The two systems have very different approaches to audio. The SNES is completely sample based. This was novel technology at the time. The MD/Genesis instead has a more traditional solution consisting of an FM chip and "sound generator" or "noise" chip. It's hard to say which is "best" in this regard. They are fundamentally different approaches and each has it's own limitations.
You can't really create sounds on the SNES. Instead, the chip is fed samples of recorded audio and then the chip allows for different actions to be taken on that sample (e.g. adjusting the EQ, Envelope, Loop, etc). The major limitations with this approach is bit depth for the samples (quality) and storage (sample length).
On the MD, the chips themselves create the sounds either through direct manipulation of the waveforms (FM) or by selecting a specific waveform and performing various modifications such as pitch or frequency (sound chip). This is like having a synth built into the unit. It's only going to sound like that particular synth and with some clever hacks, you can get close to realistic sounding audio, they are still hacks at the end of the day. The MD isn't capable of playing samples using the built in hardware.
The SNES was capable of producing a near limitless array of sounds and could even reproduce realistic sounding instruments as it was just playing audio samples. Something the MD simply couldn't do. That said, for the sounds the MD could produce, it would sound "better," if that makes sense. As an analogy, playing a song on a YM2612 based synth and the MD, which uses the same FM chip, would sound identical. However, playing a song on a YM2612 synth and then playing the same song on the SNES (via samples) would be noticeably different, or may not even be possible at all due to the limitations of the hardware at the time.
One little correction, I think: the MD/Genesis had one PCM channel so it could play samples. Hence digitized speech in some games, for example. But, yes, it had nowhere near the capability to manipulate samples (aside from involving the CPU, of course!) that the SF/SNES's chip had.
What an interesting time when there were so many valid extremely different approaches to problems like "making sound" that led to such different outcomes.
You can't really create sounds on the SNES. Instead, the chip is fed samples of recorded audio and then the chip allows for different actions to be taken on that sample (e.g. adjusting the EQ, Envelope, Loop, etc). The major limitations with this approach is bit depth for the samples (quality) and storage (sample length).
On the MD, the chips themselves create the sounds either through direct manipulation of the waveforms (FM) or by selecting a specific waveform and performing various modifications such as pitch or frequency (sound chip). This is like having a synth built into the unit. It's only going to sound like that particular synth and with some clever hacks, you can get close to realistic sounding audio, they are still hacks at the end of the day. The MD isn't capable of playing samples using the built in hardware.
The SNES was capable of producing a near limitless array of sounds and could even reproduce realistic sounding instruments as it was just playing audio samples. Something the MD simply couldn't do. That said, for the sounds the MD could produce, it would sound "better," if that makes sense. As an analogy, playing a song on a YM2612 based synth and the MD, which uses the same FM chip, would sound identical. However, playing a song on a YM2612 synth and then playing the same song on the SNES (via samples) would be noticeably different, or may not even be possible at all due to the limitations of the hardware at the time.