I am trying that crawford.mpg file. It doesn't seem to have sound, is that correct?

mplayer says that my system is too slow to play the file.

top shows that my system is at about 96% CPU no matter what command-line options I give to mplayer. I've tried...
-vo xv
-vf pp=lb
-framedrop
-cache ####

Nothing seems to help the CPU. I'm surprised that I see no change when I do -vo xv. I'd assume it would help a bit. Should I assume that?

No, 'emotion' and 'logo' have no sound, but the others do.

Not necessarily. You could try xvmc instead of vo, but the video decoder lib mplayer uses with xvmc doesn't work very well.

It may be time to consider transcoding before playback.

_________________
KnoppMyth R5.5
MythiC Dragon v2.0
Join the KnoppMyth Frappr!

I'll mess around some more. I may go the transcoding route, but I don't understand why some guys can do it on less than 2GHz and I have 2.6GHz and it won't work. It just really feels like I'm missing something that could speed it up.

Well your system is hindered by pci video, slow pc2100 memory, and a celeron processor which is not on par with a P4 or Athlon 64. It is not as simple as just a megahertz to megahertz comparison. Good hardware makes all the difference...

There are 2GHz processors that can get more work done per second than your 2.6GHz processor, though. The CPU I use in Dragon is a 2GHz processor, but it's a 64-bit processor with newer and better architecture than a Celeron.

There are people with VIA boards that use the onboard MPEG-2 decoders with a special version of XvMC who can get results comparable to yours on HDTV content, and those VIA boards have CPUs that top out at about 1.5GHz at the moment. Their results are probably more due to the fact that the VIA boards have specialized MPEG-2 decoder hardware and less to do with the CPU speed.

_________________
KnoppMyth R5.5
MythiC Dragon v2.0
Join the KnoppMyth Frappr!

Well, when I was looking around, I saw a guy that claimed to be doing it with an Athlon (or maybe it was Duron) 1.8GHz. I understand that the Celeron isn't as nice of a CPU and that my RAM is slow. I get all that.

If the problem was a PCI, bad HD, etc. bottleneck, that would cause the CPU to be utilized less, right? Would the same go for slow RAM, or does that stall the CPU so it still shows up as being in use under 'top'?

I can transplant my stuff into another PC. I can easily try a 3GHz P4 or some other machine around my office.

I'm mostly surprised that the difference between "Standard" and "Standard XvMC" isn't noticable and that the deinterlacing is such a hit on the CPU. I would think those things would offload more CPU strain because they'd use my FX 5200.

My personal feeling on the matter is if you have a fast enough machine then you would you want to mess with XVMC at all because your cpu can handle the work. I tested my Athlon 64 2.5ghz system with PCI 5200 vs onboard NVdia 6150 graphics. 6150 onboard graphics perform better with R5D1. I also have a slot for PCI Express card but right now I don't have one so I can't compare. I would guess that a 3ghz P4 would be good without XVMC.

If you want to stick with your hardware then try test as the frontend or try to recompile with less software than in the knoppmyth build. This would provide some insight for us all....

That's a very good question, and I haven't done the scientific testing to know what happens to reported CPU load during the various kinds of bottlenecks that can occur. There is a thing called a wait state that the CPU can be in, where it is holding off doing anything else because it's waiting for some information to come in. 'Top' categorizes that separately from user, system, niced, and idle usage.


That's a good way to narrow down the exact cause. If all you do is swap the motherboard, CPU, and RAM, then you can test if your graphics card or HDD is bottlenecking. If it runs smoothly like that, then it's either CPU or RAM.

Not all nVidia cards are equal. Although it can vary from release to release, I generally do see a drop in CPU usage when using XvMC on Dragon. If you're not seeing a similar drop in usage, it could be that your FX 5200 is no better at helping decode MPEG-2 than your CPU is, for various reasons.

_________________
KnoppMyth R5.5
MythiC Dragon v2.0
Join the KnoppMyth Frappr!

So, I found this interesting.

I swapped my HD and cards to another PC to test. I have yet to get MythTV up and running, but the first thing I noticed was the LILO booting.

The little dots on the LILO line (probablly when uncompressing the kernel) fly by on the 3GHz P4. On my 2.6 Celeron, they are rather slow.

I know that my motherboard is pretty junky. Perhaps there's something going on there. At least, in theory, I could replace my motherboard for much cheaper than buying a new CPU. Though, I'm still not ready to spend right now.

I'll keep tinkering.

Well, I transplanted the following items to a new PC.

Video Card
HDTV Tuner Card
Celeron CPU
512 PC2100 RAM Stick
Hard Drive

Essentially, this new PC is only a new motherboard. Today, my junky antenna isn't pulling in HDTV as well as before, but when it does, it seems to only use about 75% of the CPU. Also, the LILO dots go a lot faster.

I tried "xine -V xvmc" to play craford.mpg and tst.ts. Xine rarely goes over 50% cpu on them.

This all strikes me as rather odd. I don't get what would be so slow on my motherboard. The BIOS on it sucks, though, so I can barely configure things.

It's somewhat interesting. I hadn't messed much with Xine before.

I put all my stuff back in my original PC. I messed with Xine on it.

If I run Xine on crawford.mpg or tst.ts (http://www.pchdtv.com/downloads/tst.tar) it works fairly well on my junky motherboard.

"xine -V xvmc -D crawford.mpg" On my slow motherboard, that will only use 30-40% of the CPU. It might spike up to 60.

If I remove the "-V xvmc" option, it'll run around 70-80% of the CPU.

I can not get mplayer to work with the "-vo xvmc" option. If I specify ffmpeg12mc it gets errrors.

We already know about MythTV's performance. Why is Xine so much nicer on my CPU?

Could be that xine and mplayer are using different MPEG decoder libraries by default.