Can't belive... new machine - speed 3.4

[jmartina]

I am using CCE for almost two years now (from the times of Pioneer A03), and have encoded some 200 movies
using it.

For about a year I had Gigabyte 8IK1100 i875 motherboard with P4 3.2GHz, 1 GB dual channel DDR400.

Couple of days ago, I have replaced my machine with new Intel LGA MB with 915p chipset,
P4 3.4GHz, 1GB DDR2-533 dual channel.

On most tests (mine, or on the internet... tomshardware etc..) this new machine is slightly slower
than 875 boards like one I had before (not ALL, but MOST tests).

Something like: "Yes DDR533, but CAS latency is 4, so...blah, blah...)

I have noticed that fact in real life, using same applications that I used before, and (subjective)
there was no significant speed difference between two machines.

First movie that I get to encode on new machine was Troy (PAL).

Using CCE with most movies on my old machine resulted in speed of 2.3 - 2.4.

I was amazed when I saw speed 3.3 - 3.4.

First thing on my mind was: "There is something wrong with video, it can't be this fast".

But I didn't stopped encoding (I have watched UEFA champions league game Milan-Barcelona so I
didn't care at the moment).

On my suprise, everything was OK!!!!!

This is for the moment only movie that I tried on this new machine, but..... speed index 3.4 ????

This is just first impression, I just wanted to share with everyone,
so I don't know is there something about this movie, or what??

By the way CCE is 2.66.01.07.

[hendrix]

i have a 2.8Ghz PIV system w/1gig DDR400 - and i get around 3~3.2 on my CCE 2.67.00.27 when encoding progressive...the time is around 1.8~2.5 with interlaced material - CCE is a fast encoder

[jmartina]

Hendrix, yes I know CCE is fast, as I said I have encoded some 200 movies in two years. But I ALWAYS had speed of 2.3 - 2.4 with any material (using AVS).

What I meant with this post is to check with other people that recently changed machines, is it possible that this new technology
(new processors and DDR2) maybe works extremly well with CCE ??

As I said, most apps on this new machine works with 2-3% speed difference comparing to old machine.

EXCEPT FOR CCE !!!

[hendrix]

if DDR2 and the new processors make CCE faster then I need to get them

[br408408]

Using DGIndex, DGDcode, AVISynth 2.54, and CCE 2.5 (yes, 2.5...I like it better even though 2.67 runs faster), I get encode speeds of 3.1 on progressive NTSC material 720x480 DVD9>DVD-R (bit rate reduction) on the computer in my sig. Test with CCE 2.67 give speeds of 3.4 RT

Bill

[jmartina]

Hi br408408,

Can you specify what movie gives you speed of 3.4 ??

Like I said, I had GB 875 motherboard, 1GB dual chanel DDR400 (CL2.5), P4 3.2.

I just can't belive that your machine can be 50% faster than mine!!

Also, in past two years since I've been member, I have never seen
such high speed index. That's the reason in the first place that
I opened this thread.

There are some postings from couple of months ago from people that
have dual Xeon machines, and even their results were not so high!!

I am using MPEGDecoder.dll, and you DGDecode, could it be the key ??

[hendrix]

ive had 3.4 encodes before...i really dont remember which movies though..it isn`t impossible - some movies just encode really well

[erdoke]

Maybe it worths the time to read this article:
CPU video coding roundup (http://www.digit-life.com/articles2/intelamdcpuroundupvideo/)

[jmartina]

Nice link erdoke, but unfortunately it doesn't covers new
LGA 775 processors or DDR2 533.

[erdoke]

Originally posted by jmartina
Nice link erdoke, but unfortunately it doesn't covers new
LGA 775 processors or DDR2 533.

It has its conclusions about CCE as well. Generally speaking, what really matters for CCE is clock speed.
Considering this I would say your speed jump is more like due to the movie than your new platform. You have only 200 MHz advantage and Hyperthreading (improved in Prescott core) is not really supported before v. 2.70.
Double L2 cache size didn't seem to matter much during the test either. New instructions can't keep up with the much longer pipeline.

[jmartina]

I agree.

I will try another movie, and post result here.

[br408408]

@jmartina
The movie I used was "Insomnia" (NTSC, 16:9 progressive). I started using it for no other reason than it was within an arms reach of my computer at the time, and I own it. I have kept using it because speed test with real world sources are only valid if you use the same source...so i am kinda "locked in" to this movie if I want to compare my results with test i have conducted it the past.
I looks like you and I are getting about speeds anyway (our speed with 2.67 in RT is about = to cpu speed in ghz) aren't we?

Bill

[RB]

Keep in mind that CCE speed is also dependent on bitrate, although not dramatically. The lower the avg. bitrate, the faster CCE encodes.

[erdoke]

Originally posted by RB
Keep in mind that CCE speed is also dependent on bitrate, although not dramatically. The lower the avg. bitrate, the faster CCE encodes.

Yes, and of course also depends on aspect ratio and the selected filters.

[crespo80]

I can't believe of these speeds!!!

I have a decent PC (see in signature, for normal use I set the cpu clock to 3000/3200MHz) but I never see these speeds!
I encode at most at 0,95/1,0 ...
I use CCE 2,67.00.27.

I encode my VHS interlaced caps at 704x576 without any resize/filter, with constant quality=40. The average bitrate vary from 6000 to 8000.

The speed I achieve is quite twice faster than TMPGenc!
Which are the parameters that influence most the quality of the video?
Why do you have speeds 3 times faster than mine?
Which kind of material do you encode and with wich settings?

:confused:

[jmartina]

Hi crespo80,

"Can't belive.." as I said in the title of this thread.

But you can see in first post, I had machine like yours, and
this speed increase happened after upgrading to new technology.

I will post more details about machine and type of movies later.

********************

For br408408,

I will reencode "Insomnia" again and post result here. The difference
will be only in video system (PAL, since I am in Europe)

[shoarthing]

"Generally speaking, what really matters for CCE is clock speed." - no, I don't agree at all.

A dual Opteron at 2.4GHz encodes normal PAL source at normal 'backup' bitrates w/out filters at around 4x - a dual AthlonMP (Barton) at 2.4GHz encodes the same source at under 3x. Same CCE-SP, same AVISynth & etc & etc

- IMHO memory speed & SSE2 implementation are keys to CCE-SP speed.

[jmartina]

The movie I used was "Insomnia" (NTSC, 16:9 progressive).
Hi br408408,

I have just tried "Insomnia", and I am getting 3.2 RT,
but, the problem is that your NTSC version is progressive, and
my PAL version (16:9, 720x576) is interlaced,
so, results are not actualy comparable.


I encode at most at 0,95/1,0 ...
crespo80, there is something very wrong with this results of yours.

I assume that you are doing something wrong, because even
not overclocked machine like yours (2.4GHz) with ANY kind of
material should have index of at least 1.5 RT.


- IMHO memory speed & SSE2 implementation are keys to CCE-SP speed.
In next couple of days, I will have one machine like
mine, but with regular DDR400 memory, so that will be
oportunity to test how big influence memory have on results.

I will post results here.

[erdoke]

Originally posted by shoarthing
"Generally speaking, what really matters for CCE is clock speed." - no, I don't agree at all.

- IMHO memory speed & SSE2 implementation are keys to CCE-SP speed.

Have you visited the link I inserted above?:confused:
Of course the statement you quoted is true for similar architectures only. We were talking about the difference between Northwood P4 with dual DDR 400 RAM and Prescott P4 with dual DDR-2 400 RAM.
You'll see that there is no real difference between ANY of the A64 CPUs, doesn't matter if they have 512 KB or 1 MB L2 cache, dual channel MCH or single etc, as far as they run at the same clock speed.
The same is true to the NetBurst architecture, Northwood Celerons are just a little bit behind NW P4 CPUs with their 1/4 L2 cache and 1/2 FSB clock.

I had a 3,2 GHz NW P4 before this Opteron 146 workstation, and believe me the P4 was around 20 % faster with CCE encoding despite its 10 % lower memory speed (both supports SSE2). If you visit the above link, you will realize that a 2 GHz A64 is faster than a 3,2 GHz P4 in a couple of encoding types but certainly not with CCE.

[br408408]

Yes, memory speed seems to be more of a problem with the AMD XP line of CPU's, and not the 800 mhz fsb P4's or A64's. With my AMD machine 166 x 15 is much slower (with CCE) than 227 x 11, even though the core speed is almost the same. Lets face it, I loved socket A, but it is getting old...800 fsb P4's and A64's have better memory bandwidth than socket A, and there for it is not an issue for them with CCE...until we can push the core speed much higher (if we can).

@RB
You are right, I didn't say what bitrate I used when running my speed test and jmartina would need to know this if he wanted to run the same test. It is: 2000 min, 9000 max, 5500 ave.

Bill

[dragongodz]

erdoke - i think you need to reread the CCE conclusions on that page.

Cinema Craft Encoder SP 2.67 prefers a large cache and the NetBurst architecture, with the latter being more vital. It doesn't benefit much from Hyper-Threading, as well as from memory bandwidth.

:D

[jmartina]

You are right, I didn't say what bitrate I used when running my speed test and jmartina would need to know this if he wanted to run the same test. It is: 2000 min, 9000 max, 5500 ave.

You dont need to tell me what bitrate you used, bacause I know.

We all need to pack DVD9 movie to DVD5, are'nt we?

So If we used same movie, bitrate would be aprox. the same.

And indeed, it was.

As I said in my post before, problem is that PAL version of "Insomnia" is interlaced and NTSC (at least what you said) is progressive, so results are not comparable.

[erdoke]

Originally posted by dragongodz
erdoke - i think you need to reread the CCE conclusions on that page.


:D

If so then I agree, I really have to reread the P4 part...:D

[jmartina]

Erdoke, you said:
Double L2 cache size didn't seem to matter much during the test either.

But conclusion for CCE in that article is:

Cinema Craft Encoder SP prefers the NetBurst architecture (Pentium 4 / Celeron) and is not indifferent to the L2 cache.

[erdoke]

Originally posted by jmartina
[B
But conclusion for CCE in that article is:

Cinema Craft Encoder SP prefers the NetBurst architecture (Pentium 4 / Celeron) and is not indifferent to the L2 cache.[/B]

What you quoted from me, I said it related to Prescott vs. Northwood.
I suggested that clock speed matters much more than anything else. Later I also clarified on this, that thorough comparison is useful only apples to apples, like A64 Clawhammer vs. Newcastle, NW P4 vs. NW Celeron, etc.

[br408408]

@ jmartina
Just about all movies on DVD in the USA (where I live BTW) are progressive...the exception would be low budget movies shot on video tape, porn, and some seasons of TV series that are released on DVD such as "The Sopranos" and "Friends"...etc. I don't own any of these.


" (Insomnia on) NTSC (at least what you said) is progressive"

From this statment, you seem to doubt if I have the know how to tell a progressive movie from an interlaced one...well, I made my first SVCD before Doom9 even had this site, so I think I can tell....just maybe.

Anyway, if you have any progressive movies, if you list them, I will see if I have any of the same ones and will test one for you on my system if you would like.

Bill

[jmartina]

From this statment, you seem to doubt if I have the know how to...

br408408,

Sorry, but english is not my native language, so I made mistake.

I just wanted to say that you said that movie is progressive,
it was not my intention to said that you don't know what you doing.

I am very sorry....:(

[br408408]

Originally posted by jmartina
br408408,

Sorry, but english is not my native language, so I made mistake.

I just wanted to say that you said that movie is progressive,
it was not my intention to said that you don't know what you doing.

I am very sorry....:(

At first I thought of this, but you english seemed so good, I thought it might have been you native language....anyway, I understand....sometimes things said in one language don't "sound" the same way in an other language, so I understand.

I am sorry too. And you english is very good, by the way.

Bill

[jmartina]

Thanks for understanding Bill, and by the way, i will have a new machine for test in next couple of days.

It will be based on Gigabyte 915p motherboard that supports both DDR and DDR2, so we will see is memory
makes any difference for CCE.

And also, I will encode few progressive movies during 2-3 days, and post results here.

[jmartina]

Well, here we are:

MEMORY !!!


I have tested two equal configurations (to be exact one computer with two different memories).

First, the results (both memories dual channel):

DDR-400: 3.15 DDR2-533: 3.62



Computer:

Gigabyte GA-8I915G-Duo
Intel Pentium IV 3.2HGz (LGA775)
RAM: 1GB (2 x 512MB)
Gainward GeForce 6600GT 128MB
HDD Hitachi SATA 160GB
Windows XP SP1


Movie:

Terminator 2, PAL, 720x576, 16:9 (2.35:1), progressive, duration 2:26:25

CCE 2.66.01.07 two-pass encoding (2500:3700:8500)
Avisynth 2.5.3
MPEGDecoder.dll



Also, this means that this motherboard is much faster than mine Intel D915PBL.
Of course, I knew that Intel motherboards are not speed kings, but are known to
be very stable. For years I had Asus and Gigabyte motherboards,
but I thought: "Well, this is new technology, it's better to go with the guys that made it".

Intel D915PBLL ”Battle Lake”
Intel Pentium IV 3.4GHz (LGA775)
RAM DDR2-533 1GB (2x512MB)
Gainward GeForce 6600GT 128MB
HDD Hitachi SATA 160GB
Windows XP SP1

Result: 3.41 (With 3.4GHz CPU, while above results are made with 3.2GHz)