A real CCE SpeedTest

[kazam]

Personally I think it's better to use these free clips as opposed to an Avisynth script only for the fact that this situation more closely resembles what we're trying to do.

Let's just standardize the procedure.

I tried ConvertToYUY2(interlaced=true) and unchecked "Progressive frames" in CCE but it didn't (noticably) affect speed. So I'm keeping it simple.

mobl_080.m2v

Same settings as coona.

LoadPlugin("F:\Program\AviSynth 2.5\plugins\MPEG2Dec3dg.dll")
mpeg2source("F:\Rips\test\test.d2v")
ConvertToYUY2()
AddAudio()


P4 2.8C
2GB DDR RAM in Dual-Channel configuration
SATA RAID-0

CCE 2.50.01.00

Speed: 2.1

/kazam

[DSP8000]

Script Used:
#start of CCESpeedTest01.avs
ColorBars (720,480)
ShowSMPTE(24)
converttoyuy2()
#end
CCESpeedTest01.ecl (posted here)

Rig:
AMD Athlon XP 2800+ (Barton Core)@ 2405 MHz
ASUS A7N8X Deluxe Rev.2.0 (Bios v.1007Final)
1.5 GB DDR 2700 @ 200 MHz
60 GB Seagate (2MB cache)
120 GB Seagate (8MB cache)
Used CCE SP 2.50.01.00
Speed: 1.806 constant after 20 sec, peak 1.809.
DSP8000

[nuno]

To Kazam or Coona
Could you just put online the content of your ecl file?
Just to save time to every one.

Cheers !

[kazam]

Sure. Replace paths as necessary.

/kazam

; Cinema Craft Encoder SP -- Encoder Control List

[item]
title=Standard.avs
vid_out=1
vaf_out=1
aud_out=0
vid_file0=F:\Rips\test\Standard.m2v
vid_file1=F:\Rips\test\Standard.m2v
vaf_file=F:\Rips\test\Standard.vaf
aud_file=F:\Rips\test\Standard.mpa
file_focused=0
encode_mode=0
packet_size=4096
timecode=0x0
width=704
height=480
vmode=2
frame_rate_idx=4
cbr_brate=6000
opv_q_factor=60
opv_brate_min=2000
opv_brate_max=9000
vbr_brate_avg=2500
vbr_brate_min=0
vbr_brate_max=9000
vbr_bias=30
vbr_pass=9
quality_prec=15
use_filter=0
filter_val=6
seq_endcode=1
dvd=0
half_width=0
half_height=0
fast_mode=0
progressive=0
alternate_scan=0
non_linear=1
top_first=1
lum_level=0
intra_dc_prec=-1
aspect_ratio=2
gop_m=3
gop_nm=4
gop_hdr=12
seq_hdr=1
all_closed_gop=0
fix_gop_length=0
mpeg1=0
mpeg1_cps=0
samples_per_sec=44100
stereo=2
brate_idx=7
crc=1

[file]
name=F:\Rips\test\Standard.avs
type=0
frame_first=0
frame_last=450
encode_first=0
encode_last=450

[coona]

As I see kazam posted his ecl file so there is no need for double posting ;).

[DarkFear_DK]

I get a speed of 1.159 using CCE 2.5.

My computer is a AMD 1700+ @ 2000MHz (160x12.5 - cas 3-3-3)

Note that I'm still ( :( ) using plain old SD-RAM, so it is quite obvoius that memory also serves a great purpose when encoding.

Chipset: Via KT133A

-- DarkFear_DK

[chipvideo]

Closing program
CCE Max Speed: 100.000
Video Encoding finished.

I guess my plutonium upgrade is really fast. Can't figure that one out. LOL.

[Trimegisto]

Started using the sugested scripts and this is what I've got:

CCE 2.50.01.00 -> 1.11
CCC 2.67.00.23 -> 1.25 (with EclCCE)

I'm planning on buying a Dual Opteron system. Let's see how it will increase this values.

[Pfc Joker]

I got 2.24 with CCE 2.67

---------------------------------
Rig =
P2.4 C @ 250 fsb = 3.0
Gigabyte GA-8KNXP Canterwood
1 Gig Geil PC4000 (DDR 500)
Two Western Digital Raptors (Raid 0)
Two Western Digital JB250 (Raid 0) Thats right, half teribyte:D

[br408408]

CCE 2.5 Speed 1.392
CCE 2.67 Speed 1.44
Machine in sig

Seems a bit slow compared to some other posts as I run anywhere from 3.0 to 3.3 RT when encoding a movie to DVD specs using DVD2AVI, AVISynth 2.5 and CCE 2.67. Cut and paste the this link ( http://www.theforumisdown.com/uploadfiles/1203/CCE.JPG ) as it won't open if you just click on it. This is a screen shot of CCE 2.66 encoding the first three VOB's the movie Insomnia with the following AVI script:

LoadPlugin("C:\Program Files\AviSynth 2.5\plugins\MPEG2Dec3.dll")
MPEG2Source("C:\TwoTowers1.d2v")
YV12toYUY2(interlaced=false)
AddAudio()

Bill

[LB]

Hey man. Apologies for not replying back to your pm a loooong time ago. I was just really really really busy with work and had to put good ole doom9 on pause for a few months. Anyway, I got some time and saw your pic. Tell me a few things about that 3.0 speed. And two towers... Hmm... Tell you what. I'm gonna go rent it and try that script, and see what happens just for kicks. But tell me, why aren't you ivtcin'g that stuff? I mean seriously...

Those new speeds you got on this test seem more accurate of your system. Granted, 3ghz is fast, but it shouldn't produce 3.0RT speeds on a 720x480 full-screen video.

BTW guys, you're looking for some video source that is identical that we can use as a test? Simple. Look for a place on the net that houses free video. Here is an example:

http://science.ksc.nasa.gov/shuttle/missions/sts-72/movies/movies.html

I doubt more than 100 people every will d/l the file for the test, and there is no reason why we can't use it for our test. We all pay taxes right? :D anyway, there is a 105second video on there which is 3.5mb. I think that would be an excellent candidate. And remember, who cares if it's not at 720x480. Just insert a bicubicresize into the script. Remember, it doesn't matter how crappy the source is, as long as (1) it is a source from some video; and (2) we all use the same script + source. Whacha think?

[LB]

Okay, I did the test on Two Towers. A buddy of mine had it. With that script, DUAL 3ghz (virtually) I get 1.78 MAX. And this is with CCE 100% maxed out on both processors. I guess going direct from the d2v without even IVTC'ing the file feeds CCE even more threads. In any event, I don't get that 3.0RT speed, so something you're doing is not reporting correctly. Granted, my bus is only at 533 and yours is at 1ghz, but that difference in bus speed shouldn't equal 2.2RT "jump" (remember for other readers, I'm dual and he's a single 3.0, so halve my 1.78).

But anyway, when I do things with my own filters and what not I can get around 2.4 which I'm happy at, so that's all that matters. You're 3.0 will continue to puzzle me. :p

[br408408]

Most DVD's are film...progressive...no need to ivtc. I use "Forced Film" in DVD2AVI and my CCE output file is 23.976. I then use Pulldown.exe to enable 3:2 pulldown. As far as speeds, I am getting those speeds and they are real...can someone with a P4 HT back me up here. And it is 720 X 480. With dual 3.0's I don't know why you are only getting 1.78. My old AMD XP2000 Palamino (1.66 GHz 133FSB- 266DDR) gave me speeds of 1.5 - 1.6. My XP2500 OC'ed to 2156 MHz (227FSBx9.5) gave speeds of 2.3 - 2.45. HELP!!! someone back me up here. All speeds stated in this post are for backing up a movie, and not the "Real CCE SpeedTest". LB, what do you get when you run the CCE Speedtest. Also, keep in mind that in the CCE Speed test, I get a Speed of 1.36, but DDog said in his first post that "Although this test is primitive and will reflect a speed MUCH SLOWER than you will actually get using proper source, it should allow an exact comparison of speed on various platforms." So, the 1.36 in this test would tend to back up my real video encoding speeds. BTW,this is for video ONLY...no audio as I am making DVD's and just use the original AC3.

Bill

[kazam]

Well, like I said in an earlier thread:

If you have a P4 with 800MHz FSB and are encoding progressive NTSC material the CCE 2.67 RT speed should be about equal to your CPU clock in GHz.

No special filters needed, just a 3 line avs script.

And actually I believe a 1GHz FSB as compared to a 533MHz one would make a tremendous difference.

I can get 3.0RT quite easily by overclocking my CPU to 3.0GHz.

That's for NTSC though. PAL is much slower and I get around 2.3RT.

/kazam

[Pfc Joker]

You know if you look at the numbers it seems like everyone running a P4C on a new chipset with raid get close to the same results. Kazam's statement about the CCE speed being about equal to CPU clock speed seems to hold water for me anyway.

[br408408]

Thanks kazam and Pfc Joker
I build a fair amount of computers, and test CCE on ones of interest to me. I have found that with P4c's CCE speed = 100% of CPU speed in GHz...more or less. With an AMD Barton or Thoroughbred with a high FSB (215 mhz or more) speed can get as high as 115% of CPU speed in GHz. Now if they only made a Barton that would o/c to 2.8 GHz or more.....thats an other story.

I am no newbe at this, I made my first VCD before Doom9 had this site. The "Video Squeezer" with the Panasonic plugin. Didn't like the quality so I went to SVCD, even though most info on the web at that time said don't even try SVCD, it just too hard.

No disrespect to you LB, (and I mean that) but just because your machine won't encode that fast dosen't mean that something is wrong with what I or kazam or Pfc Joker are doing. A P4c at 3.0 GHz on an 865/875 Chipset with a 1000 MHz FSB WILL run CCE at 3.0 or more with NTSC progressive video. Just doing a slight reduction in bit rate so I can go from DVD9 to DVD5, I see no need for filters. And since the source in progressive to start with, I don't know why you think I should ivtc.

Bill

[Pfc Joker]

A P4c at 3.0 GHz on an 865/875 Chipset with a 1000 MHz FSB WILL run CCE at 3.0 or more with NTSC progressive video

Agreed.....you got to have the memory also, by that I mean do your test right after the computer boots into windows. I notice a good drop in speed if I do the test after I've been playing with Photoshop CS, Nero, Acrobat or any other memory hog type of program. I havn't seen 3.0 but 2.0 average is not out of the question.

@LB
Granted, my bus is only at 533 and yours is at 1ghz

A 1ghz FSB with memory at 500mhz is in a different ballpark than a 533mhz FSB and 400mhz memory.

[LB]

Well, I ran the speedtest.

2.5 = ~1
2.67 = ~1

So... I guess it might just be the FSB. I also thought of another reason, or addition to the slow speeds. I use ECC Registered memory, so I'm obviously going to notice quite a bit of slowdown from that alone esp. in an encoding program... (sigh). Well I guess I need that upgrade right about now, but my wallet is looking kinda skimpy :p

Anyway, glad we got it figured it and I guess congrats on the good speed my friend :D

EDIT: Sh*t... my pc is only running on 1 cpu the moment. What the hell... I gotta reboot and check into this lol because umm.. yeah, I have two cpus. lol

[chipvideo]

What I don't understand is that with using this test I get around 1.35 and when I do the real life dvd encoding I get 3.14-3.35.

These people that are getting over 2 on the test get no higher than what I get durring the real dvd encoding. So the gap is much less durring the test and real dvd encoding for them.

[Kedirekin]

That's the problem when comparing a dually to a single processor machine. A slow script really biases things in favor of single processor machines, while fast scripts really let a dually shine.

I believe the bias lies in the throughput of AviSynth, which (I believe) only runs on one processor. Both slow and fast scripts will use 100% of one processor, but the slow script will limit throughput enough that CCE only uses 2% or 3% of the second processor. In this case, MP and SP effectively become the same (i.e. using 1 processor), in which case a 3 GHZ SP will outperform a 2.4 GHz MP.

With a faster script, you might get CCE using 50% of the second processor, in which case the 2.4 MP will outperform the 3 GHz SP. In effect, CCE gets 1.2 GHz on the 2.4 GHz MP (1/2 of 2.4), but only gets 1.0 GHz on the 3 GHz SP (1/3 of 3.0) [actually, that's an oversimplification, but it illustrates the point well]. And as the script gets faster, things grow more and more in favor of the dually.

[br408408]

I think there are many things going on with duallies. If you have an AMD dually your board has the AMD 761 chipst (SIS might have a dual chipset...not sure) and it is not the best performing chipset. Secondly, more then likely you must use ECC Registered memory....not the fastest memory. And I think when we all had 1 Ghz or less CPU's, the CPU was always the bottle neck with CCE. Now that we are running up around 3 GHz's, memory, bus speed, and chipset are coming into play, as that CPU must be feed. The other possible problem with an AMD dually is the CCE 2.5 runs best on an AMD, but I think I have read that 2.66 and 2.67 are better set up for SMP. And lastly, AVISynth (which most of us use with CCE) is not set up for SMP. BIOS options are very limited on most dually boards and AMD CPU's really shine when you get you FSB speed up well over 200 MHz. Most dually boards are more of a "work horse" than a performance machine. I am not trying to condemn duallies here. 18 month ago I was thinking of building one for use with CCE, and I had the funds to do it. It was the reasons stated above that made me think twice. I have however read of some Intel duallies getting speeds of close to 4.0 or better!!!

So if you take a dually encode speed and cut it in half and compare it to what some of us are getting with single CPU's, yes, our (single CPU's) speeds look "too fast...something must be wrong"

As far as the encode speeds on my AMD XP2500 @ 227 x 9.5 / Abit NF7-S, they are very much real. Markrb, the retired mod the the DVD2SVCD forum, built a water cooled AMD XP1700 o/c'ed to around 2.4 GHz and was seeing speeds of 2.6 - 2.8 if I remember correctly...I know that is close as it was his set up that prompted me to buy an Epox 8rda+ and an XP2400.

Bill

[floslap]

cce25--1.670

athlon xp 2500+ 11x205@2.26ghz

[PleXuS]

hi all

i did buy a AMD Barton XP2500!
And a new Soltek Dual Channel DDR400 motherboard!
1GB Dual channel DDR400mhz onboard!

When i run it at 400mhz fsb and memmory bus at 400mhz
Speed=2.1 (Maybe higher but it crashed not so good overclock board!)

When i set memmory bus at 333mhz and fsb400 get following results
Speed=1.95 (even running apps on background no problem :) )

seems to be pretty good but i want a better overclocking motherboard os i gonna sell it! and buy Abit one with SATA on andso..


Weard stuff is that when memmory set at 400mhz motherboard resets and cpu gets hotter.. maybe becoz the memmory controler gets to hot don't know.. or firmware bug..

[br408408]

Yes, the Abit NF7-S is a nice board, mine ran at 227 MHz no problem. Sounds like your memory might be holding you back though, not your board. My NF7-S and AMD XP2500 @ 227 X 9.5 with the memory at 227 gave me speeds of 2.3 - 2.45 with CCE 2.5. What are you running for memory?

Bill

[Amnon82]

guys I don't know what are you doing...

I use this script:

###### DRP 2.2 Script ######

## DLL Section ##
#
LoadPlugin("c:\Filters25\MPEG2Dec3.dll")
MPEG2Source("MPEG_Video_0xE0.d2v")

Subtitle(" DRP 2.2 ",480,460,90,180,"Arial",22,$FFFFFF)
Subtitle(" (C) 2004 DRP",480,480,90,180,"Arial",22,$FFFFFF)

Subtitle(" AVAMAT 4 ",480,460,200,290,"Arial",22,$FFFFFF)
Subtitle(" (C) 2004 DRP",480,480,200,290,"Arial",22,$FFFFFF)

and my DRP 2.2 Settings (http://www.dvdr-page.tk)

I got always 1.64 RT.
http://rj-elektronik.de/avideo/images/DRP22rocked.PNG

If I use nic's MPEGDecoder it will be 2.30 RT!

Download MPEG2Dec3 1.10 (http://nic.dnsalias.com/MPEG2Dec3v110.zip)

Download Nics MPEGDecoder 2.03a (http://nic.dnsalias.com/mpegdecoder.html) (don't use it if you use QCCE for encoding or for NTSC movies! It will only work with PAL. (Thx Ddogg for this. I hadn't known this.))

_____________________________________________________________________

My PC: AMD Athlon 2.8+, NVidia GF 5700 U, 2x Maxtor SATA 80 GB, 512 333 2.5 Infinion DDR, Epox KT600, Audigy2, WinXP Home SP1a, Pioneer A06, Toshiba 16x,

[br408408]

This is my script:

LoadPlugin("C:\Program Files\AviSynth 2.5\plugins\MPEG2Dec3.dll")
MPEG2Source("C:\DDC.d2v")
YV12toYUY2(interlaced=false)
AddAudio()

Simple, as I am encoding for DVD, I don't need to resize, and I don't use subtitles. Using this script on my old AMD machine (Abit NF7-S and XP2500 Barton @ 227 X 9.5 = 2156) I got speed of about 2.45...just slightly faster that your XP2800 thats runs at 2.08 GHz.

Amnon82's AMD machine follows my AMD rule of encoding time = about CPU speed in GHz X 110 - 115%

Bill

[LB]

I finally got it all working. On re-install, I didn't install the SMP HAL, I messed up and installed the single cpu HAL, so winxp was only using 1 cpu. But, the thing was, I didn't notice right away. Why? Because my speeds stayed pretty constant with what they were in CCE prior (when I had XP setup to recognize both cpus). So here is what I found out.

On a single XEON 3ghz, I do about 1.12 in the speed test. And on my regular encodes I get about 2.8. On a dual XEON 3ghz, I do about 1.34 on the speed test, and 3.2 during a regular encode.

So what gives?

Well, on a single cpu, I max the chip out at 100%; but on a dual setup, I get about 55% on each CPU. So, this actually explains why a br408408's 3ghz 1ghz bus chip can get 3.0 on a typical encode... lost yet? Well it gets 3.0 because the chip is 100% maxed out by CCE. Now that's my theory, so it could be wrong, so br408408 correct me if I am, but I think I'm right. So, CCE apparantly isn't programmed that well to use both chips to their fullest. If it was, then both would be maxed out.

Interesting eh?

[PleXuS]

At the moment i did get stable maximum speed of
2.10!

XP2500 running at XP3200 at fsb400 Dual Channel 1GbDDR400

I use DoitFast4u! so don't know how to change those scripts!
And the speed was done with CCE2.66.01.07!
CCE2.5 was faaster for my AMD but worked not so stable with my software here!

In the near future i get a Abit NF7S Revision2.0!!
And i hope to get higher specs out of it!
I have no multiplier lock so :)
13.5x200=2700
Running at 2,7ghz stable mode! or higher but that will we see in a couple of days! :)

-=PleXuS=-

[coona]

to PleXuS

There is no need of changing scripts. The only thing you should do is to save the code from Ddogg into *.ecl file and open and run it in CCE. We all have to perform the same tests on different machines. That is the purpose of this thread ;). Real encoding speed depends on many factors like source, used filters, different settings, etc...). So it couldn´t be compared as it is.

[Amnon82]

right coona, we need the same source, script and filters.

So I start a new speed test here (http://forum.doom9.org/showthread.php?s=&threadid=70056) in the forum.:)

[LB]

Originally posted by Amnon82
right coona, we need the same source, script and filters.

So I start a new speed test here (http://forum.doom9.org/showthread.php?s=&threadid=70056) in the forum.:)

... what wasn't clear about it? there is all that stuff if you follow the directions.

1. same script
2. same filters
3. same source

it's all there; just follow the directions, don't create a new thread. :angry:

[LB]

Originally posted by PleXuS
At the moment i did get stable maximum speed of
2.10!

XP2500 running at XP3200 at fsb400 Dual Channel 1GbDDR400

I use DoitFast4u! so don't know how to change those scripts!
And the speed was done with CCE2.66.01.07!
CCE2.5 was faaster for my AMD but worked not so stable with my software here!

In the near future i get a Abit NF7S Revision2.0!!
And i hope to get higher specs out of it!
I have no multiplier lock so :)
13.5x200=2700
Running at 2,7ghz stable mode! or higher but that will we see in a couple of days! :)

-=PleXuS=-

I don't wish to sound rude, but either delete or edit your post. The reason for the scripts is so that we all have the same setup. You disregarded the scripts. You can't do that. You also used your own source, which you can't do. FOLLOW DDog's directions. It's honestly not hard. He even numbered them.

As for the source? Follow the directions! His script AUTO-GENERATES a source. yes, it really does. so again, no new threads and if you can't get it working, ask one of us before posting useless numbers. :sly:

[Amnon82]

@LB: Sorry, if somebody of the admins don't like the new thread, delete it. I only wanted to help.

[Amnon82]

@LB: Sorry, if somebody of the admins don't like the new thread, delete it. I only wanted to help.

[Edit] Sorry my connection is crasy

[Fotis_Greece]

CCE SP 2.67.00.23

2.5 ecl test-->1.12
2.67 ecl test-->1.19
You can see my PC configuration in my signature

[Pfc Joker]

As for the source? Follow the directions! His script AUTO-GENERATES a source. yes, it really does

@LB - The way the script generates the Color Bars is pretty slick, and speaking of color bars and testing. I recommend Digital Video Essentials by Joe Cane. http://www.videoessentials.com/ I thought his previous release was cool but this ones even better. :)

@PleXuS and Amnon82 - Both you two, what happened, we had a good bit of information going on here from the same script at the beginning of the thread. Stop the madness.

@Fotis_Greece - Your speed seems low for your specs, you have the exact same motherboard, burner, video card, and probably the same Western Digital drives to get your 500GB. Same amout of memory but I'm running Geil DDR500. You not overclocking? The only thing left I can see to give me a speed boost would be the two 10k RPM Raptors I also have or overclocking. :confused:

[MaXiMuS]

Originally posted by MaXiMuS
CCE 2.5
0.914
:confused:

http://www.lfchosting.com/divx/_cpu_speed3_CCE.png
http://www.lfchosting.com/divx/_cpu_speed2.png
http://www.lfchosting.com/divx/_cpu_speed.png

lil UPDATE
tested CCE 2.66 & CCE 2.64 (later on will post Real life DVD transcoding test
http://www.lfchosting.com/divx/MaXiMuS_CCE_266_CCE_267.jpg

[Pfc Joker]

You know after looking at these specs it seems to me that all things being equal, RAM, hard drives and final FSB is what makes the difference.

[Stabmaster-Arson]

CCE 2.50 Speed 1.47
CCE 2.50 Speed (One cpu processor affinity) 1.142

Asus A7M266-D with dual Athlon MP 2600+
1280 MB Reg ECC DDR cas 2.5 by SPD (ECC is enabled)

This was with a newsgroup reader decoding binary files at the same time, 800 MB ram in use before I ran the CCE test, which spiked over a gig of ram usage)
I will do another test and edit to this post after a resh reboot later.

[shoarthing]

Hi - dunno whether anyone else has graphed results to date [removing the usual offscale clutter]: if you do, AMD & Intel 'speedtest' results fall fairly well along two distinct trendlines - so at least this test is consistent . . .

. . .but I don't think it relates well to CCE/AVISynth encoding speeds - especially for PIV's, which 'test' very poorly in relation to anecdotal encoding speeds.

IMHO the crude overall impression you might draw fom these trendlines [PIV at best does 2/3 work per tick of AMD] looks a fair enough reflection of known low-level CPU performance but fails to reflect the PIV family's optimisations for encoding work.

- some more results to add to the plot:

CCE Speedtest 2.67 - 1.24
CCE Speedtest 2.5x - 1.25

AVISynth 2.54
CCE-SP 2.67.00.23
CCE-SP 2.50.01.00

Iwill MPX2: 2x Palomino MP's @ 10.5 x 156MHz '2000+' - 1.64GHz
1GB Crucial ECC 2700, timings by SPD, ECC scrub enabled
SCSI 10K rpm RAID5 [plural arrays on plural 64-bit hosts]

NB: This test ran at 58%[2.5x] - 61%[2.67x] average CPU loadings: actual PAL ZZ Progressive encodes run at approx 77% average & at 2.1-2.2x realtime using CCE 2.67x.

[Karovaldas]

Interesting thread. I'll run the tests once I get home...

This thread has focused quite a bit on standardizing the procedure. A bit ahead of the procedure, there's a question of purpose to address. What I mean is this: in order to get answers we should spend some time on identifying the variables that affect the encoding speed--or the variables that are of interest to the users here.

Here's a quick preliminary list that I was able to derive from the posts (I am including things that may be totally irrelevant and also most likely omitting very important ones):

PROCESSOR
- Processor speed (GHz)
- Processor vendor (Pentium vs. AMD)
- FBS (Khz)
- not sure how to handle SP/MP issue--they don't seem to just add up do they?
- Operating Temperature

RAM MEMORY
- RAM size (MB)
- RAM speed (Khz)
- available size at the moment of running the test
- (not sure about this) RAM ECC?
- fragmentation status
- temperature

CHIPSET/MOTHERBOARD
- Manufacturer
- Chipset (what?)
- I don't know much about them--what are the variables that change?

HARD DRIVES
- ATA/SATA/RAID: ATA-66, -100, -133, -150, etc. Is there a way to quantify RAID setup into a number along this dimension?
- RPM (5400/7200/etc.)
- I think fragmentation status also matters (at least it does when we are talking about avoiding dropped frames during capturing)
- If we can't factor RAID into the first section under this heading, then RAID/No RAID.

GRAPHICS CARD
- I have no idea if it is relevant here--does the "AviSynth>CCE>MPEG Output" process involve the graphics card at all?

POWER SUPPLY
- Watts
- Voltage (120/240)
- Frequency (50/60 Hz)

CCE VERSION

AviSynth VERSION

OS

OTHER VARIABLES?



Then, there's more to it than just running an identical the same script. As much as possible has to be held constant while running it, which means all computers should be doing the same thing. Sure we could all run some binary procedures, but the best common denomenator I can think off is doing nothing. I see a number of people reporting that they run various other tasks while running this test. Do us all a favor and sit tight for 3 minutes while running the test (do we really need 3 minutes)?

I'd be willing to put the eventual data into a statistical package of some sort once we have enough data and clean enough design. Sure there will be "noise" variation due to humidity, amount of dust in your machine, usage of Photoshop immediately before the test, etc. That's why we need a lot of users to report their numbers so that the irrelevant variation not due to our agreed upon factors of interest can be factored out statistically.

So what I propose is standardizing the following:
1. Agree on a list of variables that everybody will report along with their test results
2. Run the tests exactly as the original author recommends.
3. Do not run anything else on the computer while the test is running

* Report the amount of RAM utilized by the computer before the test is run.

If we can do this and get somewhere in the neighborhood of 100+ data points, I could figure out what aspects matter most and least. With the present data it is kinda hard to do.

What do you guys think?

[coona]

Very nice post ;).

Does Power Suply and Graphic card really matter? And to the Prosessor section - system temperature doesn´t affect encoding speed, does it?

Anyway - I think we need to sumarize all relevant posts and sort the results with configurations into one table. Any volunteer :)?

[shoarthing]

Karovaldas - Hi - it's easy enough to graph the results to date to see if these are any trendlines: I took twenty minutes out to do this & there are two, showing [per GHz] AMD:Intel 'scores' to be very roughly 5:3

Finer granularity would require, as you say, standardisation of data & more data per report - but I don't think this interesting 'speedtest' proposal has proved itself more informative than anecdotal report - the whole point in proposing the thing.

IMHO the existing 'CCE Speedtest' has three obvious flaws:

1: 'score' cannot be tied by users to actual AVISynth/CCE encoding speed, by a correction-factor common across platforms. This tells us no more than a faster AMD is faster when encoding than a slower AMD in fairly clear ratio to their gross GHz; & similarly but separately with Intel-based systems. This we surely already know . . . .

2: SMP systems' 'scores' are comparable only with other SMP systems based on the same platform. As we all know, CCE is multithreading, AVISynth isn't: in actual encodes SMP platforms have aggregate CPU loadings from roughly 75% upwards [this is one of the times Intel's HT technology may actuallly be of some use]; but in this 'test' aggregate CPU loadings appear to be roughly in the 55-60% region. I suggest it's therefore a pretty fair assumption that what is mostly being tested here is AVISynth.

3: Intel:AMD 'scores' appear, as stated above, to fall roughly in a 3:5 ratio per Ghz. This understates the PIV family's known CCE encoding performance & again, I suggest, primarily reflects how these families of CPU's handle the [FPU specific?] demands of AVISynth.

[Karovaldas]

I did some tests of my own on the original script:

CPU 2.4GHz P4 HT, FSB 800
Soyo Dragon 2 Platinum MB with Pentium 875 chipset
1 GHz of DDR400 RAM
WD ATA-100 and SATA-150 Hard Drives, 7200 RPM, multiple partitions

TEST 1:
CCE 2.50 speed 1.04
CCE 2.67 Speed 1.08

TEST 2:
I ran both tests twice on two different hard drives and the speed did not differ between them--ATA-100 and SATA-150 (both 720 RPM).

CPU usage during encoding 59-63%

TEST 3:
I also turned off the Hyper Threading for two more tests.

CCE 2.50 speed .99
CCE 2.67 Speed 1.01

CPU usage jumped all the way up to 100%, though. While the speed difference might be negligible, the advantage of the HT technology in providing additional CPU resources is obvious! With just under 40% of unutilized CPU one can still do a lot of stuff on the computer while CCE is encoding at full speed.

QUESTION:
What about AMD? What's the processor utilization during your tests? I am curious if the 3:5 PIV:AMD ratio may not be reflecting the idle CPU for the Intels...

EXPLAIN THIS!
While I was running my first test on CCE 2.5, it was running at about 1.04 speed at the 2:30 marker. At about 2:45 it EXPLODED and the speed picked up dramatically. The CPU utilization went up to 100% (this is in HT mode) and at 5 minutes the encoding speed was 2.68. I terminated the test at that point because I thought that there was something wrong with the whole picture.
I repeated the test and the same thing happened again at around 2:44. This time I let it run up to about 4 minutes and the speed again was close to 2.65. I figured there may be something wrong with AviSynth feed. So I went and checked the the output .mpg. It was nice and shiny all the way to the end.
I did not get this effect after I switched the test to another hard drive (SATA). It only happened on the drive where the OS is located (different partition though). The drive is a slave on the IDE1.

Anybody has ideas about what happened? How can I do it again!!! I would love to double the speed of my encodings on the regular basis, but at this point, I don't know how.

TEST 4:
I then did a bit of overclocking--went from 200 CPU speed to 205. It was reflected in small speed gains on both CCE versions 1.08 and 1.11 respectively.

I must say I was rather disappointed with the results. I just built this new computer and after doing so discovered that AMD's are both cheeper and faster per CPU click. Hoping to find that HT has advantages in overhead space :-)

TEST 5:
Discovered that my CPU can be overclocked a lot so I turned it up to 3GH (that's 250 rather than 200 Hz and a multiplier of 12). RAM is run at a .8 ratio it is still running at the same PC3200 (400MHz) speed

CCE 2.67 = 1.36 Hey, that's 26% gain is speed for the 25% increase in the processor speed.

[Trimegisto]

Ok, brand new Opteron machine

CCE 2.50.01.00 -> 1.285
CCC 2.67.00.23 -> 1.33 (with EclCCE)

Not much increase over my previous values with the dual 1800MP

CCE 2.50.01.00 -> 1.11
CCC 2.67.00.23 -> 1.25 (with EclCCE)

I'll see what I can do to optimize this. And wait for a 64bit version of CCE :D

[six6]

WAIT!

This thread seems to have fallen into a state of disrepair. I hope (yet wonder if) someone can harvest all/most of these results reliably.

Who knows what people are encoding?! Some have read the whole thread and are zipping some random mpeg through CCE. Others read three responses off the first page, got excited, then followed those instructions. Unless (nearly) everyone edits their posts to specifically state _what_ they've encoded with their script, what are we comparing?

I feel bad about this whole thing! Even the thread author quit half way through the thread, probably disenchanted and put off by the overwhelming critiscm and disrespect for his method. He worked hard for some results, can we just do what he asks? If not, why the heck are you reading this thread? If you want (EDIT: to discuss) SP/MP distinctions (EDIT: concerning something other than this test), start a new thread. EDIT: Do not start a new thread comparing the two with this same test.

I'm _not_ saying suggestions for other methods are bad; I'm saying put a little commentary *with your results*.

I have a similar machine to everyone else it seems...then again, alot of us are on the AMD cheap-overclock bandwagon.

Anyway, my results:

Edit: I've encoded the original script in 2.5
After 3:12 mins of encoding (making dinner instead of paying attn): 1.416 * realtime with:
AMD 2100 @ 166x13.5 ~ 2250 mhz
Asus A7V8X (should have waited for nvidia chipset...)
DDR400 512 MB @ 2-2-2-6

P.S. To anyone upset/worried/flabbergasted over their results...be happy :), you've got an amazing top-of-the-line mpeg2 encoder at your _complete_ disposal, (if only for 30 days :)), free time to sit around and encode video, and the freedom of speech to start your own thread about why your computer is waaaaaaay better than mine. But, I'll bet the guy who makes one movie with his PIII could (and might!) be as satisfied as the guy cranking out 5 movies/day on his nearly-quantum computer.

Edit: bump->Karovaldas

[Karovaldas]

Six6,

Am I missing something, or did you not specify what you were encoding, after all... :)

Version might be important, although it does not seem to make much difference in results...

[six6]

You're right, I removed that when editing the message before! I've added it back, though. How could I be so foolish? :eek:

[insite]

All who are interested, please read my overview of optimizing SMP performance with CCE and AVISynth in this forum. Done some pretty cool research and testing, and the performance increases have been significant.

[Piaf]

cce 2.50 TEST
Hi,heres my 2.50 test

http://www.asdfhost.com/members/superpolm/CCESpeedTest01.png

cce 2.67 TEST
http://www.asdfhost.com/members/superpolm/CCESpeedTest267.jpg
system:
--------
P4 3.0 Ghz/800 MHz FSB
512 MB DDR400
ATI Radeon 9800 XL 128 MB SDRAM

*big edit* :o ;)