Here's some information I found on the net maybe 8 years ago and saved to a Word file. I don't recall the source, but the guy provides an overview of the alternatives. Take it for what it's worth:
"There seems to be too little information being put forward here. How about a recap and summary to clear things up?
ReplayGain is basically a method for determining the perceived volume of a song. That's really all it is, in and of itself, it doesn't change anything. You scan the song data, do some math, and get a number.
To make songs sound the same, you have to make all your songs have the same number. To do this, you apply gain. So if a song has a RG number of 83 dB and you want to make everything sound at 89 dB, you would apply a 6 dB gain to the song as you play it. In theory, applying gain can be done losslessly. In practice, we don't deal with an infinite range of values, and so there will always be rounding error. The rounding error is basically unnoticeable, but it's important to remember it because it does add up over time, if you keep applying gain and changing a file's actual audio data over and over again.
MP3Gain will get the RG value from an MP3 track and apply a gain that's reasonably close to bringing the song to a standard. MP3Gain is limited by the fact that it can only adjust the gain in 1.5 dB increments, but it's a winner in that it can do so totally reversibly. Changing the gain using MP3Gain over and over again does not cause any degradation or additive rounding error. This is because it's not modifying the actual audio data, but modifying gain values in the MPEG stream itself. Changing the gain does change the stream, so in that sense it does change the audio in the file, but it doesn't change the actual substance of the audio, and so it's reversible and/or non-lossy. Since the stream is actually changed, all audio players will support it, and since it has tags, some players will support it more exactly.
WAVGain is a program that will scan for the RG value and apply a gain to the actual audio data in the WAV. It does this in a way that's similar to "normalizing". Figure out what gain you want to apply, convert this to a multiplier, multiply every sample in the WAV by that value. This *does* produce rounding error, and so it is technically lossy. Do it once, you probably won't hear it. The errors will be on the order of one bit, and that's way below anybody's hearing capability. Do it over and over again, eventually you will hear it, as the errors add up and tend to flatten the waveform. Advantage of adjusting gain with WAVGain is that it is exact. It's not limited to the 1.5dB steps that MP3Gain is. Disadvantage is the rounding error problem. For burning a bunch of songs to CD and leveling them, then decompressing/copying to WAV, WAVGaining, and burning them is not a particularly bad idea. Just delete the modified WAV's after you're done.
Other ReplayGain methods are usually lossless. The normal method is to add a tag to the file that specifies the desired gain and/or the track's RG value. With this information, a player that's capable of reading it (foobar, others) can apply gain on playback directly. This is the best way to do it, because it doesn't mess with the audio data, there's no additive error, and it's as exact as the resolution of your sound card is. But it does require player support and proper metadata tagging standards. As such, a lot of players do not yet support it.
Which to use:
-If you're burning to CD: Decompress the stuff to WAV and WAVGain it, then burn the WAVs and delete them from the hard drive.
-If you're putting stuff on an MP3 portable player: MP3Gain the lot. Use Track or Album Gain as you see fit.
-If you're ripping from CD: Rip and compress with no modifications like WAVGain. You ideally want to compress the original source as clean as possible, to remove possible rounding error later. MP3Gain the resulting compressed file if you feel the need to do so.
-Not using MP3: Use players that can scan for RG, store it in tags, and apply it properly (foobar 2000)."
