What FLAC compression level are people using these days? Options

[sgold]

I’m about to re-rip a fair amount of CDs and am thinking of upping the compression level to -8 from my usual -5.

Before I do this I’ve been reading some generally quite old forum posts and occasionally coming across contra-indications for switching to -8

Largely, the few negatives I’ve seen seem to be around compatibility issues and extra time to decode -8 on playback, neither of which I would imagine are relevant with today’s technology?

Before I make a decision on this has anybody experienced or is aware of any negative compatibility or functional deficits of -8 what-so-ever for example across different media players and devices or indeed native to the compression level itself over a lower level of compression?..I'm not talking about sound quality which of course is the same at any level of compression or the time it takes to compress the file. Rather I’m referring to the pure universal functionality and reliability of -8 vs any lower level. What are people mainly compressing at these days? Have folks shifted up the compression to -8 as technology has improved or do some people feel they have more peace of mind at a lower level for whatever reason?

This post has been edited by db1989: Jul 20 2012, 23:13

Reason for edit: updated figures as per request in post #23

[Blueshirt]

What are people mainly compressing at these days? Have folks shifted up the compression to -8 as technology has improved or do some people feel they have more peace of mind at a lower level for whatever reason?

I can only speak for myself, but I have always used -5 compression for my CD rips to FLAC and I see no reason to change it now. I certainly wouldn't re-rip a load of my CDs to -8 compression just because technology has improved! There is nothing wrong with my FLAC files that were compressed at level -5, and it is all going to sound the same anyway. So for me, it if ain't broke don't fix it!

[Soap]

I certainly wouldn't re-rip a load of my CDs to -8 compression just because technology has improved! There is nothing wrong with my FLAC files that were compressed at level -5, and it is all going to sound the same anyway.

Nobody in their right mind would rerip to get -8 FLACs when they have -5 FLACs, they would transcode.

I think we can safely assume the reripping is because OP either didn't do it securely the first time or only ripped to lossy.

[eahm]

-6, best one in my opinion for speed/ratio/performance.

[sgold]

Yes. The CDs i'm re-ripping were amongst the very first i did years ago and with the benefit of hindsight and the cruelty of an obsessive-compulsive mind could have been done more elegantly in terms of set-up, etc. The motivation to re-rip is nothing to do with the compression rate however.

[Porcus]

Nothing but -8 here. And right now my drives are 98 percent full. Wouldn't have made it in -0, that's for sure.

Then, I guess it depends on whether your ripping application converts while ripping. dBpoweramp (my choice) is track-oriented, so it encodes a track while reading the next one. That effectively means that I only have to wait for encoding the last track of each CD, except every now and then when long and short tracks alternate. This on a laptop with a Turion TL56 @ 1800 MHz. If your application first rips the entire album as one image and then converts, then I would probably have ripped to -0 and then reencoded to -8 over night.


(OK, I did automated ripping with a changer, but -8 is still how I do it when I have brought a handful of CDs home.)

[Frank Bicking]

Poll added. Let's see if there are any differences to 2007-2009.

[sgold]

Nice! Many thanks Frank

Personally, my very subjective instinct is that the -5/-6 region just feels safe and robust in every aspect, that's purely psychological i'm sure; but i'm getting the feeling that recently -8 is being increasingly adopted?

[pawelq]

-7. Don't realy know why I chose it. And I have no intention to change, I guess that converting using more aggressive compression would yield a minute benefit.

[skamp]

I ran a benchmark on over 5,000 FLACs and determined the overall gain in disk space of FLAC -8 over FLAC -5 to be exactly 485 MiB for a total of 134 GiB. That's 0.35%, folks. IMO, it's only worth it in situations where the increased encoding time is absorbed by some other task running in parallel, like ripping a CD. Worth transcoding your old FLACs? Total waste of time and electricity.

This post has been edited by skamp: Jul 17 2012, 18:48

[AliceWonder]

I use -8 but I use flac for archival, I don't listen to them, so I figure compress them as much as I can.

[db1989]

Personally, my very subjective instinct is that the -5/-6 region just feels safe and robust in every aspect, that's purely psychological i'm sure

“[S]afe” and “robust” imply what? Superiority of support among players, error resilience, or some other unknown thing? You will need to define your terms before your statement can have any meaning. Which is not to imply that any such differences exist: I highly doubt it.

This post has been edited by db1989: Jul 17 2012, 19:07

[Porcus]

I ran a benchmark on over 5,000 FLACs and determined the overall gain in disk space of FLAC -8 over FLAC -5 to be exactly 485 MiB for a total of 134 GiB.

That's very close to Synthetic Soul's comparison http://synthetic-soul.co.uk/comparison/lossless/index.asp . (Those figures would have saved you a mighty 499 rather than 485 :-o)

However, the difference from -0 is quite significant.
(I think I remember having squeezed some GBs of flac 1.1.lessthanfour @ -5 files down by a couple of percents though.)

This post has been edited by Porcus: Jul 17 2012, 19:34

[Remedial Sound]

- 5... set it 'n forget it!

[IgorC]

The same but old poll. http://www.hydrogenaudio.org/forums/index....showtopic=58731

This post has been edited by IgorC: Jul 18 2012, 02:06

[Soap]

- 8... set it 'n forget it!

I can't measure the difference at the wall in power consumption between encoding (nor playing) -5 and -8. -8 may offer only the most marginal of gains but it appears to cost me nothing.


This post has been edited by Soap: Jul 18 2012, 03:09

[A_Man_Eating_Duc...]

I did a test a while back on 300+ full albums and this is what I came up with



this is mainly rock music though.

http://www.hydrogenaudio.org/forums/index....c=61054&hl=

This post has been edited by A_Man_Eating_Duck: Jul 18 2012, 06:23

[probedb]

-8. It's not like one takes any noticeable length of time longer these days.

[Porcus]

By the way, I also verify after encoding (as dBpoweramp supports that). Not because I think there is any bug in the reference flac.exe which would lead to a an encoded+decoded signal differing from the original, but to increase chances that any issue (HD ...) would be detected.

In Synthetic Soul's test, this step would add 10 percent to the -8 encoding time (and 40 percent to -5).


And for those cases where I reencoded: I do not trust overwrite with --force. Encoded to different target folder, foo_bitcompare, then remove.

[pdq]

I did a test a while back on 300+ full albums and this is what I came up with ...

I notice that the difference between FLAC -5 and -7 is especially small, only 0.06%. This means that if you compressed 1600 CDs with -5, you would be able to compress 1601 with -7 in the same space.

I use -5 because it was the default, and I had no issue with either speed or file size with this setting.

[jensend]

The poll needs options for FLACCL.

The exhaustive model search that FLAC uses for -7 and higher involves a tremendous amount of added work for very little gain. I would never ever bother with -7 or higher on the standard encoder. The developers knew what they were doing when they made -5 the default.

But it appears that this tremendous amount of work is embarrassingly parallel and reasonably well suited for GPU computation. I don't own a GPU with enough power to give a real advantage, but for those who do, FLACCL rather drastically changes the tradeoff between encode speed and compression ratio.

If you're absolutely certain you need that last tiny bit of compression gain and you aren't using FLACCL, you should look at using another format instead. FLAC is not designed to push the extreme limits of lossless compression, it's designed to hit a sweet spot on the tradeoff curve. If you're really willing to endure a ~3x slowdown from flac -5 for little benefit, instead of using flac -8 and getting a negligible improvement you could get a still-small-but-10x-bigger improvement by moving to TAK's insane compression.

[Soap]

If you're really willing to endure a ~3x slowdown from flac -5 for little benefit, instead of using flac -8 and getting a negligible improvement you could get a still-small-but-10x-bigger improvement by moving to TAK's insane compression.

I'm not willing to endure a 3x slowdown for the marginal gains of -8. Yet I use -8.

Thankfully I don't compress to FLAC with a slide rule, and my demands for FLAC files are never mission time critical. My computer has more than enough processing power to perform the operation in the background without affecting what I'm doing in the foreground nor measurably affect the VA pulled by the computer (as measured by my kill-a-watt.)

This post has been edited by Soap: Jul 18 2012, 17:49

[eahm]

Just another simple test (CPU specs: http://i46.tinypic.com/110lytc.png):

Pink Floyd - The Wall (2011 Remaster) WAV 819 MB

-5: Total encoding time 004.836, 1006.87x realtime 444 MB (466,053,562 bytes)

-6: Total encoding time 005.242, 928.89x realtime 444 MB (466,043,822 bytes)

-7: Total encoding time 012.589, 386.78x realtime 444 MB (465,811,133 bytes)

-8: Total encoding time 017.863, 272.58x realtime 443 MB (465,150,733 bytes)


Can a moderator please reset/change my vote to -5 please? I did few more test and it seems more efficient, a lot more, even than -6. Thanks.

This post has been edited by eahm: Jul 18 2012, 20:05

[IgorC]

Probably the poll could be a bit more complete with some extra options like

-8 -A tukey(0.5) -A flattop http://www.hydrogenaudio.org/forums/index....showtopic=58731
FLACUDA www.cuetools.net/wiki/File:Flaccl3.png

This post has been edited by IgorC: Jul 20 2012, 21:23

[db1989]

@eahm: done!