Micro Mart Forum / Digital Photography & Video / Technical Forums / Video Players and HD Decoding / Latest Posts

RE: Video Players and HD Decoding

Slipstreem — Sat, 12 Jul 2008 14:12:28 GMT

[quote][b]FreakShow! (11/07/2008)[/b][hr]Not sure what the eventual filesize would have been though.[/quote]
Pretty huge I'd expect. SD content (720x576) in MPEG-2 requires 6Mbps to maintain commercial DVD quality. By simple extrapolation, HD content at 1920x1080 would probably require roughly five times this (~30Mbps) to achieve the same level of quality in MPEG-2.

I've been experimenting with transcoding an HD WMV file (1440x1080) to XviD MPEG-4 with [url=http://mediacoder.sourceforge.net/]MediaCoder[/url], and the results look almost passable to me personally from around 6Mbps upwards on a single-pass encoding. Most of the detail is still present although there are visible signs of solarising in places. I followed this with a two-pass encoding at the same bitrate and find it much harder to tell the transcoded file from the original on the hardware I have. Detail is preserved with much better accuracy than by a single-pass encoding and solarisation is almost eliminated completely. This would be an acceptable bitrate for me personally for home HD MPEG-4 archiving. A single-pass 90% quality-based encoding turns out a slightly higher level of clarity compared to two-pass encoding at 6Mbps but without the encoding time overhead of a two-pass encoding, so this would almost definitely be the one I'd go for.

Resultant filesizes from 1min 45secs test source file...

6Mbps single-pass = 73MB,
6Mbps two-pass = 75MB,
90% quality-based = 77MB (variable dependent upon source material).

As you can see, they all work out at roughly 2.5GB per hour, so most typical 90 to 100-minute movies would fit onto a single-layer DVD-R in two-pass mode. Quality-based mode is less predictable as the filesize relates to content complexity so varies quite considerably in size. Single- or two-pass modes offer a fixed and predictable size-per-minute in terms of final filesize with variable quality whereas quality-based mode offers a fixed level of quality with an unpredictable final filesize. If target filesize isn't critical then quality-based mode is always preferable to the fussy.

It's only turning in an encoding speed of around 12FPS even in single-pass mode though (two-pass is around 8FPS), despite apparently using both cores of my C2D almost fully, so this does make you realise that it's simply not possible to encode/transcode HD content quickly [i]and[/i] properly, especially if you expect to keep the filesize down to be comparable with SD DVD content.

Carrying the maths a stage further, to achieve a figure of 30x real-time encoding for HD content whilst maintaining the same level of quality that can be achieved using a CPU and the XviD codec in two-pass mode would require either two C2D cores running at around 250GHz or one core running at 0.5THz. It's not going to happen with AVIVO and a GPU running at less than 1/500th of that speed.

Having said that, I can believe the claim of 30x real-time with single-pass encoding at the native resolution required for smaller portable devices with the accompanying large drop in image quality that can be tolerated on a smaller screen at a normal viewing distance. A bitrate of as low as 300Kbps in single-pass XviD MPEG-4 can produce watchable results for a 2" screen with a native resolution of 320x240 according to my own personal experimentation. :)

Cheers, Slipstreem. :cool:

RE: Video Players and HD Decoding

FreakShow! — Fri, 11 Jul 2008 16:44:22 GMT

Media Player Classic runs HD content better as it can use more than one core.

VLC can just about run it, but it's limited to one core.

Encoding a MPEG2 version now. Was going to let it run overnight, but then it crashed for some reason :\

EDIT: Left it to run for an hour and a half. Came back, wiggled the mouse to bring the monitor up and then it locked :\

AVIVO has issues apparently. On the plus side I was able to play what it had converted with VLC and results aren't too bad. Not sure what the eventual filesize would have been though.

RE: Video Players and HD Decoding

Bruce R — Fri, 11 Jul 2008 07:52:02 GMT

Hi Freakshow! :)
Hi Slipstreem :)
Your exchanges in this thread are a bit 'over my head', but in case it helps, when Freakshow! first found that HDdemo *.mkv file, I found that the only player that I had that would touch it was Videolan's VLC.
In XP the Feb2008 version 0.8.6e stuttered and lost a lot of frames on my [C] rig and the Ubuntu 8.04 Linux version fared no better, but on my [A] and [B] rigs a different story.
On my [A] rig still stuttering under XP, less so under VHP, but LinuxMint4, no problem, quite smooth playback after a startup pause. Ditto on my [B] rig, where I have just quickly tried loading a LiveCD session of Mint5, going to its software portal and downloading VLC. Result not perfect, obviously dropping some frames, but quite watchable. After posting this 'reply' from my [C] rig Ubuntu8.04, I'll try out its offline XP + VLC 0.8.6e and add a PS to the post. ;)
Edit PS - using an offline copy of XP, with less overheads, produced a slight improvement, but still massive frame drop. Meanwhile VLC Player at [url]http://www.videolan.org/[/url] is now at version 0.8.6h for Windows, so give it a try ?
If you don't define its File Associations correctly, don't panic, just 'Install', ie 'Uninstall' it again and it should restore the prior situation. (However, for smooth HD playback, try a Linux distro with VLC ?)

RE: Video Players and HD Decoding

Slipstreem — Fri, 11 Jul 2008 05:49:47 GMT

[quote][b]FreakShow! (11/07/2008)[/b][hr]In Catalyst Control panel, all settings are left to the application to decide. On Adaptive anti-aliasing should the method be Super-sampling or multi-sampling?[/quote]
Use super-sampling for the ultimate in antialiasing quality that the card has to offer, use multi-sampling for a slightly less precise antialiase with less hit on framerates.

ATI Tray Tools calls the options "Quality" and "Performance", and having tried both at native display resolutions of 800x600 and 1366x768, I can't see enough difference to worry about personally. I've now plumped for "Performance" which equals multi-sampling as the X1950Pro slows down a little in some games with super-sampling at my new native res of 1366x768. You may see a very slight increase in jiggy-jaggies on fine detail with multi-sampling. See what suits your eyes best.

Bear in mind that the algorithms for super-sampling and multi-sampling are tweakable via the driver at the ATI end, so it's worth checking out both on every new driver revision as their relative effectiveness may change in future driver releases. It has done before and I'm sure that it will again as driver developments and enhancements are made for the HD4xxx series. :)

Cheers, Slipstreem. :cool:

RE: Video Players and HD Decoding

FreakShow! — Fri, 11 Jul 2008 03:10:18 GMT

True, and it works beautifully in that respect. Maxing out all (except the obvious one).

As you're the only one here, I might aswell hijack me own thread to something you know about.

In Catalyst Control panel, all settings are left to the application to decide. On Adaptive anti-aliasing should the method be Super-sampling or multi-sampling?

While looking through your links to see if that answer was somewhere, I saw that ATT had some Mpeg2 optimisations. I'll have to wait till ATT is more friendly to 4 series cards, but hopefully, it may do something for this very thread.

EDIT: Forgot to say. I'll try a couple of different formats to see if it has a problem with DIVX, or if just generally it's not worth it's weight in Mb.

RE: Video Players and HD Decoding

Slipstreem — Thu, 10 Jul 2008 20:14:04 GMT

Maybe it just doesn't figure highly enough on the list of development priorities. I'd be the first to moan if development money was siphoned away from the graphics card's primary purpose and function to achieve something that I can do perfectly well without ATI's help. Bear in mind that none of the MPEG-4 hardware encoders offered by the capture card manufacturers work worth a spit either in terms of quality versus speed when compared to a CPU and software solution, not until you start spending silly money anyway.

Maybe AVIVO encoding will be sorted out at some stage in the near future, but I'd rather have them working on the actual graphics driver personally. It is a graphics card after all. :)

Cheers, Slipstreem. :cool:

RE: Video Players and HD Decoding

FreakShow! — Thu, 10 Jul 2008 20:08:05 GMT

CPU occupancy is at maximum. It's the same with GPU folding aswell.

I figured that it's the CPU having to feed data and just not being quick enough. If what you read is correct, then it's a pity that such lies/mis-information has been spread.

RE: Video Players and HD Decoding

Slipstreem — Thu, 10 Jul 2008 16:08:06 GMT

Something I read on another forum yesterday (can't remember where now) led me to believe that most of the work is actually done by the CPU during encoding sessions and that the reason for poor quality results was down to poorly written software. Have you checked the CPU occupancy when encoding? I know that AVIVO is capable of hardware [b]de[/b]coding. I'm just a little suspicious of whether it really uses the GPU during [b]en[/b]coding. :)

Cheers, Slipstreem. :cool:

RE: Video Players and HD Decoding

FreakShow! — Thu, 10 Jul 2008 15:17:46 GMT

Well, as if to re-iterate how poor ATI are handling this atm, I re-encoded a Hi-Def film, which was 8Gb (or something).

The result was an MPEG4 (DIVX) file which was 1.56Gb. Quality was awful, very blocky.

While I'd like to use it, you are correct in that quality rules over speed of the operation.

The conversion was using the highest quality setting, so I'd dread to think what the lowest is!

RE: Video Players and HD Decoding

Slipstreem — Thu, 10 Jul 2008 06:45:38 GMT

[quote][b]FreakShow! (10/07/2008)[/b][hr]Results were dissapointing for the filesize. While the quality is better on CPU, the speed is there on the GPU. And if the GPU is able to render a game at high resolutions at near instant speeds 30 times + a second, I'd have thought allowing it to convert videos to be an oK thing.[/quote]
If you can put up with the large apparent drop in encoding quality, yes. I wouldn't use any hardware encoder I can afford for full-size, ie, non-rescaled encodes personally. Maybe for shrunk-down portable files. The improved quality that a software encoder can give offers you the flexibility of encoding properly over a longer period or using faster options and creating a file to match the hardware encoding in terms of image quality but with a potentially much smaller filesize.

I'm not sure how much of the 3D rendering hardware is actually used in video encoding mode, so it's debatable as to whether it's logical to expect it to be quick solely on the grounds of also being a graphics card. Unless the maths required to perform the encoding runs in dedicated hardware or in a language that suits the existing graphics hardware, there's no reason to assume that it can perform video encoding quickly [i]and[/i] well.

I realise that encoding on a CPU appears to be slower, but I can almost double the speed that Vidomi encodes XviD MPEG-4 at (climbs from 95FPS to 182FPS for 16:9 SD content) just by changing two settings, and the output file is still very pleasant to watch on a 32" TV. I can provide FPS figures for different quality settings and rescales if it's helpful.

I'd be very interested to know how quickly AVIVO can encode the clip in the MPEG-4 Encoding Race thread with no rescaling at various quality settings. It would give you and us a handy reference to compare the speed to our database of PCs performing software encoding if you can provide the relevant FPS figure. :)

Cheers, Slipstreem. :cool:

RE: Video Players and HD Decoding

FreakShow! — Thu, 10 Jul 2008 04:08:38 GMT

Well, I got AVIVO working (just set it and let it work. I'm just impatient).

Results were dissapointing for the filesize. While the quality is better on CPU, the speed is there on the GPU. And if the GPU is able to render a game at high resolutions at near instant speeds 30 times + a second, I'd have thought allowing it to convert videos to be an oK thing.

Converting an HD video on my CPU was going to take 6 hours. ATI claim they can do the same conversion in 32 minutes. I'll wait for updated drivers, or OMEGA drivers, but after that, if it still doesn't work, I may drop an email to them to see if there's something I'm missing.

The Video acceleration is definitely what I am looking for as this can be applied to the media centre PC.

Again, I'll see what happens when updated drivers come along.

RE: Video Players and HD Decoding

Slipstreem — Wed, 09 Jul 2008 08:54:35 GMT

See my reply in your other thread. I think there [i]may[/i] be a fix. ;)

The problem seems to be that AVIVO is actually a very old standard and only ever produced relatively mediocre encodings at best when compared to what a good software CODEC can achieve. Maybe development hasn't been kept up with since the X1xxx series because it's a lost cause when compared to what can be achieved with a dual-core CPU (especially a C2D) and decent software.

I wish you the best of luck if you try unleashing this power with the Omega driver, but don't expect perfection. It's going to be potentially buggy down to a lack of serious development recently. Playback acceleration may work acceptably well though if you can enable it. :)

Cheers, Slipstreem. :cool:

RE: Video Players and HD Decoding

FreakShow! — Wed, 09 Jul 2008 00:44:05 GMT

While my processor is able to do the work, the reason for my testing is that I am hoping to get a media centre PC HD ready.

It's not the most powerful (Celeron 2.66GHz), I'm hoping that adding a relatively low powered graphics card with said HD chip will help it along.

CoreAVC claims to be bringing GPU decoding onboard, but that doesn't seem to have actually happened yet, so no idea when it can be expected.

RE: Video Players and HD Decoding

Slipstreem — Tue, 08 Jul 2008 23:46:48 GMT

[quote][b]FreakShow! (08/07/2008)[/b][hr]MPC-HomeCinema uses two cores to decode it all so that's how it plays smooth playback (I think. It uses more than 50% of CPU).[/quote]
That would explain where it gets its horsepower from then. I did wonder.

[quote]Still not HD Decode chip. Pity there aren't more programs that do it :([/quote]
ffdshow itself has the ability to decode VC-1, if memory serves. It may use the same routine as MPC-HomeCinema to do it though. ffdshow is only a collection of custom video filter DSPs and such wrapped inside a GUI environment for easy settings adjustments. Many parts of ffdshow have been stand-alone filters in their own rights at one time or another. They pop up all over the place in modified forms. :)

Cheers, Slipstreem. :cool:

RE: Video Players and HD Decoding

FreakShow! — Tue, 08 Jul 2008 23:37:21 GMT

MPC-HomeCinema uses two cores to decode it all so that's how it plays smooth playback (I think. It uses more than 50% of CPU).

Still not HD Decode chip. Pity there aren't more programs that do it :(

RE: Video Players and HD Decoding

Slipstreem — Tue, 08 Jul 2008 23:22:11 GMT

That's a question I'd like answering too. AVIVO seems to have been around for ages now, but I never see any mention of whether you can actually use it yourself or not.

Did you ever try MPC-HomeCinema? It plays back HD WMV files and is claimed to support VC-1. I have no idea how it implements VC-1 support though. It may be hardware optimised. You never know. :)

Cheers, Slipstreem. :cool:

Video Players and HD Decoding

FreakShow! — Tue, 08 Jul 2008 23:06:57 GMT

Hi all,

So armed with an ATI 4870 I'm trying all sorts of things out. I've tried the games, I've tried the GPU folding and now I'm looking at HD video.

There's not much properly encoded HD video about. I've found one [url=http://www.techpowerup.com/downloads/530/HD-DVD_Demo_1080p_VC-1_DDPlus_5.1.html]here[/url]

Now, I know that PowerDVD and Intervideo both support the HD decoding chips from Nvidia and ATI, however, do they support using the chip from files, or is it only from Blu-Ray drives?

Are there any video players out there that support the use of HD decoding chips from file?

I've had a good search, but can't seem to find much.
TIA :)