You know I was thinking... we got all giddy about the ~20% speed bump in 3.95, but the --fast settings are like twice as quick! have any detailed listening tests been done in a while to determine that they really are inferior? Any feedback from Gabriel et al? Is the new VBR psychoacoustic model simpler or something? Because man, if we are missing out on those quick encodes for no good reason...

Maybe all further ABX testing should compare:

LAME 3.90.3 --alt-preset standard
LAME 3.96 --preset standard
LAME 3.96 --preset fast standard

we may learn something interesting. Or not.

it would be worth adding this to your list

LAME 3.90.3 --alt-preset fast standard

When I had some time a while ago (unfortunately, I haven't had much time to sit at my home computer for weeks ), I used MYF_4SEC.WAV to test LAME 3.96 --preset fast standard, --preset standard, and 3.90.3 --alt-preset standard. For that sample, 3.96 --preset standard was the clear loser, and 3.96 --preset fast standard was very slightly better than 3.90.3 --alt-preset standard. Test Results here.

It would be great if someone people could also do a little testing on 3.96 --preset fast standard against 3.96 --preset standard or 3.90.3 --alt-preset (fast) standard, but people's focus now seems to be on ~128kbps range...

For further development of the new VBR routine, it would be welcome and very interesting, if the test Jebus suggested took place.

with 3.90.3 using fast standard also meant slightly higher bitrates. is that still true with 3.96?

Don't you mean slightly LOWER bit rates? At least that's what I've been gettng.

please correct me if i'm wrong. i remember that using fast standard with 3.90.2 you got slightly lower quality with a slightly higher bitrate.

http://www.hydrogenaudio.org/forums/index.php?showtopic=596

That was my understanding as well.

To me the most interesting would be testing preset standard with full 32-320kbps VBR. This is IMO what is most lacking in current lame & size/quality in preset standard encoding, it could be better!

the link doesn't seem to back up your claims.
Anyway. The newer vbr routine has been improved and I think it would be worth testing it to see where we really are NOW--with 3.96.1.

I have been trying to ABX different samples with 3.96.1 --preset standard vs --preset fast standard, but the only sample where I was able to ABX the difference so far is myf_4sec (see above), and --pfs sounded clearly better.

With other samples, when --ps is not transparent, --pfs is also not transparent, and even when I seem to hear some coarseness with --pfs and not with --ps, when I try to ABX the two samples against each other, I fail to score <5% each time.

I re-tested myf_4sec.wav again, and got basically the same results (--ps 132kbps/--pfs 138kbps):

ABC/HR Version 0.9b, 30 August 2002
Testname: lame 3.96.1 vbr-old vs vbr-new/mrth

1R = C:\My Test Samples\MYF-4SEC\3.96.1-preset-standard.mp3.wav
2L = C:\My Test Samples\MYF-4SEC\3.96.1-preset-fast-standard.mp3.wav

---------------------------------------
General Comments:

---------------------------------------
1R File: C:\My Test Samples\MYF-4SEC\3.96.1-preset-standard.mp3.wav
1R Rating: 3.0
1R Comment:
---------------------------------------
2L File: C:\My Test Samples\MYF-4SEC\3.96.1-preset-fast-standard.mp3.wav
2L Rating: 4.0
2L Comment:
---------------------------------------
ABX Results:
Original vs C:\My Test Samples\MYF-4SEC\3.96.1-preset-standard.mp3.wav
5 out of 5, pval = 0.031
Original vs C:\My Test Samples\MYF-4SEC\3.96.1-preset-fast-standard.mp3.wav
5 out of 5, pval = 0.031
C:\My Test Samples\MYF-4SEC\3.96.1-preset-standard.mp3.wav vs C:\My Test Samples\MYF-4SEC\3.96.1-preset-fast-standard.mp3.wav
7 out of 8, pval = 0.035

edit: spelling

Thank you very much LoFiYo, for taking some time and helping.

You all may wonder why I start these threads and then don't participate... fact is even with training I can't tell the difference - just don't have the gift

Anyhow, still very interested and I'd love for some others to add their ABX results on known problem samples.