mitiok's LAME compiles for BeSweet

[sillKotscha]

Hi folks,

I've made a few simple tests comparing BeSweet 1.3RC vs. BeSweet 1.3b12. As you will see, I've used different types of lame concerning common Cmdlines. There is just a little difference to the DVD2DivX(5.02) doom9-Guide... instead of using --alt-preset 128, I'm using --alt-preset 96 --resample 48.
it sounds equal (IMHO even better, but that's subjective) to the proposed setting but saves space for the video... unfortunately it takes a bit longer ;)


TheWEF-Gordian Mod. (18th May 2002 21:37) - Topic: @theWEF...concerning BeSweet


BeSweet is faster if you use the same settings.
i'd say ~10-20% faster.

but if you are setting -alt -preset stuff encoding takes quite a while, even if you use BeSweet.

higher quality - longer encoding times...


regards Sill

P.S.:

Attachemnts must be validated by a moderator before they become visible

... I hope so ;)

[DSPguru]

the well known mitiok (http://home.pi.be/~mk442837/) had released his own compile of BeSweet's lame_enc.dll (v3.93).

please test and report feedbacks :D.
Dg.

[lark]

@sillKotscha
thanks for great comparison.
could you please let us know your HW setup?

btw why is
--alt-preset 96 --resample 48
NOT advisable for v.1.26 engine 3.91

regards
t :)

[frank]

than BeSweet's one.

Tested with SSRC and --alt-preset 128.

[DSPguru]

mitiok rulez !!

[sillKotscha]

I can't give you any exact comparisons by now, but as far as I can say... it is really fast!!

as Dg said mitiok rulez !!


btw why is

quote:
------------------------------
--alt-preset 96 --resample 48
------------------------------


NOT advisable for v.1.26 engine 3.91


...look at the final file size and as you can see it is something like a MP3 but not a real VBR one. The engine 3.91 wasn't really capable of handling VBR MP3s.

cheers Sill

[cofferscuffs]

deleted.

[mitiok]

mitiok certainly rulez, but as i see you are trying to compare two different version of lame, 3.92 and 3.93alpha. that's not very interesting for me....lame was changed during this time, you know....

could you test please

http://mitiok.free.fr/dll-20020625.zip
and
http://mitiok.free.fr/dll-20020628.zip
?

these dll's are both default compiles, dll-20020628.zip was compiled using icl4.5, dll-20020625.zip - icl6.0

just wanted to know which version of icl is better. somebody told me that icl6.0 is faster for pentium4, as i see icl4.5 is faster on my p3, so i don't know what to use know

thanks

edited (29 June): 28 - icl 4.5 and 25 - icl 6.0

[vinetu]

Hi mitiok !

some tests between dll's:

test WAV file - 1h 09min , 44100kHz mono
GUI - CDex v. 1.4
bitrate -CBR 160kbps,quality - q5(normal)

"dll-20020625.zip" is reported by CDex as "Lame MP3 Encoder v 128 engine 3.93 ..."
"dll-20020628.zip" ... as "Lame MP3 Encoder v 129 engine 3.93 ..."


Time to encode:

PIII 800 MHz (Dual CPU,2GB RAM,win2k) :
"dll-20020625.zip" - 6min 02sec
"dll-20020628.zip" - 5min 14sec


Celeron Tualatin 1.25GHz (overclocked from 1.1GHz,512 Mb RAM,winXP)
"dll-20020625.zip" - 3min 46sec
"dll-20020628.zip" - 3min 09sec


P4 1.5GHz (overclocked from 1.4GHz,1GB RDRAM,win2k)
"dll-20020625.zip" - 5min 15sec
"dll-20020628.zip" - 4min 02sec

Best Regards !

[mitiok]

Originally posted by vinetu
test WAV file - 1h 09min , 44100kHz mono
GUI - CDex v. 1.4
bitrate -CBR 160kbps,quality - q5(normal)


thanks a lot. could you make also tests with alt-preset.


"dll-20020625.zip" is reported by CDex as "Lame MP3 Encoder v 128 engine 3.93 ..."
"dll-20020628.zip" ... as "Lame MP3 Encoder v 129 engine 3.93 ..."

yeah, but there's no much difference as i know


P4 1.5GHz (overclocked from 1.4GHz,1GB RDRAM,win2k)
"dll-20020625.zip" - 5min 15sec
"dll-20020628.zip" - 4min 02sec


and even here icl 4.5 is better... :(

thanks a lot!

[Rodger]

I tested this "speed boost" by using the lame-DLL in CDex.
There is no speed improvement between org. version and this 682!

They both took 49sec. for the same file on my AthlonXP @1610Mhz.
As far as I know there can`t be any speed improvement for BeSweet to then.

Rodger

[vinetu]

Hi man!

...and the new "alt setings":
WAV lenght - 1h 9m 35s - 44100, STEREO
CDex 1.4
bitrate - VBR-NEW , min 96kbps - max 128kbps ,quality -q2(high),VBR quality-VBR 0,forced stereo

CeleronTualatin 1.25 GHz:
dll-20020625.zip - 8min 58sec
dll-20020628.zip - 8min 20sec

P4 1.5GHz:
dll-20020625.zip - 11min 30sec
dll-20020628.zip - 10min 08sec

PIII - not available at this time but the Celeron core is
almost the same (coopermine - plus-minus some litle things) .


BEST
REGARDS !

vinetu

[Rodger]

Okay...so many results can`t be wrong so here is my second test:

Testfile: Wav (102:59)

Lame V3.92 (V1.28) 8min41sec.
Output-Filesize:106MB

Lame V3.93 (V1.29 Alpha2 / 618) 8min39sec (99,62%)
Lame V3.93 (V1.29 Alpha2 / 625) 8min35sec (98,84%)
Lame V3.93 (V1.29 Alpha2 / BeSweet) 8min26sec (97,12%)
Lame V3.93 (V1.29 Alpha2 / 628) 8min80sec !!!!!
Output-Filesize:121MB (All of them!)

My System:
Epox 8Kha+ (KT266A)
AMD Athlon MP @1608Mhz
512MB DDR-Ram

So, on my system there isn`t that boost but it´s faster!
But why is the output filesize bigger???

greetz
Rodger

[DSPguru]

there's a new compile (http://home.pi.be/~mk442837/), now made with ICL 4.5 .

time to compare ICL 4.5 vs ICL 6.0 ..