Hi,

is someone able to uplaod me the orginal Lame 3.90.3 bundle? I think it is written by Dibrom. Would be nice...

I found a link to it:
http://homepage.ntlworld.com/jfe1205/LAME/lame-3.90.3.zip

Wow, isn't GOOGLE amazing?

Just for curiousity, why do you want Lame 3.90.3? 3.97b2 has been tested and will provide you with the same quality and smaller file sizes.

Not to mention noticeably faster encoding times.

Really, dude, 3.90.3 is dead.

At high bit rate 3.90.3's and 3.97b2's overall quality is pretty much the same (see current 320 kbps soundexpert test).

When it comes to robustness I consider high bitrate 3.90.3 better due to problem sample behavior (when avoiding the vbr modes including aps, ape).

All a matter of taste, but there's no reason to bury 3.90.3.

It is not a matter of taste anymore than you using 3.89 or 3.90.2, or r3mix.

It has already been buried.

3.90.3 was originated in this site. It was hosted on Rarewares. Now it is not. It has been officially pronounced dead.

But hey, if slower encodes put your mind at ease, be my guest.



Oh, and you base your opinions on a site that posts this?

It must be a joke.

I don't want to start another discussion like this again (and I've been warned several times not to do so).

So I will definitely do only this reply:

It's the other way around as you stated it. I've done a lot of listening tests on my own which led me to the conclusion that quality increase of current lame is in the low and moderate bitrate range whereas in the high bitrate range (which I'm interested in) overall quality did not increase.
As for certain problem samples 3.90.3 was even superior (I contributed on several threads on that).
And when I saw Serge Smirnoff's test was coming up I proposed to try 3.90.3 as well as current lame. To me that's verification.

The problem is that many people generalize a quality statement. If they like 3.97b2 they do not only like the advantages at the moderate bitrate they use, but they think it is to be universally better.

So everybody should do what he thinks is best, and let other people do the same.

I think you are mistaken. There are a few of the test samples (like velvet) that are commonly used for testng that I'm not able to reliably discern between PCM and ATRAC3plus when using Hi-MD, which is the codec used in this "test". I'm not defendng this actual test or results -- but an inadequate test doesn't mean that the codec is automatically crap.

I have no doubts in my mind that ATRAC3plus will do better than some non-recommended LAME CBR mode, especially using an older version of the LAME codec and the current version of the ATRAC3plus Hi-MD codec. I'm fairly sure that using recommended settings would be a different story as well.

You might be uninfomred. ATRAC3plus is a new codec, and it is not compatible with the original ATRAC codec, nor is it that "ATRAC3 ACM" that is floating around, which has incredibly bad quality even when compared to the older generation of MiniDisc technology. On some of the known problem samples, ATRAC3plus at 384kbps is better than LAME 3.90 to my ears, but both still have some noticeable artifcats (castanets).

You can do your own blind tests by transferring some test samples encoded in MP3, ATRAC3plus, and PCM to HiMD and have a friend randomize the track order. See how you do.

I know this isn't a `true` ABX test, so I hope this isn't a TOS#8 violation by posting this, but it is a blind test and it is repeatable by anyone. It's more of a violation to knock a codec if you never actually tested it. If this post is out of line, feel free to edit it.

When most people think of MD codecs, they think of that old ATRAC ACM. It's not any more correct to assume all MP3 codecs sound "bad" because some old BladeEnc sounds bad. Over the years, all codecs have made big improvements in quality, ATRAC3plus included, and are all much closer in quality.

Edit: Ah, not even LAME, the MP3Pro codec in Audition. I don't even know the quality of that codec, but I find it hard to imagine being better than LAME. While the methodology of the test linked may certainly be flawed, I still urge you to test HiMD's ATRAC3plus yourself.

Edit: Corrections.

Addition: To avoid going too far off-topic, I believe that listening tests in general confirm that the latest LAME codec is a general improvement. The samples used are not "killer" samples, and the improvements were actually remarkable.

Thanks and this is the complete orginal bundle of the Encoder by Dibrom?

Dibrom created 3.90.2. I made the small change, as specified by Dibrom, and compiled the 3.90.3 bundle using the ICL4.5 optimisations originally used by Dibrom during his tuning exercises.

hm... if so many people still prefer lame 3.90.3, why doesn't upload rarewares this compile again? It's just a small file... although lame is to be said it's dead.

Hmm, yes, I may as well do that. I'll put it down the botton of the page.

Because some people also "prefer" to learn about Intelligent Design vs. Evolution. The scientific community is going to ignore them, as well.

EDIT:

I Guess that's a bit harsh. I'll save that one for r3mix users

Why not put it on ReallyRareWares?

yuhuuuuu

Yes, it could go on ReallyRarewares, but since there seem to be a number of people who wish to use 3.90.3 it doesn't hurt really, does it? After all, freedom of choice should be king, shouldn't it?

Probably because they don't want to encourage people to use it anymore.

hm... It's your choice of course, could also be added on RRW (probably it's better to do that, because there are also Lame compiles which are also already "dead", huh? ) but then, for newbies, it could be harder to be finded again

Why provide people with a gun if you know they are going to shoot themselves in the foot with it?

Some casual HA visitors probably don't realize we've updated to 3.97. If you leave a link to 3.90.3 on the download page, they'll continue to use it because that was always the case (having links to both the current version, and the recommended one).

If you move it to RRW, then they'll dig a little and realize why it was replaced. Just my 2 cents.

Ditto.

And don't forget about those crazy stubborn ripping groups

The fact that 3.90.3 was removed from Rarewares was the reason that I started posting here holding up 3.90.3's qualities in the high bitrate range.

I do understand that providing a lot of versions can confuse novice users. But this is the case already now with 3.98a, 3.97b, 3.961.1, 4.0a being provided. What may be missing are some statements for any version as for why there are people appreciating this very version.

I do not understand why 3.90.3 was taken from Rarewares but was not added to the older Lame versions collection on ReallyRarewares.

The general public and newbies don't need to know about 3.90. Linking 5 year old versions is very bad from a psychological point of view because the average joe will suspect serious (unfounded) flaws with current versions ( like many release groups) .

So if you link 3.90 then you must link all other versions, other wise it looks dodgy.

3.90.3 should not be linked to anymore. Period.

3.90.2 was not recommended, even though it provided no real advantage over 3.90.3. Just some samples that could be improved by the use of the -z switch. But how many?

3.97 offers clear advantages over 3.90. It is faster. Who cares if some people feel it doesn't deliver on some high bitrate samples? Let them prove it. It has always been the HA way. Just because somebody wants to hold on to a 5-year old version, doesn't mean the whole forum should confuse new users by offering two versions.

3.90.3 is dead. I loved it. It brought me lots of pleasure. But it is time to move on.

since there are so many old posts about LAME 3.90.3 it is hard for people to realize that Lame 3.97 is now recommended and not Lame 3.90.3. Many people still have the old propaganda in their head that Lame 3.90.3 is better and that's what they base their opinion on. If it is no longer available people might start to wonder why

Thanks for all the opinions. I recall now that Roberto and I intended placing it on ReallyRareWares but I guess it got forgotten. I'll have Roberto put it up when he returns from his Easter break.

A wise solution IMO.

actually for me, I feel new lame doesn't quite deliver on low bitrate samples (90-130kbps), but of course it's a matter of opinion as to what properties make one *better* than the other.
I, for example, will gladly take some *minor* artifacts in exchange for greater audio bandwidth.

I had never even tried lame 3.90.x untill near the end of last year (had always used lame 3.95-3.96).. after a little experimentation, it certainly impressed me.

Now what kind of *proof* should I provide for my preference of 'old' lame at low bitrates? I don't think ABXing with the original would be of much use (as transparency is not the point). I don't think ABXing between 2 encodes would do it either, since being *different* doesn't say which is *better*. And then there is the issue of my 'tweaked' settings making people call into question whatever the results happen to turn out as. So what kind of test would work in this case?

I think you are badly mistaken here. 3.90.3 is not tweaked for such low bitrates. 3.97 is.

And I don't think your "tweaked settings" would help much. Remember, everything that could obejetively (through ABX) help transparency is already there, by the programmers. And they are smart people, you know.

If you feel you can do better, post a lossless sample of a piece of music in which you think your choice of codec and settings improve on the comparable bitrate preset on 3.97

Use ABC/HR like all listening tests here use.

3.90 is clearly worse than 3.97 around 128k due to ringing, although 3.97 also introduces another distorsion. My results are somewhere under listening tests.

http://www.hydrogenaudio.org/forums/index....showtopic=43771
there it is everyone (i started a new thread cause it would be off-topic in this one)