Speed : Duron 900Mhz to Athlon XP 1800 !??!?!?!?

[ultradna]

Some hours ago I had a DURON 700Mhz --overclocked---> 900 Mhz


DURON : encoding speed of 0.45
(after the VIA driver update) = 0.54

But now I have the Athlon XP 1800+ :

the speed is now 1.075 ... I doubled the speed for hundrets of $$$ ?

It's a Athlon XP 1800+
256 MB DDR
ATI Rage 128 32 MB
ATA HD
MSI Board
FSB @ 133 Mhz (maybe overclock???)

What's up with this thing ?

BTW I use 4 pass Q.Factor 17 and I get superior results, nearly DVD
quality. But it still took so long.

And when I listen to that song from Emmy Bundon
"What took you so long" ... I could cry !=!=!"=?)342$(="§$/§)"

Maybe someone can help?

THX

[chainsaw135]

I use a 1.4ghz THunderbird for my encodes use only 1 pass with q factor at 5 or i use cbr .but with using either one i get speeds of 1.444 if not atleast 1.4 1.075 seems a bit slow.

[Mozart]

@ ultradna

have a look at cce's tab in DVD2SVCD. Qfactor is not the same thing of image quality priority. Qfactor is only used in 1passVBR.

[ultradna]

@chainsaw135
I heard that if you use "One Pass" with Q. Fact. 5 you should get the same results BUT it is much faster. Maybe I should try this!?

@Mozart
Oh, I didn't know that Q.Fact. has nothing to do with settings higher than "One Pass" :o/

So, but what is the matter ... I got a speedfactor of 1.075 with a
Athlon XP 1800+ ... *ufff*

BTW : I use Win 2ooo with SP2! Do I need any patches?

[chainsaw135]

Yes that is true if you use 1 pass vbr is faster then 4 Pass.
CBR is also faster then 4 pass ..but what i dont understand is why your only getting 1.075 are you clicking on temporal smooth?
if so then that would be why other then that your encode speed atleast in my case not sure if this is the general rule but that doesent change. So if i do an encode with 1pass 4 pass or cbr i still am getting 1.4-1.444 ect

[markrb]

I have an Athlon XP 1700 + overclocked to 1.6Ghz with 512MB DDR.
I get about 1.5 in CCE.

Open your DVD2SVCD log and see if Forced Film was used. If not it could be using IVTC and this will slow you down a ton.
Also if you have TemporalSmoother on you drop 40% in CCE.

Post your DVD2SVCD log file.

If you have a motherboard based on a Via chipset head over to www.viahardware.com and grab the PCI latency patch by George Breese. Current version is .19.

Mark

[ultradna]

No TemporalSmoother
No IVTC (I have PAL DVDs only)

...

[markrb]

If you want us to help you will need to post that log file. We need more info.

Email me and I will go over your system with you to see if you have good settings in your bios. Sounds like something is slow.
If Sisoft ever gets their download page back up download Sandra.
Do a google search on Sisoft Sandra they are in the UK. Get at least version 2001.?.?.11 or newer. The lastest is 2002.?.?.?.
Run it and click on system info.
Email me the following info from the output of Sandra not what you know is in the system. I need to know what the system thinks it has.

Processor Model
Processor Speed
Memory bus speed
System Chipset
Installed memory

Sandra link http://www.sisoftware.demon.co.uk/sandra/
Get it here they have the latest shareware version http://www.simtel.net/pub/pd/57463.shtml


Mark

[mrbass]

Ssounds about right to me. PAL has more resolution to encode than NTSC. I know you said it wasn't interlaced but still. If it's a 16:9 it takes longer than a 4:3 source. If you can get a hold of a 4:3 NTSC DVD non ITVC then you should hit about 1.7 CCE RT. I'm putting together a AMD 1700 tommorrow/today/tonight and am expecting around that speed myself.
On a AMD 1600 I get around 1.5-1.6 for 4:3 NTSC and around 1.2-1.3 for 16:9 NTSC.

where's the longest thread about CCE speeds? ok here it is
http://rilanparty.com/vbb/showthread.php?&threadid=5777

[markrb]

Didn't think of that, but send me the info anyway. It will only take a second for me to see if there is any obvious problem.

I get about 1.5 16:9 NTSC no IVTC.


Mark

[ultradna]

Ok, ... I tried this with "One Pass" and Q.Fact. 5 -- It took something about 4-5 hours. Quality wasn't as good as I set it to 4-pass and quality priority 17 and I got 3 files! Normaly I get with 4-pass 2 Files (~798mb each)

+++ @markrb +++

SiSoft said it is :

AMD Athlon(tm) XP 1900+ @ 1.61GHz
Performance rating : PR2143 (estimated)
L2 On-board Cache : 256kB ECC synchronous write-back

System Chipset :
Model : VIA Technologies Inc. VT8366 Apollo KT266 CPU to PCI Bridge
Bus(es) : ISA AGP PCI USB
FSB speed : 2x 140 MHz (280 MHz data rate)

Installed Memory : 256 (82% true allocated load)

[chainsaw135]

What took 4-5hours the encoding part or the whole process?

Also how long is the movie that you tested this with?

With a movie that is 146minutes i can encode it with 1pass vbr q factor at 5 in 1 hour 15minutes with my 1.4ghz amd tho i dont use the via chipset.

3 more questions under the bitrate tab..how many disks did the length of that movie say you would get? and what bitrate range did dvd2svcd say you'd be getting?

And did that bitrate get carried over 2 CCE?

[ultradna]

@chainsaw135

Q: What took 4-5hours the encoding part or the whole process?
A: The whole process (maybe faster, I was sleeping)

Q: Also how long is the movie that you tested this with?
A: 121 mins. - Return of the mummy (PAL)

Q: 3 more questions under the bitrate tab..how many disks did the length of that movie say you would get? and what bitrate range did dvd2svcd say you'd be getting?

A: If I use 4-pass I got 2 CDs (each 798 mb)
Bitrate, ... I'm not sure, I think that bitr. showed next
to xx - 120 min. But the quality with 3-pass is superior!

[chainsaw135]

Well if you have updated all your via software, then you should be set...maybe someone with more knowledge of Pal verses NTSC can say if one encodes faster then another, meaning maybe NTSC is faster at encodes then doing a PAL ect.

[ultradna]

When I encoded in one pass and q. fac 5 I got a speed of 1.6 !
But multipass has 1.0xx

[ultradna]

After installing the "4 in one" VIA Driver (Update) I got ...
( 3-pass / image quality pr. 15 / anto noise filter 2 )
... a speed of 1.157

It took 1:3o hours for the "VAF" file & 4:30 hours for the
"MPV" file ... and for the AUDIO it took ~20 - 30 min.

6:3o hours for a "101 min." - "PAL" - "16:9" ... is this okay?

[markrb]

I really have no experience with PAL so it's hard for me to judge.

If you want to check your system and see if it's running up to snuff run these Sisoft benchmarks. Make sure they are running alone and on a cleanly booted system.

Memory Bandwidth Benchmark
CPU Arithmetic Benchmark

They will give you two scores each.
Email me all of them.

Mark

[Slick]

Originally posted by ultradna
It took 1:3o hours for the "VAF" file & 4:30 hours for the
"MPV" file ... and for the AUDIO it took ~20 - 30 min.

6:3o hours for a "101 min." - "PAL" - "16:9" ... is this okay?

No!

if cce quotes 1.0 then the movie will take the same time to encode as the original movie length.
I.e: "101 min." will take 101 mins to encode at a speed of 1.00
a speed of 2.0 will encode in 50.5 mins

Dvd ripping varies enormously. but should be completed in around 15 - 30mins.
sound ripping & encoding time varies but should take around 30mins.
muxing will take around 15mins
cue file creation around 15mins.
Total time incl Cce: 191 mins / 3.18hrs or less with cce speed => 1.00

hope this helps


----share what you know...learn what you don't----

[Fmazzanti]

@ultradna,
a couple of months ago I upgraded my system from 1GHZ Th to a 1800+ XP, and found something similar. Right now my CCE encoding speed for a 16:9 PAL movie goes around 1.225 with 5 pass VBR. And that's perfectly OK. If you get something slightly slower then you're ok, no matter what other people say. All those guys who claim to have speeds of around 1.4 or 1.5 encode NTSC which is certainly lower resolution.
So take it easy... if you want to improve further over those figures... wait until AMD K10 or K12 is released :D
Have fun in the meantime.

[Flinch]

i'm using an AthlonXP1600+ and i get about 0.85 when encoding PAL-movies with VBR. encoding pal-movies is of course somewhat slower than ntsc-movies because of the different resolutions:

NTSC: 720x486
PAL: 720x576

[ultradna]

I am satisfied with the results, but I just thought that
the new MSI board & cpu & 256 DDR RAM speeds the system up.

For a 121 min. movie it takes about 8-9 hours (all steps) to have 2 "ready-to-burn" MPG SVCD files (4-pass).

For a 121 min. movie Multipass : 1 (onepass in multipass) +
"qual. prior. 1" it takes for all steps nearly 2 hours, but I don't
think that the quality is as good as with 4-pass.

I optimized my bios ( THANX markrb ) and it got more performance!
At DVD 2 AVI it takes before that new bios settings 5:3o min for
a 121 min. movie and now 3:24 min !!!!

But if (@ encoding) I see a speed of 0.5 that means :
movielenght 121 min x 2 = 242 min = 4h 02m encoding time?

But speed 1.0 and < means 1:1 or faster encoding? I allways have
speeds of 1.073, that should take less than 121 min. - But that is
not right! I got encodingtime of about 6 hours !

Anybody can tell me how to calculate the encodingspeed? And how
long it should max. take for a movie on a Athlon XP 1800+ (its
overclocked to FSB 141 = 1600 Mhz = Athlom XP 1900+ )

Cheers,

[mrbass]

in the Q&A see Q61

[goldfinger]

Will you earn some speed by having dual mp1800+ instead of a single xp1800+ in cce video encoding?

[chainsaw135]

I think this point has been brought up before with dual cpu's and what I got out of the posts was the cost verses the increase in speed is not worth it. Alot of programs still wont run under DUal Cpu's. So I mean there is alot more in to the idea then just slapping in 2 chips and encoding. In some cases games might not work other programs that are not threaded for DUAL CPU would not work right if at all. Not woth it in my eye's.

[mkronen]

Don't deinterlace if possible.
If the source is anamorph, don't try to use the 16:9 setting, but 4:3, this evades rescaling.

With these tips, i'm able to encode a PAL source in 0.75 on my Duron-900, 512MB SDRAM, 120Mhz FSB!
:)

[ultradna]

Why shouldn't I select 16:9 if the movie IS 16:9. If I select 4:3 then the movie will be resized in height and I got long faces. I tried it and it didn't work. - BTW: How could this help, fasten up encoding!?

Why does ripping the subs take so long? ~ 2-4 mins! It just have to be seeked and saved as bitmap.

Last question ... what pass is best? 2,3 or 4? If 4 & 3 is similar, then I could use 3-pass. Now I get with 4-pass nearly DVD quality on 2 discs.

THX

[Wh00pS]

Q-1 Read the sticky in the newbie forum on 4.3 versus 16.9

Q-2 dont know,i never use subs.

Q-3 I personally cannot tell any difference in picture quality between 3,4&5 pass so for me 3 pass is best as it is quicker.

[mkronen]

Q2-2

Well, creating 4000 - 6000 subtitle bitmap takes time. This is a proportional to your seek time, and depend on your File System.
NTFS will be much faster than FAT.

[daxab]

I've started using VobSub for subtitles. I describe this in detail in another thread (that got moved to the Newbies forum). If you're doing permanent subs I recommend this; the results look better. A side benefit is that it doesn't create any bitmap files. The VOB scan still takes a few minutes, though.

[daxab]

Also, more on topic, I have an almost identical rig: XP 1800+, 512 MB DDR, no overclocking, Win2K SP2, and I get the same result: CCE encode speeds of between 1.0 and 1.1, very consistently. This is with TemporalSmoother, without it, with VBR, with CBR, with force film, and so on -- nothing really changes it. Even GreedyHMA (I only use it for anime) is the same speed (as compared to force film). This is all NTSC material that's been fudged back to 23.976 fps.

I think part of this is disk/file system speed. I may try moving to a striped RAID to see if that speeds things up. I've also been wondering if FAT would be faster than NTFS -- doesn't seem like it should be.

[da franksta]

UltraDNA, how big is your virtual mem? CCE uses 1/4 of total memory (i.e. including virtual mem).
Besides, 256Mb seems a bit low. Is your cpu 100% used? Probably not.

Check your system resources. There must be something to hold you back.

Good luck!

[daxab]

Follow up:

Well as I said in an earlier post, I have a nearly identical rig: Non-overclocked XP 1800+, 512 MB DDR, Win2K SP2. I had already installed the 4in1 4.37 drivers and the PCI latency patch.

I don't know why, but today I thought I'd reinstall the 4in1 and the PCI latency patch -- and my encode speed went up to 1.5 to 1.7. (It was around 1.0 before.)

I'm guessing that when I installed SP2 it stomped something that the 4in1 and/or the latency patch was doing.

Anyway, I recommend trying a reinstall of these two items.

[daxab]

Follow up:

Ugh. I take it back. My speed is still 1.0 or so. I just happened to get 1.7 on the VAF pass of one particular movie. It had nothing to do with the 4-in-1 VIA drivers. I'm starting to wonder if FAT32 or if Win98SE would be faster than the Win2K/NTFS setup I have.

[markrb]

All I can add on the FAT32 vs NTFS thing is that in benchmarking fat32 blows away NTFS.

Alas the daunting 4gb file limit makes me keep my one large partition NTFS.

I suspect it will make no difference in encode times in CCE though. CCE doesn't really push a file system that hard. Some of the other programs do, but compared to CCE everything else is just a drop in the bucket in time.

I now get CCE speeds of about 1.45 with my overclocked XP 1700 @ 1.65 Ghz, but I have been using IEEE 1180 reference which according to mrbass is slower, but better quality. I will most likely hit over 1.5 with 32 bit SSE. This is with 16:9 NTSC material though.

You might see some gains in speed if you up your systems RAM over 256mb. Not as much as you see from 128 to 256, but some. However have you seen the cost of PC2100 lately? And I complained about paying $43 for 256mb.

Mark

[daxab]

Another update:

I reformatted my NTFS volume as FAT32 and got an immediate increase in CCE encoding rates, from 1.0 up to 1.5.

However, other operations, like muxing and image authoring, were incredibly slow -- hours instead of minutes. So I took the plunge and "upgraded" to Windows 98 SE.

Now everything is zippy.

I don't really believe that Win98SE performs better than Win2K, in general, but it certainly seems to in my case.

[DDogg]

fast DDR memory and a fast FrontSideBuss speed seem to make a large difference. My machine will run the FSB at 150 without becoming unstable (checked with Prime95 torture test). This seems to make the biggest difference in the RT. I normally get around 1.35 using bicubic on a 480x360 resize. With Tom's new SimpleResize filter I have been getting 1.73

TB 1.4-266A chipset-FSB 150-512 meg ddr 2100-XP-NTFS

[markrb]

DDOGG I know that most people use Prime95 as a test to check system stability and it seems to work ok, but have you tried GoldMemory v5.07 - PC Memory Diagnostic Tests? This tortures your memory subsystem independent of any Windows OS. It will show errors that other tests fail to bring out. While it's main goal is to test the memory by doing this it also tests your CPU, PSU and motherboard in general.
If you get even 1 error you should back things down until you pass 100%.
It isn't perfect and does not push disks, PCI or video, but the majority of errors tend to be memory or CPU related.
I used to use memtest86, but I find this program better.

You can get it here:
www.goldmemory.cz

Give it a try an see what you think.

Mark

[DDogg]

I'll give it a try but I am most concerned with computational errors. This is what screws CCE, IMO. Machines that seem perfectly stable when OC'ed will many times have computaional errors until slowed down a bit. Maybe this memory checker does something similar.

Thxs

[markrb]

It may not show all the weaknesses in a system, but it is a great way to test without exposing Windows to potential corruption. Since this runs in stand alone mode there is no risk of data or OS loss.
I will bet that if this program passes 100% you will find that over 95% of the time Prime95, Seti and all the others will pass just fine as long as it's not an OS issue.
Many of the people I know run Prime95 and Seti at the same time to really toture there system.

Mark