Back Forum Reply New

New setup kicking serious Butt

After much testing and working on my new system I now have a DVD2SVCD butt kicking machine.

This is from an AMD system with the following spec.

XP2100+ @ 2.47 Ghz
Water Cooled
512MB PC3200 @ 197FSB

Fully tested with Prime95 and Memtest.

JPG of CCE encode of a CBR at 5500 Kbit for DVD-gt;DVD conversion of an NTSC Force Film source.

I never did expect this type of speed.  It is not an error as I did an earlier encode that came out fine.

After posting the pic the speed is now up to 2.537.
Switched to VBR and with the same settings got up to 2.648.

Mark

And your rubbing this in, because? hehe  I need to upgrade

Not really rubbing it in.
More of really amazed.
When I first started with DVD2SVCD I was begging to get .73 and forget it if I had to use IVTC.
I was very happy doing 20 hour encodes and not being able to use my computer at all during this time.

I can't wait for 5.0.  I guess I will have to wait for AMD 4000+.

Mark

Hey, that's a BIG speed difference. I guess you have the biggest one on the block now

i am getting about 2.84 in cce so far in the conversion of pal dvd -gt; dvd.
system is an abit nf7-s with an xp 1700+ running as an 3300+ with 2,53 ghz real. 11,5x220regards

eternal

Hi-

I'm curious. Are these speeds faster than doing DVD-gt;SVCD, maybe because there's no resizing or other filtering involved? Or are they slower, maybe because fewer pixels are encoded when making SVCD with its lower horizontal resolution? Or are they pretty much the same, maybe because the two pretty much cancel each other out? In any event, those speeds are kickass.

Damn Mark! Give us the real scoop on your hardware. What motherboard, HDs, brand of water-cooling system, etc.

-=lt;halcyongt;=-

Yeah, it might be fun to have one OT thread that isn't the same old same old stuff.

It's nothing super special.

FIC AU11 (exact copy of Epox 8RDA+) Nforce2 board.

AMD XP 2100+ T'bred B. AUIHB @ 2.47.  Can do some more, but tops out around 2.5Ghz.  Decided to back it off a little to assure good longterm stability.

Memory is Twinmos PC3200.  Good value.  Many think it is almost as good as Corsair at a much lower cost.

FSB @ 197 Dual Channel.  Needed to do a mod on the board to get this high.  Can do more, but backed it off for the same reason as the CPU.

Water cooling mostly home made.  Water Block is the Spir@l from Dangerden.  Should have bought the TC-4, but this is still a good block.
Pump Sen 500.  87 Camaro Heater Core with a 192MM fan.
Most of the setup is external to my case.  Not a real pretty site, but it's my private office so I don't care.  It is also much quieter then my old Alpha 8045 80MM heatsink.

@eternal nice.  Do you have one of those famous 1.5V 1700's?

Mark

yep exactly, got a few of those famous jiuhb with tbred core and 1,5v

abit nf7-s
1700xp+ jihub 0308 xpmw @ 2,53 ghz
2x 256MB corsair pc3200LL running sync
water cooled

regards

eternal

Hi-

I decided to try and answer my own question. It turns out that encoding for DVD, without resizing and adding borders as you would when making SVCDs, is considerably faster. So I also tried a 1 Pass CBR CCE encode @ 5500 Kbit of a 16:9 anamorphic widescreen DVD. Also Force Filmed. Also at DVD resolution.

P4 2.66 @ 3.166GHz
Stock Intel Heatsink/Fan
2x512 PC3500 @ 211

Couldn't quite reach eternal's speeds, though:

manono, good point. Mark, could you do the same encode above for a few minutes with bicubic resize and addborders? That would give the svcd'ers an apples to apples instead of an apples to oranges.

I've a similar setup and I can get 3.4 with simple risizing 4 passes.

Hi-

I can get 3.4 with simple risizing 4 passes.

I can get that when converting .avis to SVCD, but not DVD-gt;SVCD. If you're getting those speeds with DVD-gt;SVCD, I guess I'll have to go back to tweaking some more.  Or maybe you're making a Huffy.avi first?

I attached a .jpg to the last post, but it hasn't been approved yet. I think it said 2.704 RT. And I tested with a 16:9 DVD. If I used a 4:3 widescreen DVD, of course the speeds would be faster.

Edit: Using AviSynth 2.51 and changing to quot;iDCT_Algorithm=5quot; in the .d2v.

@all
So, Avisynth 2.07,2.08 or 2.51? With or without idct=x tweak? showthread.php?s=amp;threadid=51028

Just did as you asked, but with a different movie.  So the results won't be 100% the same.

Using FF, Avisynth 2.51, bicubic resize of 480 x 480 on a 16:9 anamorphic NTSC source.  

I got a CCE speed of 2.3.  A bit slower then the 2.648 I got with no resize on a different movie.

I also used the defualt mpeg2dec.dll.  I didn't swap whatever one DVD2SVCD uses so maybe I could have been faster had I used a different one.

Mark

No, dvd2svcd is using mpeg2dec3 when set up for 2.51 so you should be as fast as possible with that setup. Only thing to tweak is adding the idct parameter to your mpegsource(quot;stuff.d2vquot;,idct=2). You may have already been using this but you did not say one way or the other.

It turns out I had not changed a setting or two to fit what the average DVD-gt;SVCD conversion would have.

After changing things closer to the defaults I now get 2.773 with Bicubic, FF, 480 x 480, 16:9, Avisynth 2.51, mpeg2dec3.dll.

I did try the tweak settings mentioned of idct=2 for my AMD, but it had no effect.  I tried both in the source line and in a .DEF file with the dll.

When I get more time I will do the same movie at as close to the same time as both an SVCD conversion and DVD conversion and then truly see a head to head comparison.  I can't see how resizing could increase the speed.  So something must be screwy.

Mark
¥
Back Forum Reply New