A real CCE SpeedTest

[DDogg]

Originally posted by vipera

My results using the original script test:

CCE 2.5: 1.920
CCE 2.67: 1.88

A64 3000+ @ 2470mhz (247x10)
That's interesting. Using CCE 2.67.00.27 I am at 1.85 RT running at 2.74 gigahertz (210x13) whereas you are at 1.88 RT running at a lower 2.47 gigahertz but with the much higher FSB of 247 megahertz.

Your CCE 2.5 speed is about where I would think it would be, but your very fast speed of 1.88 with 2.67 would seem to indicate that 2.67 better utilizes the faster ram speeds [or maybe it just likes AMD 64's, or both].

[vipera]

Originally posted by DDogg
That's interesting. Using CCE 2.67.00.27 I am at 1.85 RT running at 2.74 gigahertz (210x13) whereas you are at 1.88 RT running at a lower 2.47 gigahertz but with the much higher FSB of 247 megahertz.

Your CCE 2.5 speed is about where I would think it would be, but your very fast speed of 1.88 with 2.67 would seem to indicate that 2.67 better utilizes the faster ram speeds [or maybe it just likes AMD 64's, or both].

Athlon 64 has SSE2. Not sure if it helps the Athlon as much as it does the P4, or if CCE even uses it.

In most applications the A64 is much faster than an XP at the same clock. In CCE it seems that for the most part only raw mhz matters, or maybe it could be optimized even more for A64 in the future. ;)

[djan]

My results using the original script test:

CCE 2.5 : 1.350

Using AviSynth 2.5.5, ASUS A7N8X-X, Barton 2600+ @ 1.91 Ghz (11.5 * 166), 512 DDRAM.

Is it a correct score ?

[djan]

Nobody is interrested anymore ?

[coona]

Uff. I sorted all (relevant) posts into table. Unfortunatelly in some posts there are not enough info (such as size of RAM, Dual/single channel mode, FSB speed). I guess (as Karovaldas suggest) there should be some additional variables which can help to analyze whole issue properly (his post is on page 5 (http://forum.doom9.org/showthread.php?s=&threadid=66134&perpage=20&pagenumber=5)). I hope that this attachment will be processed quickly by mods of this forum ;).

[influenza]

I don't see any attachments to approve?

[Turd_Ferguson]

alright. using CCE 2.67.00.27 and Avisynth 2.54, i ran 2 tests to see the difference OCing my P4 2.5GHz 400FSB (no HT) makes...

0.99 @ 2.5GHz 100FSB x4
1.18 @ 3.0GHz 120FSB x4

i was planning on upgrading to a socket 939 AMD64, but so far it looks like the 800FSB P4's are doing better at this test when OCed. Anybody run this test in the 64-bit WinXP?

oh yeh, this is the rest of my system:
Gigabyte GA-8IEXP mobo (single channel memory)
512MB DDR400 (mobo will only run it at ddr266 though:mad: )
1x WD 80GB 7200 8MB cache (boot drive)
2x Maxtor 200GB 7200 8MB cache in RAID-0 (ran the test with these drives)

i ran the test on my old P3 733MHz just for a laugh using the same versions of CCE & Avisynth and it got 0.25

[coona]

I don't see any attachments to approve?

Me too :confused:. But when I want to edit my post attachment seems to be present. I´ll do some update because of Turd_Ferguson´s inputs and I´ll post it again...

Here (http://web.firma.cz/~sedlacek/Doom9/A_Real_CCE_SpeedTest_Results.zip) is a link to download it form other source.

[influenza]

Hopefully I'll be more succesful posting the attachment

[DDogg]

coona, thanks for putting those results in a spreadsheet. Great job! Here is another result.

Recently I tested one of the modded bios for the NF7s V2 which allowed me to run at a higher FSB with my HyperX bh5 PC-3500 ram. The results bear out how important ram speed is.

For those that have followed this thread, my previous best speed was obtained with a 210 fsb and a 13X multiplier which yielded a CPU speed of 2.74 gigahertz. With these settings I was just able to break 2.010 RT with CCE 2.5, and 1.85 RT with 2.67.00.27.

With this new bios I was able to increase the fsb to 222 with a 12X multiplier yielding 2.64 gigahertz. I expected slightly lower scores with the lower CPU speed, but speeds actually increased [very slightly]. The increased fsb seemed to be what made the difference. SiSoft showed: http://img25.exs.cx/img25/4300/222x12_final.jpg

Speed test results were 2.015 RT with CCE 2.50 and 1.89 RT with CCE 2.67.00.27.


ABit NF7-S Rev 2 | Mobile XP 2600 @2.64 Ghz (12*222) VCore @1.85 volt, Chipset Vdd @ 1.70 volt |(2) 256 meg PC3500 HYP_X (11-2-2-2 VDim 2.9 v Dual Channel @ 444 MHz)

[coona]

New testing machine

CCE 2.5x: 1,201

Intel Creede D865PCD | Celeron D325 2,53 GHz (256/533) | PQI 256MB DDR - single channel

P.S. I can´t add an attachment again :confused:

A_Real_CCE_SpeedTest_Results.zip (http://web.firma.cz/~sedlacek/Doom9/A_Real_CCE_SpeedTest_Results.zip) (FSB and Multiplier added)

[influenza]

The attachments need to be validated by a mod before they show up. More specifically by a mod of this particular forum, so that's why ;)

[Turd_Ferguson]

so i got bored and ran DDog's test on every boxed computer set that is sold at my workplace. the results are in the attachment. i used Avisynth 2.54 and CCE 2.67.00.27.

The suprise of this test was the Celeron D powered Emachine that scored a 1.22. it's a $600 complete system that beats out my own friggin' OCed P4, but not for long... i will post the score on my new OCed and watercooled 2.8E as soon as i get my parts from newegg.

[DDogg]

Bumped settings up again and it made a little difference.

New data point for 2775 MHz (12.5 X 222):

CCE 2.50 = 2.064 RT
CCE 2.67.00.27 = 1.94 RT



ABit NF7-S Rev 2 | Mobile XP 2600 @2775 MHz (12.5*222) VCore @2.00 volt (actual), Chipset Vdd @ 1.70 volt |(2) 256 meg PC3500 HYP_X (11-2-2-2 VDim 2.9 v Dual Channel @ 444 MHz)

[soujir0u]

I get 1.835 with CCE 2.50.

My system specs are:

Athlon XP-M 2600+ at 2.4GHz (218x11) [1.675V]
1GB Corsair PC3200 Value Select at dual-channel 218MHz, 2.5-3-3-11 timings [2.7V]
Abit NF7-S v2.0 [1.6V]
Samsung 160GB 7200RPM PATA 8MB Cache

I could probably bump up my overclock for higher results but this is the setting I run it at for everyday use.

[coona]

i´m getting

CCE 2.5 : 2.130 RT

with AviSynth 2.5.5 with the same settings like my first testing months ago in this same thread at default XP 2400+(2.0Ghz) speeds!!

Are you sure you can get 2.13 with XP 2400+ at 2200MHz. What did you change from your last testing? Only version of AviSynth?

[Turd_Ferguson]

new system is complete! here are the specs:

3.2 prescott @ 3.76GHz (235FSB, HT enabled, 46C max temp)
Asus P4C800-E Deluxe (Turbo and PAT enabled)
1GB (512x2)OCZ DDR550 (running 1:1 ratio w/ FSB)
WinXP Pro SP1

now i can get a test score of 2.15 (CCE 2.67, Avisynth 2.54) when i raise the FSB up to 245, but CCE is too unstable to do a full encode at that speed. i normally get a real time speed of about 4.30 (4.17 while i'm writing this).

something to note: when i was building this system, i tried to match some of the "record high" test scores set by the P4 2.4C northwood. i was able to reach a stable 280FSB with a 2.4C, and was still running 1:1 on my memory ratio. but the best test score i ever got out of it was 1.51. so i want to know how the scores on this processor got so high.

anybody?

[KeyserSoze]

Originally posted by coona
Are you sure you can get 2.13 with XP 2400+ at 2200MHz. What did you change from your last testing? Only version of AviSynth?

I deleted my post...i did a mistake :P

[DDogg]

Turd_Ferguson, I've wondered about some of those P4 2.4C scores. Memory timing does play a large part on AMDs and CCE. I assume also for P4's. What are your timings? Btw, thanks for posting some results that can be depended upon.

Well, I got a Swiftech 6002-A water block and boy did it make a difference.
http://img101.exs.cx/img101/2876/2860_12kB.th.jpg (http://img101.exs.cx/my.php?loc=img101&image=2860_12kB.jpg)

No, I can't run 2860 on a daily basis unless it is colder outside so that I can open a window and have ambient < 24c, but it is Prime stable and I could run the CCE tests with no problem. This thing is cooking now! (proud papa) :).

CCE 2.5 = 2.1 RT. Finally!
CCE 2.67.00.27 = 1.98 Grrr! I really wanted to hit 2.


ABit NF7-S Rev 2 | Mobile XP 2600 @2860 MHz (13*220) VCore @2.125 volt (bios) | Chipset Vdd @ 1.70 volt |(2) 256 meg PC3500 HYP_X (11-2-2-2 VDim 2.9 v Dual Channel @ 440 MHz) | DIY Water cooled - Swiftech 6002-A block, Fedco 2-342 rad < O'Reilley Auto part # 399090 < $20.00

[Turd_Ferguson]

DDogg, i use 3-4-4-8. i haven't played around with these timings much since my system is fast enough for me right now, and it is rock solid stable.

btw, it was my understanding that the higher the frequency you run your memory, the less latency timings matter. from the tests i have tried w/ CCE, the things that seem to matter the most are the attributes of 1)processor cache size, 2)processor GHz, 3)HyperThreading, and 4)memory bandwidth.

[br408408]

Turd_Ferguson, WOW...that Prescott is crankin'. With the Northwood in my sig I found that CCE speed in RT is about = to CPU speed in ghz going from DVD>DVD (bit rate reduction)

DDogg, Just picked up a Swiftech 6002-P myself. can't wait to give it a try...gotta get the rest of my water cooling setup finished...already modded my 77 b'ville heater core to single pass. See pic (http://www.ocforums.com/showthread.php?p=3073711#post3073711) I'm sure my 3.0 has more in it as it is running at the my sig at less than 1.5 vcore. You might want to give a P4 with HT a try...even though an oc'ed XP mobile might beat a P4 in the sped test, I think a fast P4 will win when doing a real encode with DVD2AVI AVISynth and CCE

Bill

Edit: Running speed test my AMD system shown below beats my P4. But doing a real encode the P4 wins by about 2.8 to 3.3~3.5 RT


Abit NF7-S
XP2600 Mobile @ 230 X 11 = 2530 1.8 vcore L12 mod
Thermalright SLK900A w/Panaflo 92x25mm High Speed Fan 48C prime95 load
2x256 Twinmos w/ Winbond CH-5

[DDogg]

@Turd_Ferguson - Regarding the timings I did an experiment the other day. I used 8rdavcore (www.hasw.net) to adjust Trcd from 2 to 3 while running the test. I don't remember off hand exactly what the difference was, but it was readily apparent. I'm sure there is a utility like that for your board. It might be fun for you to play around with it.

@r408408 - Hey, nice mod job on the rad. I think my 342 is already SP, not sure. You may be right about the real world performance. One thing for sure though, the comparison source needs to be 4:3 full screen for them to be meaningful. Let me know how the 6002 does on your Intel. I wonder if it is handling the Prescott's OK? I read they are real heat generators. Sounds like TF is not having any problems though.

TF, are you WC, or air?

[Turd_Ferguson]

@br408408
man, i like my black ice extreme 2 and all, but i wish i could say that "i have parts from a 70's GM muscle car in my computer". that is bitchin'. it's even cooler if it makes your system smell like gasoline and oil.

next watercooled system i build is definatly gonna use components from some american street classic... like the AMC javlin.

@DDogg
watercooled, but just the processor. the Asus P4C800-E Deluxe uses passive cooling and never breaks 30C. this is what i have:
Swiftech MCP600 waterpump
Swiftech MCW5000-P waterblock
Bay-Res 5-1/4" half height bay reservoir
Black Ice Extreme 2 Radiator
this is all tied together w/ 1/2" ID tubing and connectors to maximize flow. btw, the MCP600 is an awesome pump.

[br408408]

@DDogg I have a 16:9 HDTV so I don't own any 4:3 material to test. A few years back, I knew that everything in the future was going to be 16:9 so I always bought all my movies in the 16:9 format. Also, I have sold the memory that was in my AMD computer, so it is not up and running right now, so I can't run any tests on it. I had problems running the CH5 in dual channel and not having CCE crash (at any FSB), but in single channel it (CCE) would run fine. So I sold it to someone that I built an Intel system for that would never be overclocked and it is running fine for them. I have not replaced it yet so my AMD is down. Sometimes my AMD would make it through a 3 pass encode with CCE in dual channel so I could get through a movie to see how fast it would run.
So my question is this:
I have the the same software installed in both my AMD and my Intel system:
DGIndex, DGDecode, AVISynth 2.54 and CCE 2.5 (I like the image quality of 2.5 better, even though the new versions of CCE do run faster on my P4, I don't use them)
I use CCE to do bit rate reductions when going from DVD9 to one DVD-R....so I do not resize or add borders like I used to do when I made SVCD's
My P4 at the speed shown in my sig is faster that my XP2600 Mobile @ 230 X 11 = 2530 when backing up the same movie (2.65~2.7 vs 3.1~3.2 for the P4) but the AMD is faster in your speed test. I think this is because the P4 with Hyper-threading is better at multi tasking (and I don't go for all of Intel's marketing BS... "hyper-threading", quad-pumped FSB...I "like" AMD better and would build a dual Opteron system if I had the money)
My question is, why are my "real-world" tests not valid because they are not 4:3. I could see if I added borders or resized...but I don't. I am not asking you this in any kind of a smart-a** way...I would just like to know. I respect your knowledge

Bill

Edit: Thanks for the complements on the heater core...I modded that core besause two 120mm fans fit very nicely on it.

[DDogg]

As for what you mentioned, so long as the source used to compare machines is exactly the same, the comparison ought to be accurate.

As for my comments, I was unclearly speaking to the comparison of speeds between the test 'source' which is 720x480 fullscreen color-bars, to real world source. It should be of the same active resolution - meaning it should be full screen material without black bars. I see many smart people say some real dumb things like, "This test is slow, I get x.xx on "regular" source without mentioning the particulars of the source. If a person is going to compare the full screen color-bar test to a full screen active source then fine (to what purpose I have no idea). What you can't do is compare the speeds from the test fullscreen color-bars script to a 16:9 DVD source with blackbars.


TF, and here I thought I was the only guy that thought the AMC Javelin was a very cool car back when.

[br408408]

@DDogg - I sure your color-bar test is accurate and i think our Mobiles at high clock speed and high fsb's do beat "faster" P4's. I think it is just when you add two more pieces of software into the mix (DGIndex & DGDecode) the hyper-threading of the P4 helps out. I wish someone (like Nvidia) made a nice dual cpu (SMP) chipset that we could run 2 Mobiles on, with a high fsb....I know I only dreaming, it's not going to happen. I would like to see what that would do with a real-world encode.

Bill

[LB]

Well you know, it would be interesting to test the accuracy of the speed test script. What do I mean? Well, lets get a person with an AMD and a P4. Encode the same movie to Huffy avi. Then send feed the movie into CCE. If the speeds aren't very close to identical, then it means that the video generated by the speed script is somehow "better" on an AMD than a P4.

Would be interesting to test out, no?

Oh, and just for the record, when doing a real encode I average about 7RT for 720x480 and around 8RT for 480x480. Not too shabby.

[Turd_Ferguson]

@LB
no fair! you've got twice the processors i have!
no, seriously, WOW. i knew that duallies were faster, but how do you get that kind of speed? i could understand it running 2 sessions of CCE simultaneously at over 4RT, but i had no idea you could get that kind of efficiency on a single session. also, what processor is a 2.4_LV?

another general question: whenever i am encoding fullscreen video, i get about 2.25RT, half what i get on 16:9 video. is this common for anyone else?

[LB]

Originally posted by Turd_Ferguson
@LB
no fair! you've got twice the processors i have!
no, seriously, WOW. i knew that duallies were faster, but how do you get that kind of speed? i could understand it running 2 sessions of CCE simultaneously at over 4RT, but i had no idea you could get that kind of efficiency on a single session. also, what processor is a 2.4_LV?

another general question: whenever i am encoding fullscreen video, i get about 2.25RT, half what i get on 16:9 video. is this common for anyone else?

"LV" simply means they are low-voltage xeon chips; which only means they are better overclockers. What's your source video? I always encode my video using virtualdub into a huffy avi file. Then I send this into CCE. That is the fastest way to encode a video.

[TuRiSOft]

CCE 2.50 : 1.285

CCE 2.67 : 1.21


AMD Athlon XP 2600+ @1917(Barton) : Asus A7V8X-X : FSB 333MHz : 768 Mb DRAM pc2700 2.5-3-3-7.

Screenshoot attached if allowed!

[jwu42]

Dell Latitude D600 (P-m 1.8 - 512MB PC 2700RAM)

CCE 2.50 = 1.183

I couldn't get 2.67 to work :rolleyes:

[Bob01605]

I created and loaded and ran the TEST avs file - At 3 minutes I had a CCE 2.5 speed of 5.156 with this test. Did I do this right? This is a new computer built for me by my nephew. Components listed below..

AMD Athlon 64 3400 processor
1 gig of DDR 400 memory
MCI motherboard
200 gig Western Digital SATA drive

When I now convert a widescreen movie from DVD to SVCD and DVD2SVCD converts to FILM I get CCE 2.5 speeds of about 2.9. If its a 4:3 conversion from DVD to SVCD and again FILM I am getting a CCE 2.5 speed of about 3.35

What speeds do others get with their computers? Did I do the speed test right ?

Bob

[br408408]

Look at your output file...I bet you have encoded an error message. Your speed test should come out slower than a real encode

Bill

[XadoX]

AMD 3200XP
1GB DDR333
Nforce2 Driver V5.1

http://www.8ung.at/xadox/forum/CCE250.gif

http://www.8ung.at/xadox/forum/CCE267.gif

a little bit too slow for an AMD XP3200, I think :(

[SniperKilla]

http://sniperkilla.home.comcast.net/CCE.jpg

1.764 on CCE 2.50

Athlon 64 3000+ s754 overclocked to 2200mhz
DFI lanparty ut nforce 3 250Gb mobo
512mb corsair valueselect pc3200 @ 220mhz 2.5-3-3-8

[SniperKilla]

1.820 after moving processor to 2250mhz and ram to 225mhz

1.841 @ 2290mhz and ram @ 229mhz