My CPU is...

[EddyH]

Odd with the P4s.... I looked up a number of benchmarks to see what kind of acceleration factor I would get outside of plain Mhz increase moving from this Duron (in the perfect world, where I could afford to get a 3.4ghz with a snap of the fingers). Yes, even video conversion was covered.

Turns out the P4's performance per mhz is almost dead-on with that of the Duron, within 1 or 2%... of course it varied a little from test to test and arena to arena, but it was never a phenomenal, blow-you-away difference (you know... 10% or more). That is to say - the speeds of P4 and Duron didn't quite cross over, but if you were to do the maths and ie make a Duron 3400mhz it would be as quick as intel's finest, or a P4 0.85ghz would be no faster than my current box. Quite a surprise and very disappointing!
And the Athlon gently outpaced the Duron (depending on task), and the AthlonXP outpaced the earlier flavours.... so the XP walks all over the P4. If AMD can get their act together and push the AthlonXP speeds maybe towards the 4ghz equivalency rather than the current mid-2's (where intel currently shines is their clock speeds, the new 3.4ghz will beat any sub-3000 XP), i'm definately sticking.

:/

Going to hang on probably until something the equivalent of a 4.25ghz Duron rolls along though.. yknow... to get that essential 5x power boost. 85m videos that once had to be left overnight suddenly coming through in better than real time, aw yeah.


And yeah, I am trolling the polls tonight.... i'm bored

[Questor]

Eddy H - I have a Pentium 2.5 (not overclocked)- and I get my movies* coming in better than realtime.


*90 minute Divx - VCD in about 75 minutes
*90 minute Divx - DVD in about 3 hours

And yeah, I am trolling the polls tonight.... i'm bored also

[racer-x]

Interesting poll. I use AMD XP1700. What about setting up a benchmark. We could see how long it takes to encode an AVI - Mpeg2 DVD and report results. For simplicity sake we could use a 1 minute AVI and encode it in Tempenc. We could use same sttings like NTSC DVD, no filters, best frame search, and priority set to high. Then report how long it took to encode. Also report CPU, RAM type,O.S., and Direct-X version ect.

[turbosteve]

Recent Intel P4 2.53 convert here from my AMD 1800+, decoding times aren ight and day difference. P4 does it in half the time sometimes even less. Also have a Tbird 1200 24/7 server doin decodes.

Steve

[EddyH]

Ahh, but Questor, what settings are you using?

Hmmm, user-supplied benchmarks might be a good one, poll everyone on VCDHelp - and ask for serious answers for once

The numbers I'd personally like to see are for TMPGEnc, MPG1... For the sake of argument, a 120m00s (makes for easy maths!), 640x304 25fps source Divx, with 44.1 128kbit CBR MP3 sound (such differences can actually affect your time, seeing as DivX is somewhat CPU heavy, especially near-DVD rez on a sub-ghz system), converting to letterboxed 4:3 PAL, 192 vertical pixels in the 288 frame. Standardised source if possible, some easily available / hostable / everywhere-on-kazaa movie

((EDDYH DOES NOT IN ANY WAY CONDONE PIRACY AND NATURALLY HIS WORDS DO NOT REFLECT THE VIEWS OF VCDHELP!! The movie would consist of a travelogue film created by yours truly with a digital camcorder alternately mounted to the car dashboard or a head helmet and spread through the internet by spiderbots...))

Those who think their system may take a long time over this will be allowed to use a representative 60, 30, 15 or 7.5 minute sample. Anyone whose system is too blazing fast to get an accurate result from the 2-hour movie should do it twice in succession and halve the total, and then be attacked from behind whilst distracted by this task and handcuffed to the bed with a sack over their upper body as their computer is "temporarily" removed for "evaluation".

First, around CBR-1150 standard with motion estimate and no filters, but Floating point DCT, scene detect and moderate block smoothing turned on, and internal 224-stereo-crc audio, muxing to vanilla VCD on the fly (test run/quick convert).
Second, CQ encode of approx 75, 0/20 spoilage, 2400/48 max/min, maximum search precision, moderate cropping, and custom colour correction with two steps plugged in. Audio ripped out to WAV in Virtualdub (time recorded) and encoded to 160 and 192k joint stereo and 224 'normal' stereo in an(y) external encoder at the same time as doing the video, followed up by a muxing speed test using the 192k soundfile to non-standard VCD.
And a viewing/listen-test to see whether it was "worth it" or not
(Especially as the CQ would likely require those numbers multiplied by about 1.75x to account for a second video-only run in the majority of realworld cases!)

In both cases, one run in the 'overnight' setting (most system tray items closed down and a couple behind-the-scenes things C-A-D'd out of existance; video encode priotity set to high whilst, where applicable, the three audio tracks go simultaneously behind it; no other programs run and screensaver disabled), and then repeated in the 'background' style. All of the default nonsense windows loads left in place, a couple of other low-load programs in the foreground "doing stuff" (eg, excel with a macro that auto-recalcs a simple sheet every couple minutes, behind a copy of word with a weight left on one of the letter keys). Video encoder priotity set to normal, whilst the sound files (for test type 2) are encoded one at a time - which gives the multimedia just that little less CPU share from the WP than doing them all at once, but de-stresses the HD and RAM.

And after one or the other, the resulting pair of video files dropped into VCDEasy / Nero, and a stopwatch used to see how long it takes for them to read and process the file - exclusive of any user interaction - to a point where all you then have to do is hit the BURN button.

Using extreme maths voodoo all those numbers can then be mushed together, and some kind of highly improbably statistical evidence can be produced from the mess showing that, yes, my 850mhz duron will whip your 2ghz P4... hehehe... :P


Oh the things I write when waiting for cooledit to finish off with the compensating-for-30.000fps-xvid-that-thinks-its-23.976 conversion, and getting started on the next stage of the project is just too daunting...

[i-on-s]

Was running dual amd 1.5ghz t-birds but i got a new pc using pentium 4 processor 3.06ghz, 533fsb, 512k cache, intel gigabit nic and i also rock out on my p1 133mhz dell with 32mb of ram and a sony vaio 600mhz with 512mb of ram. i am still amazed that my 133 still runs smooth since 1993 good ole dell's

[EddyH]

i-on-s
Wowzers what kind of incredible conversion time do you get on that Dell beauty? It's gotta be under three days, surely... a new record


/can still remember when 133mhz seemed like an insanely fast upgrade\

[i-on-s]

well lets just say a 5min file would take 6hrs never tried a full movie or would dare to try one also. i remember the day i got it...thought i was bad shit with my 2x cd burner and my very own copies of doom 1 and duke nukem 3d

[SLICK RICK]

AMD ATHLON XP

[EddyH]

Hehe...
I got Quake demo and DN3D working "playably" on a DX2/66.. thought I was so badass
Ah the days of having a measurable wait to copy a 10mb file..

They don't make them like they used to, thank goodness.

[LanEvo7]

I use an Amd 1700+. Depending on the settings I choose for encoding, I can encode faster than real time on tmpgenc. For a regular VCD file with low motion searh precision, i do it on about 3/4 real time. If normal, then its a little less than real time. If its high, then it takes about 3x real time.

[Donny661]

I just got my old dx2/66 working again so my mom could do some word proccessing on it. Ahhhh, windows 3.1 and ms-dos, good stuff. It'd be fun to try encoding with tmpg on it; it couldn't even play mp3's when I had win 95 on it, but I haven't tried in dos.

[EddyH]

Donny... apparently MP2s will just about work for cd quality in realtime on a 66 (probably with a DOS player only)... easy to make in TMPGEnc or Cooledit, or just rip them from a music MPG. Certainly will be trying it on the 75mhz tpad once it's essentials are soldered back together - 160k joint stereo with a moderate lowpass filter sounds about as "good" as a 128k mp3. Or mono 32-44khz mp3 might work, 22-24khz stereo.
Dunno, how far did the winamp guys get with optimising their decoder before dropping all semblance of 486 compatibility?

Otherwise, to get retro, your best bet is 16khz mono WAVs using microsoft ADPCM 4-bit compression or IMA ADPCM 3-bit. Sounds terrible at first... but works and you get used to it. The best way to show off computer leetness in 1996 was getting an entire "radio quality" *real* song - not midi - onto one floppy disc... now everyone's doing it

Had a pocket full of floppy discs with a song each at one time in the lower end of high school..

[Localboy8o8]

P4 Power baby

Would own nothing BUT an INTEL CPU ...Motherboard, another story...But that can be the next poll...LOL

gotta supprt the product that i work at.

[EddyH]

Low motion search precision?

bleah.... block city/lots of VBR rate max-outs.

Dear lord, man, have you no shame? At the very least use motion estimate or Normal. That little extra wait is worth it, particularly if even Very High only takes 3x (high vs v.high doesnt seem to make much diff eye-wise or bitwise, apart from crashing on some quicktimes or MJPGs, but if you put a couple of divxs on to batch overnight it's good to be sure