IPB

Welcome Guest ( Log In | Register )

2 Pages V  < 1 2  
Reply to this topicStart new topic
OptimFROG stability
Soap
post Dec 30 2010, 21:35
Post #26





Group: Members
Posts: 1001
Joined: 19-November 06
Member No.: 37767



QUOTE (Porcus @ Dec 30 2010, 15:08) *
If not: Sure you can't get foobar2000 + TAK to work with FreeBSD+Wine? Considering how slow OptimFrog decoding is, you might even save CPU time.

WINE is a translation layer, not a classic emulator or virtual machine, it is just a set of libraries. It costs almost nothing in CPU time. If TAK is faster than OF when both are in Windows it almost assuredly will be when TAK is running through WINE and OF is native.
QUOTE (Porcus @ Dec 30 2010, 15:08) *
By the way, since you mention ZFS: Myself I tried WinXP + foobar2000 virtualized under OpenSolaris, without success. I also tried NexentaStor virtualized (with raw drive access) on a WinXP box. I couldn't get realtime music playback in either, though I didn't try very hard before I gave up. YMMV.

foobar2000 runs through WINE just fine in all my experience on *nix. I don't know why one would bother to launch a VM for it...


--------------------
Creature of habit.
Go to the top of the page
+Quote Post
jetpower
post Dec 30 2010, 22:41
Post #27





Group: Members
Posts: 67
Joined: 6-September 05
Member No.: 24363



QUOTE (_m_ @ Dec 30 2010, 14:56) *
I am talking about compression ratio. And I don't need more drives, my current ones are well sufficient. I'm just a compression hobbyist and like when things are small.

then:
http://www.lossless-audio.com/
or if you're extreme/patient
http://encode.ru/threads/1137-Sac-(State-o...dio-Compression
Go to the top of the page
+Quote Post
greynol
post Dec 30 2010, 22:46
Post #28





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



QUOTE (_m_ @ Dec 30 2010, 03:51) *
Thank you greynol.

You're welcome. smile.gif

QUOTE (_m_ @ Dec 30 2010, 03:51) *
Hmm... checking seems the right thing to do. The only problem is that ofr decompression takes time and will push the slowness by anither notch.

This is precisely why I wrote:
QUOTE (greynol @ Dec 29 2010, 11:21) *
This should give you a good idea about OptimFROG's performance.
You can't just pretend that decoding time isn't an issue. I did that with Monkey's Audio a while back and despite getting warnings from people far and wide I pushed on, later only to regret my bull-headed decision; and I was only using the high preset.


--------------------
Your eyes cannot hear.
Go to the top of the page
+Quote Post
Destroid
post Dec 31 2010, 00:48
Post #29





Group: Members
Posts: 544
Joined: 4-June 02
Member No.: 2220



QUOTE (greynol @ Dec 30 2010, 21:46) *
QUOTE (greynol @ Dec 29 2010, 11:21) *
This should give you a good idea about OptimFROG's performance.
You can't just pretend that decoding time isn't an issue. I did that with Monkey's Audio a while back and despite getting warnings from people far and wide I pushed on, later only to regret my bull-headed decision; and I was only using the high preset.
I concur, the problem with APE and OFR in their strongest compression modes have ultra-slow decoding (I think OFR slightly lessened) basically because of their "symmetric" designs (also true for that other ultra-compressor I can't remember the name of). For those who don't know, this symmetric design equates to relatively equal performance (time) for decoding as it took to encode. I think WAVPack is by default symmetrical but can be asymmetric with -hx or similar switches.

At any rate, I did not ever find any problems with OFR causing corruption, although I haven't tested for a while and don't know if anything gets mangled using different file systems (can't imagine why, but it would be disastrous for the codec).

@greynol: I think MAC got less efficient when the best-compression war got into high gear. At any rate, I am still primarily using a outdated version of APE at default -c2000 setting that I found was the most efficient, and- to this day- still pretty comparable to TAK in encoding speed, falling a bit short on compression ratio and slower at decompression, of course. Why keep using this old software? Because I want the library to be homogeneous tongue.gif I guess I'm not paranoid enough to believe some of the FUD out there biggrin.gif


--------------------
"Something bothering you, Mister Spock?"
Go to the top of the page
+Quote Post
_m_
post Dec 31 2010, 11:59
Post #30





Group: Members
Posts: 231
Joined: 6-April 09
Member No.: 68706



QUOTE (Porcus @ Dec 30 2010, 21:08) *
Are you sure that OptimFrog does out-compress TAK on your particular material?

I have a rather large collection that's not biased towards any genre, I could bet 100 to 1 it will be stronger.

QUOTE (Porcus @ Dec 30 2010, 21:08) *
If not: Sure you can't get foobar2000 + TAK to work with FreeBSD+Wine?

Yeah, I thought about it and I'm sure I could make it work. I'd fist start with XMMS because it claims compatibility with Winamp plugins, but somehow I'm sceptical...nontheless I would surely find some fairly good player that would work on Wine. But I don't want it, I prefer to use something supported natively.

QUOTE (Porcus @ Dec 30 2010, 21:08) *
By the way, since you mention ZFS: Myself I tried WinXP + foobar2000 virtualized under OpenSolaris, without success. I also tried NexentaStor virtualized (with raw drive access) on a WinXP box. I couldn't get realtime music playback in either, though I didn't try very hard before I gave up. YMMV.

I considered a similar, though reversed setup several months ago. smile.gif
A Windows host, Solaris guest, communication via SMB or so. But I'm really dissatisfied with where Windows is heading and after a bit of using Linux / playing with BSD I'm sure they are viable choices. For me better ones than Windows.
QUOTE (jetpower @ Dec 30 2010, 22:41) *
QUOTE (_m_ @ Dec 30 2010, 14:56) *
I am talking about compression ratio. And I don't need more drives, my current ones are well sufficient. I'm just a compression hobbyist and like when things are small.

then:
http://www.lossless-audio.com/
or if you're extreme/patient
http://encode.ru/threads/1137-Sac-(State-o...dio-Compression


Yeah I gave both a thought. With SAC it was a very short one, amount of work needed to make it useful is simply too large. With LA I needed a short research, but, like I said already, I decided against it because ID3v1 tags are insufficient for me.

QUOTE (greynol @ Dec 30 2010, 22:46) *
QUOTE (_m_ @ Dec 30 2010, 03:51) *
Hmm... checking seems the right thing to do. The only problem is that ofr decompression takes time and will push the slowness by anither notch.

This is precisely why I wrote:
QUOTE (greynol @ Dec 29 2010, 11:21) *
This should give you a good idea about OptimFROG's performance.
You can't just pretend that decoding time isn't an issue. I did that with Monkey's Audio a while back and despite getting warnings from people far and wide I pushed on, later only to regret my bull-headed decision; and I was only using the high preset.

Judging by Synthetic Soul's benchmark I estimate that highnew would use 18-21% of one of my cores. I'm intend to use bestnew, possibly later backing off slightly with compression mode and/or seek length. Might be wrong, but I don't expect troubles with decompression speed. I don't think that converting my current collection will be a problem either, I will just leave a computer running for a weekend or a couple. I'm somewhat afraid about encoding new albums; I use image+cue setup, so I will have full slowness of ofr using only 1 core. But singlethreaded TAK -p4m is blazing fast for my on my 5 years old PC, so I think ofr will be OK on the new one.
Go to the top of the page
+Quote Post
lvqcl
post Dec 31 2010, 12:44
Post #31





Group: Developer
Posts: 3214
Joined: 2-December 07
Member No.: 49183



QUOTE (_m_)
I'm intend to use bestnew, possibly later backing off slightly with compression mode and/or seek length. Might be wrong, but I don't expect troubles with decompression speed.


According to my tests on Core2 2400 MHz, OFR bestnew mode has 2.2x realtime encoding speed and 3.7x realtime decoding speed. (highnew: 5.0x and 7.6x, respectively).
Go to the top of the page
+Quote Post
alvaro84
post Dec 31 2010, 15:21
Post #32





Group: Members
Posts: 128
Joined: 9-August 06
Member No.: 33830



QUOTE (lvqcl @ Dec 31 2010, 12:44) *
QUOTE (_m_)
I'm intend to use bestnew, possibly later backing off slightly with compression mode and/or seek length. Might be wrong, but I don't expect troubles with decompression speed.


According to my tests on Core2 2400 MHz, OFR bestnew mode has 2.2x realtime encoding speed and 3.7x realtime decoding speed. (highnew: 5.0x and 7.6x, respectively).


Wow! Then it can be said that OFR has a very real energy consumption and carbon footprint...
Go to the top of the page
+Quote Post
WonderSlug
post Dec 31 2010, 17:23
Post #33





Group: Members
Posts: 299
Joined: 6-February 08
From: San Diego, CA
Member No.: 51066



QUOTE (lvqcl @ Dec 31 2010, 03:44) *
QUOTE (_m_)
I'm intend to use bestnew, possibly later backing off slightly with compression mode and/or seek length. Might be wrong, but I don't expect troubles with decompression speed.


According to my tests on Core2 2400 MHz, OFR bestnew mode has 2.2x realtime encoding speed and 3.7x realtime decoding speed. (highnew: 5.0x and 7.6x, respectively).


Just for the sake of comparison, could you give the encoding and decoding speed of the most popular lossless codec FLAC, at say -6 or -8, on that Core2 2400 ?

Go to the top of the page
+Quote Post
lvqcl
post Dec 31 2010, 18:36
Post #34





Group: Developer
Posts: 3214
Joined: 2-December 07
Member No.: 49183



IIRC I posted them somewhere but cannot find that post...
Results, again:

CODE
Codec compression(%) encoding/decoding(x realtime)
FLAC
0 74.256 254 376
1 72.973 236 360
2 72.716 201 366
3 71.228 205 358
4 69.569 170 339
5 69.342 131 337
6 69.339 127 339
7 69.307 53 341
8 69.162 36 325

TAK
p0 69.342 308 359
p0e 68.941 181 361
p0m 68.882 105 359
p1 68.368 239 350
p1e 68.207 141 350
p1m 68.119 80 351
p2 67.725 170 319
p2e 67.496 78 315
p2m 67.416 48 317
p3 67.381 83 286
p3e 67.289 56 284
p3m 67.249 31 285
p4 67.230 46 244
p4e 67.144 37 244
p4m 67.117 20 248

WavPack
fast 70.537 188 232
fast x1 70.338 126 233
fast x2 70.233 91 233
fast x3 70.172 54 233
fast x4 70.108 22 232
fast x5 70.095 18 232
fast x6 70.090 15 233
normal 69.280 155 186
normal x1 68.879 95 188
normal x2 68.811 62 188
normal x3 68.787 33 188
normal x4 68.635 8.4 188
normal x5 68.614 5.8 189
normal x6 68.565 2.9 188
high 68.443 117 140
high x1 68.274 68 142
high x2 68.230 41 142
high x3 68.206 21 142
high x4 68.054 5.2 143
high x5 68.030 3.9 142
high x6 68.014 2.7 143
ex.high 68.056 93 112
ex.high x1 67.974 51 111
ex.high x2 67.941 29 111
ex.high x3 67.932 14 111
ex.high x4 67.880 3.4 112
ex.high x5 67.859 2.2 112
ex.high x6 67.851 1.6 112

Monkey's Audio
fast 68.471 112 84
normal 67.396 75 63
high 67.153 63 54
extra high 66.695 29 28
insane 66.443 8.8 8.3

ALAC 70.143 100 121

WMAL 68.079 43 50

TTA 68.928 141 149

LA
normal 65.907 7.9 9.4
high 65.717 5.7 6.4

OptimFrog
fast 67.552 37 49
normal 67.000 25 33
high 66.862 16 21
extra 66.814 10 14
best 66.785 6.1 8.0
highnew 66.246 5.0 7.6
extranew 66.170 3.9 6.1
bestnew 65.922 2.2 3.7
exp. fast 67.463 12 42
exp. normal 66.908 11 29
exp. high 66.770 8.7 20
exp. extra 66.722 6.7 13
exp. best 66.693 4.6 7.7
exp. highnew 66.162 4.0 7.4
exp. extranew 66.087 3.2 6.2
exp. bestnew 65.841 2.0 3.7
Go to the top of the page
+Quote Post
greynol
post Dec 31 2010, 21:51
Post #35





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



http://www.hydrogenaudio.org/forums/index....showtopic=74114


--------------------
Your eyes cannot hear.
Go to the top of the page
+Quote Post
Destroid
post Dec 31 2010, 23:45
Post #36





Group: Members
Posts: 544
Joined: 4-June 02
Member No.: 2220



QUOTE (lvqcl @ Dec 31 2010, 17:36) *
IIRC I posted them somewhere but cannot find that post...
Results, again:
edit: nevermind, I should have read hyperlinked post above.
CODE
Codec compression(%) encoding/decoding(x realtime)
...
Monkey's Audio
insane 66.443 8.8 8.3

LA
normal 65.907 7.9 9.4
high 65.717 5.7 6.4

OptimFrog
...
best 66.785 6.1 8.0
highnew 66.246 5.0 7.6
extranew 66.170 3.9 6.1
bestnew 65.922 2.2 3.7
...
exp. best 66.693 4.6 7.7
exp. highnew 66.162 4.0 7.4
exp. extranew 66.087 3.2 6.2
exp. bestnew 65.841 2.0 3.7
Yikes! (Don't let California know if this is the lossless program you use. /sarcasm)

This post has been edited by Destroid: Dec 31 2010, 23:50


--------------------
"Something bothering you, Mister Spock?"
Go to the top of the page
+Quote Post
WonderSlug
post Jan 1 2011, 06:09
Post #37





Group: Members
Posts: 299
Joined: 6-February 08
From: San Diego, CA
Member No.: 51066



QUOTE (lvqcl @ Dec 31 2010, 09:36) *
IIRC I posted them somewhere but cannot find that post...
Results, again:


I see. So FLAC is about 7 times faster than OptimFROG in encoding speed, and about 50 times faster in decoding.

What's the advantage of OptimFROG again?

Oh, and as for this:

QUOTE
I am talking about compression ratio. And I don't need more drives, my current ones are well sufficient. I'm just a compression hobbyist and like when things are small.


I have a single external 2 TB hard drive that cost me $130 USD. On it I have stored FLAC encodings of more than 2000 CDs. That's two thousand audio CDs. Along with those FLAC encodings are both MP3 and AAC transcodes of those FLACs. The 3 most supported formats of 2000 CDs on one external drive. I still have about 500 GB of space free. The FLACs of those 2000 CDs took up only about 800 GB. The MP3 and AAC transcodes took up about 250 GB each.

I purchased a duplicated drive, copied the contents of the first drive to it, and have it stored someplace else.

Go to the top of the page
+Quote Post
Porcus
post Jan 1 2011, 09:18
Post #38





Group: Members
Posts: 1780
Joined: 30-November 06
Member No.: 38207



QUOTE (Soap @ Dec 30 2010, 21:35) *
foobar2000 runs through WINE just fine in all my experience on *nix. I don't know why one would bother to launch a VM for it...


Limitations mentioned here: http://www.hydrogenaudio.org/forums/index....st&p=689528


--------------------
One day in the Year of the Fox came a time remembered well
Go to the top of the page
+Quote Post
Porcus
post Jan 1 2011, 09:36
Post #39





Group: Members
Posts: 1780
Joined: 30-November 06
Member No.: 38207



QUOTE (WonderSlug @ Jan 1 2011, 06:09) *
FLAC encodings of more than 2000 CDs.


... i.e., about 2 months of music. Or one month of transcoding, with speeds like reported above. Add another couple of weeks for decode+verify.

Compare FLAC -6 speeds from same table: Half a day for encoding. Plus 4 hrs for decode+verify.


--------------------
One day in the Year of the Fox came a time remembered well
Go to the top of the page
+Quote Post
Soap
post Jan 1 2011, 13:14
Post #40





Group: Members
Posts: 1001
Joined: 19-November 06
Member No.: 37767



QUOTE (Porcus @ Jan 1 2011, 03:18) *
QUOTE (Soap @ Dec 30 2010, 21:35) *
foobar2000 runs through WINE just fine in all my experience on *nix. I don't know why one would bother to launch a VM for it...


Limitations mentioned here: http://www.hydrogenaudio.org/forums/index....st&p=689528

EDIT:
Don't see the problem.


This post has been edited by Soap: Jan 1 2011, 13:20


--------------------
Creature of habit.
Go to the top of the page
+Quote Post
lvqcl
post Jan 1 2011, 13:38
Post #41





Group: Developer
Posts: 3214
Joined: 2-December 07
Member No.: 49183



http://www.hydrogenaudio.org/forums/index....showtopic=46018 -- "good old times" wink.gif
Go to the top of the page
+Quote Post
DARcode
post Jan 1 2011, 16:07
Post #42





Group: Members (Donating)
Posts: 679
Joined: 10-January 05
From: Italy
Member No.: 18968



A tad off topic and maybe silly too, but I've just realized how competitive The True Audio (TTA) codec is.


--------------------
WavPack 4.70.0 -b384hx6cmv/qaac 2.32 -V 100
Go to the top of the page
+Quote Post
Destroid
post Jan 1 2011, 22:13
Post #43





Group: Members
Posts: 544
Joined: 4-June 02
Member No.: 2220



Nah, you can't really go wrong with lossless, I think that message has always been clear. It's just the matter of being aware of the pros & cons of each program since the idea is to stick to one instead of re-encoding every few months. Decoding performance is a very important factor, unless one does not plan to frequently access the audio files. Compression is always highly regarded given that the trade-offs are not substantial. With OFR in high compression modes, there is a high price in decoding performance, and if the idea is to have a frequently accessed library I would recommend another codec. For archival to be put in the dungeon and accessed 3 times a year, OFR high compression is fine.


--------------------
"Something bothering you, Mister Spock?"
Go to the top of the page
+Quote Post
lvqcl
post Jan 1 2011, 22:33
Post #44





Group: Developer
Posts: 3214
Joined: 2-December 07
Member No.: 49183



QUOTE (Destroid @ Jan 2 2011, 00:13) *
For archival to be put in the dungeon and accessed 3 times a year, OFR high compression is fine.

...and if recovery time is not an issue...
Go to the top of the page
+Quote Post
db1989
post Jan 2 2011, 00:29
Post #45





Group: Super Moderator
Posts: 5159
Joined: 23-June 06
Member No.: 32180



Just how much space do you estimate OFR would save? Would it really be worth all that extra time? Dungeon or no, it seems a bit like you want ‘maximal’ compression (clearly not really possible) for its own sake, forsaking a lot of practicality and compatibility.

This post has been edited by dv1989: Jan 2 2011, 00:32
Go to the top of the page
+Quote Post
_m_
post Jan 2 2011, 14:26
Post #46





Group: Members
Posts: 231
Joined: 6-April 09
Member No.: 68706



QUOTE (dv1989 @ Jan 2 2011, 00:29) *
Just how much space do you estimate OFR would save? Would it really be worth all that extra time? Dungeon or no, it seems a bit like you want maximal compression (clearly not really possible) for its own sake, forsaking a lot of practicality and compatibility.


Got home, will be building the computer today. There are some hardware incompatibilities, I might need a new wifi card. I expect to have workable OS either tomorrow or later this week.
Then there will be a lot of testing, I have to ensure I can do everything I need to before I start converting...and I already noticed some troubles, after using foobar I assumed that it's usual for players to support media databases, replaygain, cue files, flexible tagging and Unicode...turns out I was wrong. So I have to test XMMS with several plugins, a couple of forks of it (some with own plugins) and Amarok claims compatibility with XMMS plugins. Though most mention visualisation which makes me wonder if playback ones work too. If neither does what I need I will have to rethink the whole deal.... :|
I may end up using foobar+ofr (--experimantal?). Of foobar+tak. Or some unix player with ape. One thing is sure, I'm not going to use anything weaker than what I have now and if good unix players don't support it, I won't use them.

As I got lvqcl data by hand, I did some *rough* estimations of how much time will it take and how much will be saved. As I mentioned earlier, I intend to use bestnew at first. And I didn't mention, but I want to use --optimize best.
Assumptions:
*My Athlon is as fast as lvqcl's Intel clock for clock.
*It doesn't run into new bottlenecks when using all 4 cores.
*Decompressing taks, script overhead and such are negligible
*I have 0.5 TB of music stored as TAK. It's not entirely true, I didn't make the tak transition fully because of Unicode related issues, so I have some flacs too. And I don't know how much music do I have because a good part of my collection is taken by artwork scans.
*lvqcl's test files are representative. I doubt it because LA scored too well compared to other tests that I've seen.

And I did the calculations yesterday, now I'm writing from memory, so they may be a bit off wink.gif
ofr encode+decode is 1.39x realtime of lvqcl's CPU, ~7.2x realtime when using 4 cores of my CPU.
A week straight to compress whole library. I won't be running my computer 24x7 and I will be doing other stuff in the meantime, so it's likely to take 1 month. With --optimize best much more.
It saves almost exactly 2% over TAK. 2% of already compressed size, not original size.
10 GB saved over entire collection. 2 copies on ZFS + backup, so 30 GB in total. Total of 192 MB saved per hour of computations.
Go to the top of the page
+Quote Post
Remedial Sound
post Jan 2 2011, 20:51
Post #47





Group: Members
Posts: 501
Joined: 5-January 06
From: Dublin
Member No.: 26898



A polite request for a mod to close this thread.

The OP's question has already been answered for all intents and purposes. I can't see any reason to further discuss one random person's obsession with doing something completely pointless, impractical, and time/energy consuming.
Go to the top of the page
+Quote Post
Soap
post Jan 2 2011, 21:57
Post #48





Group: Members
Posts: 1001
Joined: 19-November 06
Member No.: 37767



QUOTE (Remedial Sound @ Jan 2 2011, 14:51) *
A polite request for a mod to close this thread.

Thread wouldn't "need" closing if people stopped posting to it! wink.gif


--------------------
Creature of habit.
Go to the top of the page
+Quote Post
greynol
post Jan 2 2011, 22:29
Post #49





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



QUOTE (Remedial Sound @ Jan 2 2011, 11:51) *
I can't see any reason to further discuss one random person's obsession with doing something completely pointless, impractical, and time/energy consuming.

I don't see any rules of this forum being broken except possibly a very polite TOS #2 violation and other posts of a similar nature.

This post has been edited by greynol: Jan 2 2011, 22:47


--------------------
Your eyes cannot hear.
Go to the top of the page
+Quote Post
larryfine
post Feb 14 2011, 15:24
Post #50





Group: Members
Posts: 48
Joined: 21-September 10
Member No.: 84040



QUOTE (_m_ @ Dec 29 2010, 11:55) *
I'm considering converting my collection for OptimFROG, but seeing that the latest version is alpha...,

Good News,
Florin Ghido is back and released a new beta stable version (4.910b).
The speed of encoding/decoding at normal settings is good but very slow at high, extra and best compression modes.

www.losslessaudio.org


--------------------
loquor mee menti: factus de materia, cinis elementi...
Go to the top of the page
+Quote Post

2 Pages V  < 1 2
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 21st April 2014 - 11:58