LAME 3.99 is out, 2012-02-28: version 3.99.5 has been released Options

[psycho]

My results (samples taken from http://ff123.net/samples.html):

My CPU: AMD Athlon 64 X2 4000+ (Brisbane)
Endodes done with command line: -V 0

Rarewares 3.99 build:

LisztBMinor sample:
15.219x - 246.6 kbps

ATrain sample:
14.614x - 256.3 kbps

BachS1007 sample:
15.988x - 229.8 kbps

DaFunk sample:
14.185x - 287.6 kbps

ExitMusic sample:
15.921x - 250.4 kbps


My 3.99 build (MinGW 5.1.6 with MSYS-1.0.11 with yasm-1.1.0-win32.exe for nasm.exe):

LisztBMinor sample:
16.363x - 242.7 kbps

ATrain sample:
15.729x - 254.0 kbps

BachS1007 sample:
16.583x - 227.2 kbps

DaFunk sample:
15.520x - 285.6 kbps

ExitMusic sample:
15.909x - 250.1


It's curious to see how my build is faster and produces smaller mp3 files...

[psycho]

I forgot to mention:

OS: WinXP SP3 32-bit


Can't edit my previous post, so I will post my next test results here:

Here are the results with 3.98.4:

Rarewares 3.98.4 build:

LisztBMinor sample:
18.444x - 217.9 kbps

ExitMusic sample:
18.947x - 225.1 kbps


http://lame.bakerweb.biz/ 3.98.4 build:

LisztBMinor sample:
14.077x - 217.9 kbps

ExitMusic sample:
14.648x - 224.1 kbps


My 3.98.4 build (MinGW 5.1.6 with MSYS-1.0.11 with yasm-1.1.0-win32.exe for nasm.exe):

LisztBMinor sample:
15.770x - 217.9 kbps

ExitMusic sample:
16.742x - 224.6 kbps


Question for john33: I have Microsoft Visual C++ 6.0 - how can I compile lame with it?

This post has been edited by psycho: Oct 26 2011, 19:57

[Alex B]

Rarewares 3.98.4 build:

LisztBMinor sample:
18.444x - 217.9 kbps

ExitMusic sample:
18.947x - 225.1 kbps

Which Rarewares build? The old "main" compile:

From: http://web.archive.org/web/20110722042821/...lame-bundle.php

LAME 3.98.4
2010-03-23

Bundle compiled with Intel Compiler 11.1.
Download (563kB)

... or the currently available "VC6/Intel Compiler 9.1" compile:

From: http://www.rarewares.org/mp3-lame-bundle.php

LAME 3.98.4
2009-06-06

Bundle compiled with VC6/Intel Compiler 9.1 - intended for older Windows OSs.
Download (494kB)

I used the "Intel Compiler 11.1" version in my 3.98.4 test.

This post has been edited by Alex B: Oct 26 2011, 21:00

[robert]

Question for john33: I have Microsoft Visual C++ 6.0 - how can I compile lame with it?

With LAME sources comes a Makefile.MSVC, I used that to build LAME from the command line using VC6, back in the days.

CODE

lame v399: -V0 --noreplaygain LoveSexy.wav

m:ss    x    kbps    

2:06    21.3    260.8    gcc4.1.2    x86/makefile.unix CFG=RH_SSE

2:42    16.617    257.5    VC9 Express    x86/Release
2:02    22.099    259.8    VC9 Express    x86/ReleaseSSE

2:23    18.790    256.3    VC11 D-Prev    x86/Release
1:52    23.925    256.3    VC11 D-Prev    x86/ReleaseSSE

2:02    22.152    256.3    VC11 D-Prev    x64/Release
1:45    25.689    256.3    VC11 D-Prev    x64/ReleaseSSE

machine: Athlon 64x2 4000+ Brisbane @2.4GHz 8GB Win7 64bit / Suse 10.2 x86

[psycho]

Alex B, I used the "VC6/Intel Compiler 9.1" compile of Rarewares' 3.98.4.

john33, thanks, I'll try Makefile.MSVC.

[Alex B]

Alex B, I used the "VC6/Intel Compiler 9.1" compile of Rarewares' 3.98.4.

When I tested the 3.98.4 compiles in early October "VC6/Intel Compiler 9.1" was sligthly slower than "Intel Compiler 11.1" -- but perhaps not insignificantly because on average the speed difference was constantly reproducible (I ran the tests several times):

3.98.4 (the main bundle from Rarewares)
Total encoding time: 0:40.922, 134.34x realtime

3.98.4 (the VC6 compile from Rarewares)
Total encoding time: 0:41.625, 132.07x realtime

I really would not like to sacrifice any of the encoding speed just because of the new version is out. In the past a slightly improved encoding speed has often been a good reason to upgrade to the new version. 3.98.4 has been a fine LAME version and I may continue to use it (at least for now).

This post has been edited by Alex B: Oct 26 2011, 21:55

[Alex B]

Results of "lame -S -V2 --noreplaygain test.wav nul" on Intel Core2: encoding time in seconds

Rarewares compiles:
x32: 148 s

My compiles (MSVS 2010, fast math + SSE enabled):
x32 NASM: 125 s
x32 SSE2: 135 s

Also, Rarewares Lame 3.98.4:
x32: 130 s

Your 32-bit 3.99 compiles are interesting. The NASM compile is actually faster than the Rarewares 3.98.4 version. What does "NASM" mean in this case?

This post has been edited by Alex B: Oct 26 2011, 22:11

[psycho]

john33, I managed to compile lame.exe with VC6, however it doesn't encode.

If I just run lame.exe with no command line options, it displays the version info normally, however if I want to encode with it, I get an error:



Well... I'm ready to give up... just wanted to see if VC6 compile will be faster on my system... If someone is willing to compile 3.99 with VC6, I'd be happy to test it.

[robert]

Your 32-bit 3.99 compiles are interesting. The NASM compile is actually faster than the Rarewares 3.98.4 version. What does "NASM" mean in this case?

NASM compile uses "Netwide Assembler" code for some routines. Btw., using the very same VC9 and compiling 3.99 and 3.98.4 (ReleaseNASM configuration), there is not much of an speed difference:

CODE

LAME 3.98.4 32bits (http://www.mp3dev.org/)
CPU features: MMX (ASM used), 3DNow! (ASM used), SSE (ASM used), SSE2
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding w:\cd\Prince\1988-Lovesexy\01 Songs are in a continous sequence.wav
      to x.mp3
Encoding as 44.1 kHz j-stereo MPEG-1 Layer III VBR(q=2)
    Frame          |  CPU time/estim | REAL time/estim | play/CPU |    ETA
103492/103492(100%)|    2:22/    2:22|    2:22/    2:22|   18.913x|    0:00
32 [     0]
40 [     1] %
48 [     0]
56 [     1] %
64 [     0]
80 [     4] *
96 [   115] *
112 [  1580] %**
128 [  7331] %************
160 [ 40039] %%%***************************************************************
192 [ 33304] %%%%%%%%%%%********************************************
224 [ 11836] %%%%%***************
256 [  5489] %%********
320 [  3792] %%*****
-------------------------------------------------------------------------------
   kbps        LR    MS  %     long switch short %
  185.5       12.5  87.5        88.0   6.4   5.6
Writing LAME Tag...done

LAME 3.99 32bits (http://lame.sf.net)
CPU features: MMX (ASM used), 3DNow! (ASM used), SSE (ASM used), SSE2
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding w:\cd\Prince\1988-Lovesexy\01 Songs are in a continous sequence.wav
      to x.mp3
Encoding as 44.1 kHz j-stereo MPEG-1 Layer III VBR(q=2)
    Frame          |  CPU time/estim | REAL time/estim | play/CPU |    ETA
103492/103492(100%)|    2:20/    2:20|    2:20/    2:20|   19.266x|    0:00
32 [     1] %
40 [     0]
48 [     0]
56 [     0]
64 [     0]
80 [     0]
96 [     2] %
112 [    81] %
128 [  5400] %********
160 [ 43295] %%%%%%************************************************************
192 [ 37019] %%%%%%%%%%%%%%%%%****************************************
224 [  8887] %%%%%*********
256 [  6608] %%%********
320 [  2199] %%**
-------------------------------------------------------------------------------
   kbps        LR    MS  %     long switch short %
  184.8       19.1  80.9        88.0   6.4   5.6
Writing LAME Tag...done

CODE

LAME 3.98.4 32bits (http://www.mp3dev.org/)
CPU features: MMX (ASM used), 3DNow! (ASM used), SSE (ASM used), SSE2
Using polyphase lowpass filter, transition band: 16538 Hz - 17071 Hz
Encoding w:\cd\Prince\1988-Lovesexy\01 Songs are in a continous sequence.wav
      to x.mp3
Encoding as 44.1 kHz j-stereo MPEG-1 Layer III (11x) 128 kbps qval=3
    Frame          |  CPU time/estim | REAL time/estim | play/CPU |    ETA
103492/103492(100%)|    3:17/    3:17|    3:17/    3:17|   13.677x|    0:00
-------------------------------------------------------------------------------
   kbps        LR    MS  %     long switch short %
  128.0        7.8  92.2        92.9   4.1   2.9
Writing LAME Tag...done

LAME 3.99 32bits (http://lame.sf.net)
CPU features: MMX (ASM used), 3DNow! (ASM used), SSE (ASM used), SSE2
Using polyphase lowpass filter, transition band: 16538 Hz - 17071 Hz
Encoding w:\cd\Prince\1988-Lovesexy\01 Songs are in a continous sequence.wav
      to x.mp3
Encoding as 44.1 kHz j-stereo MPEG-1 Layer III (11x) 128 kbps qval=3
    Frame          |  CPU time/estim | REAL time/estim | play/CPU |    ETA
103492/103492(100%)|    2:58/    2:58|    2:58/    2:58|   15.177x|    0:00
-------------------------------------------------------------------------------
   kbps        LR    MS  %     long switch short %
  128.0        8.5  91.5        92.6   4.1   3.3
Writing LAME Tag...done

[krafty]

No one is going to explain the lowpass thing?

Robert can you give any insight on that...

This post has been edited by krafty: Oct 29 2011, 18:12

[ShotCaller]

Considering -V 0 now uses no lowpass filter, would it make sense to disable the lowpass filter for my CBR 320 encodes via the command "--lowpass -1"? Also, does anyone know if the "--buffer-constraint" command is still useful?

[/mnt]

As a result I see a certain progress with 3.99. The spot at sec. 3.0 of eig shouldn't be overestimated. For judging pre-echo behavior we hopefully get other listening test results from members who are more concerned about pre-echo problems than I am.

I have done a quick test with LAME 3.99 against 3.97 at 320kbps, since new tunings have been introduced on the CBR modes.

Everything Is Green - sounds worse on LAME 3.99, an easy to spot artifact appears at the start on the 3.99 encode, plus the pre-echo artifacts throughout the sample is more noticable to me.

Musique Non Stop - sounds worse on LAME 3.99, really bad disortion at the 0:04 mark.

That's What I Get - sounds better on LAME 3.99, less smearing on the synth.

Homecomputer and Show Me You Spine - sounds both equally bad on both encodes.

[chichazor]

¿There is a regression in sound quality with -V2? A made a quick test with the album Soundchaser of Rage, and with the 3.99 the bitrates are 20-30 kbps lower than the 3.98.4 and I can abx without problems :S

[DARcode]

Is there a reson why the old "main" 2010-03-23 Intel Compiler 11.1 LAME 3.98.4 compile has been removed from RareWares please?

[john33]

Is there a reson why the old "main" 2010-03-23 Intel Compiler 11.1 LAME 3.98.4 compile has been removed from RareWares please?

Yes! I have always removed the previous released version when a new release becomes available.

If you require the older release, you can d/l it here:

http://homepage.ntlworld.com/jfe1205/LAME/lame3.98.4.zip

[DARcode]

Thank you John, I was actually looking the the LAME 3.98.4 64bit 2010-06-08 build please, sorry for the confusion I quoted Alex B to be quicker.
EDIT 2: Got it modding your link, thanks.

Is version 3.99 (I know it's beta 1 renamed) the HA recommended one already?

Anyway, why leave the 3.98.4 VC6/Intel Compiler 9.1 compile instead of the Intel Compiler 11.1 one available on RW? More universally compatible?

EDIT 1: grammar.

This post has been edited by DARcode: Oct 31 2011, 13:19

[john33]

http://homepage.ntlworld.com/jfe1205/LAME/lame3.98.4-64.zip

There you go!

I'm not sure whether the concensus is that 3.99 is better, or not, as yet!

[halb27]

I have done a quick test with LAME 3.99 against 3.97 at 320kbps, since new tunings have been introduced on the CBR modes....

Since several Lame versions there are some question marks about improvements with CBR. Do you mind testing -V0?

This post has been edited by halb27: Oct 31 2011, 20:06

[john33]

New compiles at Rarewares reflecting id3 tagging changes, plus 64 bit libsndfile 1.0.25 compile added.

[GeSomeone]

Thanks John,
this seems to be 3.99.1 beta (or even before that)
One of the files has a release date of next week

[john33]

Thanks John,
this seems to be 3.99.1 beta (or even before that)

Why do you say that?

One of the files has a release date of next week

Doh! Fixed with yet another id3 update!

[sld]

Has anybody tried to compile with the Open64 compiler?

http://www.open64.net/home.html

[GeSomeone]

Thanks John,
this seems to be 3.99.1 beta (or even before that)

Why do you say that?

I looked at the updated history.html, not meaning to cause confusion.

[john33]

Ah, OK, I thought it was perhaps something I had got wrong!!

This post has been edited by db1989: Nov 2 2011, 20:14

Reason for edit: deleting unnecessary full quote of above post

[john33]

So as to be sure that my compiles work on both Intel and AMD CPUs, I've changed my development system to a Phenom II X4 840 with 8GB of Corsair DDR3.

Checking the latest compiles, I also produced VC10 compiles for comparison. The Intel compiles, after running through icc_patch are between 5% and 10% faster on the above system than the VC10 versions.