[CSMR]

I posted this on hydrogenaudio too.
I used dbpoweramp to encode in WMA lossless, FLAC, Monkey Audio, and Wavpack.
Compression speeds obtained (AMD XP 2400+):
WMA: 23*
APE extra high: 21*
FLAC medium: 40*
FLAC high: 2.1*
Wavpack high quality: 32*
All quite fast enough, except for FLAC high, which curiously gave compression ratios somewhere between FLAC low and FLAC medium for the two tests I made of it. So FLAC high is out of the tests.

Test 1: Beethoven Piano Sonata no. 1 Allegro, Artur Schnabel, recorded some time between 1932 and 1935
wav: 33,585kb
wma: 11,066
FLAC low; med; high: 12,734; 11,566; 11,918
APE fast; normal; high; extra high: 10,812; 10,608; 10,586; 10,559
Wavpack: 10,834

Test 2: Beethoven string quartet no 1 in F major, adagio, Vegh Quartet, c. 1973
wav: 99696kb
wma: 45,107
FLAC (medium): 46,060
APE (extra high): 43,561
Wavpack (high quality): 45,346

Test 3: Bach Johannes Passion, Gardiner 1986 Choral: Wir hat dich so geschlagen
wav: 15,665
wma: 5,842
FLAC: 6,102
APE: 5,630
Wavpack: 5,786

Test 4: Mozart Symphony no 35 in D, Minuet-Trio, Boehm 1960
wav: 36,521
wma: 16,879
FLAC: 17,223
APE: 16,399
Wavpack: 16,834

The winner: APE. The competition between wavpack and WMA is close, but wavpack takes second place, WMA third, and FLAC loses. This is a clear ranking, with the same positions in all four tests. APE had similar ratios at all settings, and would still have won in fast mode.

 

[ooheadsoo]

I'm not familiar with your "medium/high/etc" scale for flac. AFAIK, it scales from 1 to 8. 8 is definitely screwy. It takes forever on my computer and offers no better performance than 5. Shifting down to 7 puts the encoding speed back into line with the rest of the pack.

 

[CSMR]

These must be dbpoweramp default settings.
In that case I should think that high is 8 and medium is 6 or 7.

 

[Mr.Radar]

The FLAC website has a pretty good codec comparison that isn't very biased.

 

[ooheadsoo]

I'd guess medium is around 5, though I really have no idea. Just guessing from the scale of 1-8. With the flac gui that comes with flac, the default is 5.

Any chance you'll do a cpu usage comparison next?

 

[lan]

Quote:

Originally Posted by ooheadsoo I'm not familiar with your "medium/high/etc" scale for flac. AFAIK, it scales from 1 to 8. 8 is definitely screwy. It takes forever on my computer and offers no better performance than 5. Shifting down to 7 puts the encoding speed back into line with the rest of the pack.

Well any of the lossless compresssors start to take infinitly longer the higher the setting. FLAC8 isn't so bad for me. What computer are you on?

 

[ooheadsoo]

I'm on a 2400+. Was clocked to 2900+ or so but I clocked it back down to reduce fan noise. It takes about 4 minutes or so (I'm guessing) with flac8. With flac7, it's done in like ~2 minutes. These are all rough estimates.

I guess I should mention that these are for a full album, not 1 song

 

[Edwood]

You're testing just encoding speeds, right?

Have you tried to subjectively listen and measure the difference between the different compression formats?

I, for the life of me, cannot hear the difference between FLAC and APE's.

But I like APE's better since I like being able to rip and compress with one mouse click with EAC.

-Ed

 

[Glassman]

Ed, seems like you've gone mad! or you're drunk once again

[hint]lossless means no losses at all, nobody can hear any difference, not even a single hamster[/hint]

 

[CSMR]

Quote:

Originally Posted by Mr.Radar The FLAC website has a pretty good codec comparison that isn't very biased.

Oh, didn't realise that. At least it doesn't include wma, so there's something for my effort! And it could have been that classical music tests would give different results - not as it turned out. I think wma will become a popular format as ordinary non-techy people start using lossless instead of mp3.

 

[zum]

Quote:

Originally Posted by Edwood Have you tried to subjectively listen and measure the difference between the different compression formats?

You could even do an objective comparison instead of subjective by extracting the files to WAVs and then using a tool like WinDiff to compare them to the originals. (Actually, there might be differences in the WAV header, so use some tool to lop off the header first.)

If you want to compare how they sound, you'll have to compare different playback engines. FLAC/foobar2000, FLAC/winamp, APE/xmms, etc. Who knows, they MAY matter, if not because of an actual difference then perhaps because of the placebo effect. (Someone mentioned that he found WAV to be a better format than FLAC, foobar2000 to sound better than winamp, and someone else probably thinks the opposite)

If you go this far, you'll have to take into account the operating system, speed of the computer, soundcard... but indeed who knows they may have some very minor effect

Quote:

But I like APE's better since I like being able to rip and compress with one mouse click with EAC.

No problem doing that with FLAC as well, provided that EAC is configured...

 

[MikoLayer]

FLAC is supposed to take much lesser resources for playback right? i wouldnt mind giving up a tad bit more space for better performance...

 

[Edwood]

Quote:

Originally Posted by zum No problem doing that with FLAC as well, provided that EAC is configured...

LOL, it would be that obviously easy, wouldn't it?

Thanks.

-Ed

 

[Mr.Radar]

Quote:

Originally Posted by MikoLayer FLAC is supposed to take much lesser resources for playback right? i wouldnt mind giving up a tad bit more space for better performance...

Yes, FLAC is very fast requiring only integer calculations to decode which makes it perfect for playing back on slower computers and/or CPUs with crappy FPUs (like Via's C3 series). FLAC and WavePak are also a completely open formats so anyone is open to write their own implimentation while APE/Monkey's Audio is only semi-open and WMA Lossless is completely closed.

 

[Dixie Flatline]

Quote:

Originally Posted by MikoLayer FLAC is supposed to take much lesser resources for playback right? i wouldnt mind giving up a tad bit more space for better performance...

This is true -- if you look at the comparison at the FLAC sourceforge site, you'll see that -Q 8 (max compression) decodes in exactly the same time as -Q 5 (typical compression), or even slightly less. The design intent of FLAC, as I understand it, is that it's asymmetric in favor of decompression speed, and it's designed to be decompressed using strictly integer math (no floating point hardware required, and therefore well suited for portable/embedded devices). This is one reason why it's the only one of the lossless compressed codecs currently supported on a portable (Rio Karma). So yeah, if you're trying to play back lossless audio on a hardware-limited platform, it's a good choice.