The limiting factor here would be the actual MP3 format, and to a lesser extent, the software (or hardware in the case of the stand alone CDP) MP3 decoder. This has nothing to do with which sound card you use.
I suggest you download Foobar2000 as it has the best decoding engine for MP3s out of the software players (in my opinion). It has internal support for 64 bit precision, and it can output up to 32 bit floating point (but your SB Live will limit you to either 16 bit or 24 bit fixed point). This will have limited sound quality benefit, if any, but there is something very useful that it can do.
In addition to being able to output in various formats, it can work around that evil little 48 kHz internal up sampling problem that all Emu 10k1 or 10k2 chip-based sound cards (such as the SB Live, Audigy, or even high end cards like Emu's APS) have. This problem arises because the EMU chip can only do DSPs and other effects at 48 kHz, so everything that passes through it is resampled at 48 kHz, processed, and then converted to the output format. This wouldn’t matter if the resampler was of high quality, but it is not. So how do you work around this? This is done by enabling the resampler DSP and then setting it to 48 kHz at high quality (slow mode). In the future, if you chose to use digital output, make sure you leave the output format at 48 kHz, or else you will be introducing yet another sampling rate conversion.
Yet another feature of Foobar2000 is that it will let you select if you want to enable dithering, and which type you would like. I suggest Soft ATH noise shaping at 24 bit fixed point or Strong ATH noise shaping at 16 bit. Try both of these settings and see if the sound quality gain (if you notice any) is worth the extra CPU cycles.
