PDA

View Full Version : Changing iDCT during the encode will affect the video?


jamewoong
1st September 2007, 18:42
I can't determine what iDCT is better for my PC which is a PIV 2.0 Ghz. I thought the more RAM I give to it, the fastest the job will be done, but I don't think this is right (CPU speed affect the encode).

I encode a movie with 512 MB of RAM and it could take up to 868 mins (used 32 Bit SSE2MMX) with no background application running. After upgrading to 2GBs of RAM. The encoding time still the same.

I want to test to see what is the preference for my PC, but can't find anything like this...

So, just wondering if changing the iDCT during the encode will cause problem to the video that is curently encoding? By this way, I can probably determine the best setting for my PC.

Thank you.

jdobbs
1st September 2007, 19:17
It typically won't make a visible difference -- but it can make a considerable difference in speed.

I've found that memory isn't a big factor in encoding speed unless you are using the PC for other purposes while encoding or the memory is so low that it starts causing virtual memory swapping. The biggest factor by far is the processor and processor speed. Second would be the hard drive -- you'll see significant slowdown if you are using a USB connected drive instead of ATA or SATA. Most import -- keep the drive defragged. That is one of the biggest reasons you can see slowdowns. One other note: It's a lot faster to read the disc to to hard drive and run against it than encoding straight from the disc (using AnyDVD, for example).

I have to say that 800+ minutes is way long. Are you doing numerous passes? I have a computer I use for testing with similar specs (an Athlon 2400), and it usually only takes about 3 hours (and that includes the burn time).

jamewoong
1st September 2007, 22:23
I have to say that 800+ minutes is way long. Are you doing numerous passes? I have a computer I use for testing with similar specs (an Athlon 2400), and it usually only takes about 3 hours (and that includes the burn time).
Honestly, I'm really jealous that people take less than 8 hours to encode a DVD. I have encoded about 50 DVDs and the fastest speed I can get is 8 hours (528 mins) via PIV 2.16 Ghz (1 GB of RAM) that has MMX, SSE, SSE2, SSE3 (FSB Rated as 531.8 MHz) - no background application is running during the encode). Also, the Ultra DMA is 5.

When I check the speed in the Procoder Window, the speed was 0.62 X Realtime. But don't know why, sometime, the speed drop to 0.32 X Realtime. When the speed drops to 0.32 X Realtime (by random case), it will take about 12 hours for an encoding.

Does it have anything to do with the encoder - Procoder (Mastering Quality)? Or the % of the compression? Normally, it's around 45-60%.

Or DVD Rebuilder with CCE is a couple? That's why it take less time?

Between, thanks for the tips.

jdobbs
1st September 2007, 23:33
ProCoder is very definitely very slow compared to other encoders. Have you tried HC encoder? It is faster than ProCoder, gives excellent quality, and it comes with DVD-RB.

feedback
1st September 2007, 23:47
Try using "32 Bit SSEMMX(Skal)" iDCT, it the fastest iDCT for me.

Also, make sure your source is ripped to a separate drive. So you need 2 hard drives to do this method.

Procoder (Mastering Quality) is the slowest of it's choices.

Also, More than 2 passes in CCE will take more time, of course.

jamewoong
2nd September 2007, 02:48
ProCoder is very definitely very slow compared to other encoders. Have you tried HC encoder? It is faster than ProCoder, gives excellent quality, and it comes with DVD-RB.
Yes, as you said, it is probably the slowest, but it's the easiest for setting: just have to choose "Mastering Quality" and that's it. With others encoders such HC, I need to set the bias or something like that. It's not easy to determine the number to unput...

Well, will try it next time.
Try using "32 Bit SSEMMX(Skal)" iDCT, it the fastest iDCT for me.

Also, make sure your source is ripped to a separate drive. So you need 2 hard drives to do this method.

Procoder (Mastering Quality) is the slowest of it's choices.

Also, More than 2 passes in CCE will take more time, of course.
Fastest to you doesn't mean that it will be the same for me.

What do you mean by "separate drive"? When I encode a DVD, the source is always located on the Primary Master Drive which is C: and the output is in the same drive.

Correct me if I'm wrong, but when moving a file from a different drive such C: to D:, it will take time (copying), so your method of using 2 different HD will slowing down, no?

Some people said that 32 bit simple MMX (XVid) is the best and precise option. Other people choose the Skal one. Fewer one choose the SSE2MMX. Looks like there's no end with this...

Is there anyone here that got a PC that has a lower system specs than mine? If you have a system lower than mine and got a faster speed, then something could be wrong with my PC.

For others, probably your system is similar to mine, but it could have a hidden feature to enhance the multimedia or duo core system?

That's why it is faster...

feedback
2nd September 2007, 07:11
Fishman0919 did the following test which is located HERE (http://forum.doom9.org/showthread.php?t=105932&highlight=iDCT+speed)>IDCT SSIM Time
--------------------------------------------
1 32-bit MMX 99.63% 2:15
2 32-bit SSEMMX 99.63% 2:14
3 32-bit SSE2MMX 99.63% 2:14
4 64-bit Floating Point 99.88% 2:41
5 64-bit IEEE-1180 Reference 99.93% 2:52
6 32-bit SSEMMX (Skal) 99.67% 2:11
7 32-bit Simple MMX (XviD) 99.74% 2:12


All idct were compared with a simple script and DGIndex 1.40 rc5 from the movie "The Pacifier"
(I did 4 diff movies but all gave the same results)
and was the first 7000 frames... time is how fast each idct was
encoded with CCE Basic with an AMD 64 3700+ OC (about 4000+)....

LoadPlugin("C:\Test\DGDecode\DGDecode.dll")
mpeg2source("C:\Test\test.D2V",idct=?)
ConvertToYUY2()

and compared dirtectly to the Original D2V file

Source and output need to be on 2 different drives.
For instance, I rip the source from (D.)DVD-Rom drive, to my internal HD (C.)then have Rebuilder output on my external (E.)hard drive.

You will see marked increase in speed by using 2 Hard drives (I am using a laptop).

jamewoong
2nd September 2007, 08:28
Source and output need to be on 2 different drives.
For instance, I rip the source from (D.)DVD-Rom drive, to my internal HD (C.)then have Rebuilder output on my external (E.)hard drive.

You will see marked increase in speed by using 2 Hard drives (I am using a laptop).
Can anyone explain me why it will boost the speed if we use 2 HD?

Try to move a DVD content from your "C" to your D, E, F.. It will take more time as it will copy the whole content to your other drive. So, how come when using the output to other drive, it will boost the speed? I still don't get it...

Well, will try this too for the next time! Thanks!

EDIT: Yup, I think the test of Fishman0919 was right. When using the SKAL iDCT, it's 0.32 X RealTime when I'm using the PC at the same time. If I choose another one like 32 Bit MMX, MMX SSE/MMX SSE2, it will go down to 0.18 X Realtime.

linx05
2nd September 2007, 09:36
jame: have a read of this thread (http://forum.digital-digest.com/showthread.php?t=79471) to learn a bit more about that question.

jamewoong
2nd September 2007, 21:47
jame: have a read of this thread (http://forum.digital-digest.com/showthread.php?t=79471) to learn a bit more about that question.
Thank for pointing me at that thread.

Between, does it have the same effect if the 2nd HD (output) is on the network or via a partition that I splitted?

jdobbs
2nd September 2007, 22:48
You'd be well advised to avoid using a network drive. There you are probably talking about 100Mbs.. that's is very, very, very slow in disk read/write terms. Even a USB-2 drive will slow it down, and it can trasfer at up to 480Mbs.

A modern SATA drive can transfer data in the 3Gbs range. That's 30 times faster than your network.

You're better off with a single drive than with second network drive.

jamewoong
21st September 2007, 04:58
I just notice something really weird...

Took me 12 hours to encode a movie using this PC (2GHZ - SSE, SSE2 - 2 GB RAM) comparing to my laptop (toshiba satellite 1.6 Ghz 1GB RAM (SSE, SSE2, SSE3)) which is 8 hours.

But when I mux something, this PC is faster (took 2-4 mins to mux) than the laptop one (took 10 mins to mux).

How come it is?

Any clue?

neuron2
21st September 2007, 05:33
Muxing is disk I/O bound. Encoding is CPU bound.

jeffy
23rd September 2007, 20:24
Muxing is disk I/O bound. Encoding is CPU bound.
So simply said:
CON:Your laptop drive is slower than your desktop drive.
PRO: Your laptop CPU is more powerful than your desktop CPU.