QUOTE(DreamTactix291 @ Oct 29 2005, 01:39) 
--vbr-new is faster than CBR which is faster than ABR which is faster than --vbr-old.
That's mostly true, except it looks like ABR is (a little bit) faster than CBR. That's what this little test I did indicated, anyway.
CODE
Track 1 (pop punk) - 3 minutes 51 seconds
-V 5 -h 29 seconds
-b 128 -h 24 seconds
--abr 128 -h 22 seconds
-V 5 -h --vbr-new 11 seconds
Track 2 (hard rock) - 3 minutes 44 seconds
-V 2 16 seconds
-b 192 12 seconds
--abr 192 12 seconds
-V 2 --vbr-new 12 seconds
Track 3 (industrial) - 10 minutes 43 seconds
-V 0 46 seconds
-b 256 38 seconds
--abr 256 38 seconds
-V 0 --vbr-new 35 seconds
-V 5 -h 29 seconds
-b 128 -h 24 seconds
--abr 128 -h 22 seconds
-V 5 -h --vbr-new 11 seconds
Track 2 (hard rock) - 3 minutes 44 seconds
-V 2 16 seconds
-b 192 12 seconds
--abr 192 12 seconds
-V 2 --vbr-new 12 seconds
Track 3 (industrial) - 10 minutes 43 seconds
-V 0 46 seconds
-b 256 38 seconds
--abr 256 38 seconds
-V 0 --vbr-new 35 seconds
For "Track 2" I used a stopwatch to make sure I put them in the correct order. When put next to a time device that measures hundredths of a second, it becomes apparent that LAME always rounds the encoding time down. The cat's out of the bag, LAME devs.
I tried to keep the bitrate in a somewhat similar range too. I'm not sure if that really makes any difference, but I wanted to eliminate as many variables as possible.
Edit: I just realized that I messed up the last ABR/CBR comparison. It looks like CBR beats out ABR by a very small amount. Maybe that general rule is true after all. I can't seem to get consistent results, so I'll just say that they're the same speed.
