, and you will need lots of cheaper / integrated GPUs supporting the real-time decoding of HEVC in order for it to get off the ground, similar to how earlier PCs used to struggle with realtime playback of anything greater than 720p, now we are leaping up to 4K and at higher bit depths and dynamic range too.
That's why fixed-function HW like QuickSync, VPx and UVDx are extremely useful in terms of speed and efficiency for video decoding and going back to OpenCL assistance using shaders for 4K H.264/H.265 is a very bad thing.
That's why VP5 and QuickSync fixed-function HW with 4K H.264 acceleration are useful for a consistent experience of fast and efficient video performance across the board - without dependence of the GPU performance (shaders).
BTW, AMD tried in the past to introduce to the world another OpenCL assisted video decoding, called OVD (OpenVideo Decode), which was a wrapper over DXVA, I think and used only UVD.
That project didn't have any success and they forced to eliminate it because nobody ever developed a decoder or player using OVD.
I'm not sure if I want the same thing happen again for this new OpenCL assisted video (which seems different because is using shaders and not only UVD), in order for AMD to leave the tricks and deliver a decent fixed-function HW for 4K H.264 and H.265 decoder.