2016-02-29, 16:43
Just had an update that my upcoming PiTop Ceed is now going to be coming with a Pi3 rather than a Pi2.
Me a happy bunny about that
Me a happy bunny about that
(2016-02-29, 15:35)noggin Wrote:(2016-02-29, 12:51)jurm Wrote: Does this new pi3 play x265 yet?
It plays it the same way that previous Pis do. Existing Pis have GPU compute accelerated (but not VPU hardware decoded) HEVC decoding I believe - so it's GPU+CPU rather than VPU (which is what does MPEG2, VC-1 and H264 on the Pis). The Pi 2 B is doing HEVC at 720p and modest bitrate 1080p I believe. The New Pi 3 may well increase the bitrate ceiling for HEVC 1080p content with the faster CPU and GPU - and may also benefit from some optimisations down the line?
Also the new Pi 3 has default overclock (so probably not considered overclock?) for the VPU - so can now reliably decode H264 1080/60p apparently.
(2016-02-29, 17:56)TheShoe Wrote: had been hoping they would update the video chip as the current Pi2 - while a great value - exhibits too much color banding in video playback. done several A/B tests and the pi2 seems to render @ 16bit color, or something else is causing the banding.
(2016-02-29, 16:24)natethomas Wrote: Here's a quick video I just made showing the Pi3 running OpenELEC. This build is from a week or two ago though, so I hadn't gotten Bluetooth up and running yet.
(2016-02-29, 18:36)Glockyyy Wrote: I can see quite a lot of combing in your 1080i clips which is not present in your 1080p clip. Not sure if it's in the encoding of your Youtube video but seems to me a false film deinterlacing error. There was lots of mixed graphics in the material which could have caused it, but either way it was far from perfect.
(2016-02-29, 19:57)ZwartePiet Wrote:(2016-02-29, 18:36)Glockyyy Wrote: I can see quite a lot of combing in your 1080i clips which is not present in your 1080p clip. Not sure if it's in the encoding of your Youtube video but seems to me a false film deinterlacing error. There was lots of mixed graphics in the material which could have caused it, but either way it was far from perfect.
I think he was referring to there being no dropped frames or stuttering. It's unlikely that the Pi 3B has enough processing power to software decode 1080i video and de-interlace the output. You'll need an MPEG-2 license for that.
The Pi 2B can already de-interlace (YADIF 2x) 1080i MPEG-2 video with the license. IMO, the more interesting metric will be how far they can push h.265 software decoding.
(2016-02-29, 21:04)noggin Wrote:(2016-02-29, 19:57)ZwartePiet Wrote:(2016-02-29, 18:36)Glockyyy Wrote: I can see quite a lot of combing in your 1080i clips which is not present in your 1080p clip. Not sure if it's in the encoding of your Youtube video but seems to me a false film deinterlacing error. There was lots of mixed graphics in the material which could have caused it, but either way it was far from perfect.
I think he was referring to there being no dropped frames or stuttering. It's unlikely that the Pi 3B has enough processing power to software decode 1080i video and de-interlace the output. You'll need an MPEG-2 license for that.
The Pi 2B can already de-interlace (YADIF 2x) 1080i MPEG-2 video with the license. IMO, the more interesting metric will be how far they can push h.265 software decoding.
Yep - think there was a little confusion between CBS being a 1080i station (so always broadcasting an interlaced signal) and the content within that signal being 1080i native. When Nate Thomas was talking about 'interlaced content' the material looked like a commercial, the vast majority of which will be native progressive 'film-'mode (shot 24p or occasionally 30p - though probably not actually on film), which won't have the same issues as native 60i interlaced 'video'.
(2016-03-01, 11:13)Glockyyy Wrote:(2016-02-29, 21:04)noggin Wrote:(2016-02-29, 19:57)ZwartePiet Wrote: I think he was referring to there being no dropped frames or stuttering. It's unlikely that the Pi 3B has enough processing power to software decode 1080i video and de-interlace the output. You'll need an MPEG-2 license for that.
The Pi 2B can already de-interlace (YADIF 2x) 1080i MPEG-2 video with the license. IMO, the more interesting metric will be how far they can push h.265 software decoding.
Yep - think there was a little confusion between CBS being a 1080i station (so always broadcasting an interlaced signal) and the content within that signal being 1080i native. When Nate Thomas was talking about 'interlaced content' the material looked like a commercial, the vast majority of which will be native progressive 'film-'mode (shot 24p or occasionally 30p - though probably not actually on film), which won't have the same issues as native 60i interlaced 'video'.
Film material interlaced to 1080i should have been easier to deinterlace since it would have involved no interpolation of the material. The error was combing which was perhaps caused by the graphics on-screen confusing the deinterlacer into thinking it could straight weave when in fact it needed to interpolate (i.e. it needed to treat it as video not as film).Either way, net result is there were errors it was not handling it perfectly at all. Decoding license required. Doesn't take anything away from the new Pi, I know I have mine on order already!