Crunchie Compression Speeds

[eisaiasjr]

I run fairly modest setup on my MCE machine and a very nice setup on my "gaming rig" and I have noticed that Crunchie takes forever to complete!

I wanted to post some of the times I have seen my machines take, and compare them to the times you guys may have seen...

MCE Machine: 6 hours for every 1 hour of video (running MCE on the background with about 70% free CPU)
1.8 ghz intel (non-HT) @ 2.4ghz (100% stable)
1 gig of ram
60 gig 7,2k rpm HDD

Gaming Rig: 2 hours for every 1 hour of video
FX-55 @ 2.7 ghz (100% Stable)
2 gigs of ram
36 gig raptor 10k rpm HDD

Everything gets stored (while it gets compressed) on my FileServer which is a 1.8ghz @ 3.0 machine with 8 HDDs of 250 gigs on Raid 5 formation.

I would love to hear some of your times of compression... post away!

Regards,

E

[Mike]

I haven't checked times for a while now, but the last time I did it was taking +/- 10 minutes more than the recorded length of the show.

My files are crunched to a second drive on the same machine and then manually transferred to my file server.

(machine specs below)

[bella]

Hi E.,

I get about the same time results as Mike with converting DVR-MS files. My MCE runs on a 3.4HT processor. CPU load around 50%.

My second cheapskate machine runs on a 2.66ghz Celeron. It takes more then 2 hours to do a movie conversion at 100% CPU load.

If it really takes you that long you can do the following to speed up the process:

Reduce the bitrate to 1000 and see for yourself if you still like the result. I have tried it often at this bitrate and I cannot see the difference between the high bitrate and the lower. I also skip the double interlace feature, although I do not now if that makes it run faster.

[arkay]

The bitrate has no bearing on the encoding time. The double rate deinterlace should only adds a minimal amount to the encode time. Though on a slower machine the lengthening will be more dramatic. Double rate playback also requires more CPU as it's playing back at 50fps instead of 25. Either way nothing in the settings will really affect the time it takes..

Unfortunately it takes as long as it takes...

Cheers,
Arkay.

[bella]

This is strange Arkay. I am sure you are right but I converted the same movie twice, once with a 1000 bitrate and no interlacing, and once with 5000 bitrate and double deinterlacing.
The difference in time was ca. 20 minutes. And I noticed the same thing with other conversions. Could it be something else that plays a role? When I run chrunchie I never have anything else going in the background, so that avenue is closed.

[natrat]

Originally Posted by bella

Hi E.,

I get about the same time results as Mike with converting DVR-MS files. My MCE runs on a 3.4HT processor. CPU load around 50%.

Likewise. I'm considering turning HT off in BIOSS for some testing so mencoder then uses 100% of the CPU, at least until it's a threaded app. When I ran online game servers for a living (1u dual xeon boxes) we did a lot of testing with HT on and off and the overall difference, for game server performance, was minimal. Possibly similar for what I use the MCE box for.

nat

[arkay]

The bitrate doesn't change it. The double rate deinterlace can. As it technically has to do a lot more work to convert a 25fps clip to 50fps progressive. i.e. It is generating twice the amount of input frames to encode so yes, it does lengthen the time. I didn't think it would be to that extent though I confess I've never really timed it.

I have a differing view on all this "it's not fast enough" stuff anyway. I've always been of the opinion that's it's better for it to take twice as long for excellence in quality vs half the time for poor quality. Given that you only ever encode it once and for me personally I always want it to look it's best when I watch it.

However, if your drive is full and your CPU isn't crunching fast enough I can understand that people want it to be quicker. I'll look into speed issues at some point though it will only be when a newer version of mencoder is released that I will be able to do anything about it.

mpeg4 compression is very CPU intensive and will always be slower than we'd like

Cheers,
Arkay.

[TiggerK]

Yes, at the end of the day, it's the CPU that makes the difference, and not a lot else.

Here are some interesting CPU comparison specs from tomshardware.com, encoding a small 182MB VOB to divx test. Tested with appropriate good quality RAM/chipsets for that era of CPU hardware. Time in minutes.

Pentium 233 MMX 113.02 (included just for a laugh)
Pentium 3 600MHz 11.33
Pentium 3 1GHz 7.28
Pentium 4 1.8GHz 4.37
Athlon XP 1800+ 4.26
Athlon XP 2600+ 2.49
Pentium 4 2.8 Northwood 2.47
Sempron 3100+ 2.46
Pentium 4 630 (3GHz) 2.03
Athlon FX-55 1.51
Pentium 4 660 (3.6GHz) 1.45

So a P4 1.8 takes more than twice as long as a P4 3GHz... Thought this was interesting anyway.

Cheers
TiggerK

[gussey]

Hmm i find that last chart very interesting...when i upgraded from a AthlonXP 1700+ to a AthlonXP 3200+ I didnt notice a hell of a lot of difference when encoding VOBs, was probably limited by the speed of the optical drive i was ripping off!