Time to level-set. What's your fastest CCE speed and any new ways to encode?

[Roveer]

It's been a full 3 months since I did my last CCE encode. Geez, like AA. Since then DOCCE4U has come to be and I tried it last night. Seemed pretty good. Makes things nice and easy.

Here's my questions. (running CCE 2.66 on a Dual Xeon 2.8ghz)

What's the fastest CCE speed that anyone is seeing? I'm getting 3.0 on VAF creation and 2.9 on actual encode. Anyone doing better? If so, whith what? Opterons should cook.

Also, anything new in the way we encode? I saw some stuff on a new AVISYNTH that did a new colorspace or something like that. I remember that it was the conversion that was holding up CCE. Anything there change that might give us better CCE encodes?

I'd do all the reading myself, but I figured I'd ask and if there is a lot that has changed, I'd spend the time to dig in. I dont' have a lot of time to devote so I have to be smart and ask dumb questions.

I got caught up in Pinnacle Instant Copy which looked really good, but between the machine lockups and the few movies I encoded and had the movie jump all over the place I have now abandoned it. Let me know if any of the newer versions are better. When it would run, it ran pretty fast on my Dual Xeon.

Thanks and look forward to the comments. Hope the replies will help others too.

Roveer

[RB]

Your speeds look pretty good. I'm getting about 1.6 on my single 2.4 GHz P4 (no HT).

About the color space conversion: this is in the FAQ. Basically, with AVISynth 2.5, we could get a noticable increase in speed if only CCE would natively decode YV12. But no version of CCE currently does and Custom Technology has "currently no plan for that".

[redeemer-dk]

Nice speed. I'm getting around 1.3 on my Athlon XP 1800+ with a script using only MPEG2Source() and YV12toYUY2().

[FilipeAmadeuO]

How much RAM menory do you have ??
I have a AMD 2000+ with 256Mb Ram and i get a 1.20 speed in CCE

[Roveer]

My machine has 1gb of RAM. I've seen it report in the 300mb range for memory consumption so the more the better.

Roveer

[ADLANCAS]

I have a AMD Athlon 1600+ (256DDR) and normally CPU usage is 100%when I encode CCE.
Is that normal? Maybe better with 2200+ or 512DDR or what?

I've made another test with a single script:
Avisource(c:\test.avi")
BicubicResize(480,480)

Speed in CCE was 1,06.
Somebody knows if is it normal?

Thanks for any comment,

Alexandre

[redeemer-dk]

I have 256MB of DDR RAM. Recently removed one of my memory sticks since it was bad. I don't think it makes a big difference though, since (according to taskmanager) the peak memory usage is about 175MB. It actually went up speed 1.35 yesterday. But this speed also depends on the complexity of the MPEG2 stream you're trying to decode. Some can be decoded faster than others so you can't compare it with different sources. Mine is also PAL, and the speed in FPS is:

FPS*Speed

So if your source is NTSC 29.97 FPS then your speed is slower even though you might be decoding a larger number of frames per second.

-Rune Svendsen

[coona]

I have a AMD 2000+ with 256Mb Ram and i get a 1.20 speed in CCE

I have the same CPU and amount of RAM and usually get 1.35-1.45 in CCE. I´m PAL user like FilipeAmadeuO (I guess). I use W2000 SP3 and avisynth 2.08. Maybe some bios settings difference ;).

[FilipeAmadeuO]

I use Windows XP professional, Avisynth 2.52 and CCE 2.67.00.10
The speed depends on the source, but the top speed that i ever get was 1.30. And normaly it´s 1.10-1.20.
What Avisynth decoder do you use ??
Is it any speed increase changing the decoder ???
What is the CCE version that you use ??
Do you use ConvertYUV2() ???

[coona]

I tried to switch to new version of Avisynth when it was firstly realesed but there were some errors (which may be currently repaired) so I returned back to previous version. I want to try it again in near future but still have no time to do it :(.

I use Avisynth 2.08 and CCE 2.5.01.00. (I´m at work and I´m not sure about decoder). I don´t use ConvertYUV2().

What motherboard and chipset do you have??

[Zot]

Athlon dual MP 1800
1GB RAM
Windows 2000
CCE 2.50

Average speed using DVD2SVCD front end is 1.5-1.8 unless its a hybrid/needs de-interlace. Then it drops down to about .9

Zot

[tao52]

When not using DoCCE4U and just doing the reencoding myself I get 1.8-1.9 on my P4 2.53 with 512 MB PC1066 Rambus. However when I use DoCCE4U I get 2.2-2.3, and this is with CCE 2.50. I dont know what accounts for the speed difference but it is there, and the only change was using DoCCE4U.

[SVCD4Me]

CCE speeds of 2.22 on XP 2500+ @ 2136 MHz, 256 MB PC2700 Dual channel. FSB and RAM running @ 170 MHz. This is using DoCCE4U with the following script:

import("C:\Program Files\DoItFast4U\new.avs\addaudio.avs")

LoadPlugin("C:\Program Files\DoItFast4U\mpeg2dec3.dll")

Mpeg2Source("VTS__04_P01.16~9_1.d2v")



AddAudio()

ConvertToYUY2()

EDIT: The speed varies now from 2.2 to 2.8 or so. Usually settles in around 2.5. It seems to depend on the original/new bitrate.

[asifanwar]

Hi,

System:
Abit IS7-E mobo (Intel 865 Springdale with 800MHz bus)
P4 2.4 o/c to 3.3GHz
2 x 256 Corsair DDR
WinXP SP1

DVD2AVI 1.76
AVIsynth 2.08
CCE 2.66.01.07

Encode Speed: 3.54x

I used to have a 2.4 P4 o/c to 2.62 with a Gigabyte Intel 850 mobo and RAMBUS and was running at 2.8x encode speed. Much happier now of course!

[shoarthing]

Very interesting & useful [am dithering between dual Bartons/Dual Xeon HT's to replace the current dual Palomino MP's at 1533MHz w/ 512MB DDR]
- the MPEG2 streams dealt with are mainly interlaced PAL; using more-or-less 'standard' DoItFast4U *.avs scripts for FieldDeinterlacing in AVISynth then encoding as progressive, speed with MPEG2 streams of 'normal' complexity & target bitrates is just about exactly 1.3:

import("C:\Program Files\DoItFast4U\new.avs\addaudio.avs")
LoadPlugin("C:\Program Files\DoItFast4U\new.avs\mpeg2dec3.dll")
LoadPlugin("C:\Program Files\DoItFast4U\new.avs\decomb.dll")
Mpeg2Source("VTS__01_P01.I-TFF.16~9_1.d2v",idct=7)

FieldDeinterlace(full=true,dthreshold=7)
AddAudio()
ConvertToYUY2()

Encoding as interlaced, speed averages 1.8-ish [never better than 1.85]

import("C:\Program Files\DoItFast4U\new.avs\addaudio.avs")
LoadPlugin("C:\Program Files\DoItFast4U\new.avs\mpeg2dec3.dll")
Mpeg2Source("VTS__01_P01.I-TFF.16~9_1.d2v",idct=7)

AddAudio()
ConvertToYUY2(interlaced=true)

[BTW: processing as interlaced gives far, far better quality results; & on the rare occasions I process progressive source, speed is approx 2.05]

. . . AVISynth & plugins versions & [to a lesser extent] idct# seem to affect speed more than CCE version [obviously enough when it's waiting for a stream being deinterlaced]: above speeds are with AVISynth 2.52/MPEG2DEC3 1.08/DECOMB 4.x; CCE is 2.67x, priority idle, CPU usage [on the rare occasions the system is encoding alone] approx 83% average.

- Roveer - Hi - what speed-difference if any over default settings does icdt=5 make for your Xeons? - this as I'm sure you know is meant to force/enable SSE2.

[Blackout]

Ive got a AMD 2400+ and i get 1.06 speed for CCE 2.67.

I used to get crashes also when doing anything else with the pc at the same time until i got 1gig of DDR ram. :) now its yet to freeze once. I think it helped that the two sticks of ram are exactly the same brand and speed now too.

ppl with similar setups but intel eg those with a p4 2.4gig seem to get a lot more speed than AMD, even tho our processors are "theoretically" matched. disappointing.

blackout

[SVCD4Me]

I've noticed it matters what the original/new bitrate are. I average anywhere from 2.2 to 2.6 on average. Haven't seen it go lower than 2.2 even on a 60% encode.

[shoarthing]

SVCD4Me - Hi - pair of Bartons at 16x 141 [ie 2.25GHz @ default vcore] encode CBR 6000Mbps at 2.4x with the same sort of *.avs as yours.

May drop the multiplier & go for a bit more FSB when I find the upper limit for these AQXEA stepping Bartons.

[DaRat]

Well, somehow memory SHOULD be an issue for CCE since I saw it eating down about 676 megs of my 1 gig.

[DDogg]

CCE speed comparisons are not valid unless it is the same source (ntsc, pal), cce version, script and mpeg2dec version with IDCT variant specified. Anything else is futile, and frankly a little silly.

To that end we once worked on an avisynth only script that generated patterns and color bars and such so that it could be run by anybody and actual yield factual and useful information. I don't know what ever happened to that project.

Probably the closest thing around is a copy of the Matrix and demo CCE. Seems everybody owns or can easily obtain the Matrix. Using a easily downloadable demo version of CCE would then give a good benchmark IF the version of Mpeg2dec and the script was specified.

If anybody is interested I can try to put this together so we can really compare apples to apples.

[Edit: I might be able to do a script that would generate a small text report with all the timings and amount of frames encoded and other pertinent information which could then be posted. Better yet, I think I will ask a person (Bilu)who I think of as an expert in this reporting stuff, to consider doing a report script for us.]

[Edit2: As a start here is a fairly standard svcd script for the matrix (NTSC) using mpeg2dec10 from the avisyth development thread and CCE 2.67.00.13 Demo. The d2v is standard forced film from dvd2avi 1.76:

mpeg2source("D:\matrix\DVD2AVI_PROJECT_FILE.d2v",idct=6)
BicubicResize(480,368,0,0.6,8,0,704,480)
AddBorders(0,56,0,56)

CCE set as CBR 2584, no vaf,no audio. Progressive and zigzag the only things checked in quality tab.

My speed is about RT 2.43 after 6 minutes elapsed on a xp1700 OC at 183X12 (2.2 ghz) w/ 512 megs HyperX running single (64 bit) channel 6-2-2-2 on a cheap Biostar Nforce Ultra 400 board (M7NCD Pro). I'll post my CCE 2.50 times tomorrow.]

[nuno]

Hi DDogg !

I think you refer to your proposal for a "cee stress test".

Here goes the link.

http://forum.doom9.org/showthread.php?threadid=17518&highlight=stress+test

PS. I'm thinking on upgrading my system and i'm curious about your results.

Cheers

[DDogg]

nuno, thanks for reminding me of that old thread. I'll start a new topic (http://forum.doom9.org/showthread.php?s=&threadid=66134) and hopefully we can get some meaningful test data.