Re:The Legend of Nintendo
by mrob on Monday June 25, 2018 @07:53PM
(#17396865)
>combination of Z80 and YM2612 used by Genesis had 8KB of RAM, supported 6 voices and had 22 kHz sampling rate
Technically true, but misleading, because the 22kHz only applied to the DAC mode. The effective sampling rate (sampling rate needed to reproduce the output using PCM synthesis) of the FM synthesis mode is about 53kHz. And the Genesis also had a 4 channel (3 square, 1 noise) SN76489 PSG integrated on the VPU die, with effective sampling rate of about 224kHz, which could be used at the same time as the YM2612. In practice the frequency response is bottlenecked by the analog parts, but it still beats the SNES. The limited RAM is also of little importance because FM synthesis can generate complex tones without it.
In contrast, SNES music is entirely sample based, so it's entirely dependent on sample RAM size (unless you're streaming samples from the cartridge, which is too expensive to do during gameplay). Theoretically it could use extremely short looping samples for PC Engine-style wavetable synthesis (which to my ears sounds better than standard SNES music), but AFAIK no commercially released game did this.
>Until you realize that all instruction execution times are multiples of four cycles. Yes, the CPU is 7.6 MHz, not 3.58 MHz, but effectively SNES CPU is slightly faster for FastROM games
The Genesis not only had a faster clock, but also a wider data bus, more registers, wider registers, and higher memory bandwidth. You're right that "megahertz myth" comparisons are invalid, but you're wrong that this lets the SNES win. https://segaretro.org/Blast_processing has some detailed technical comparisons.
Technically true, but misleading, because the 22kHz only applied to the DAC mode. The effective sampling rate (sampling rate needed to reproduce the output using PCM synthesis) of the FM synthesis mode is about 53kHz. And the Genesis also had a 4 channel (3 square, 1 noise) SN76489 PSG integrated on the VPU die, with effective sampling rate of about 224kHz, which could be used at the same time as the YM2612. In practice the frequency response is bottlenecked by the analog parts, but it still beats the SNES. The limited RAM is also of little importance because FM synthesis can generate complex tones without it.
In contrast, SNES music is entirely sample based, so it's entirely dependent on sample RAM size (unless you're streaming samples from the cartridge, which is too expensive to do during gameplay). Theoretically it could use extremely short looping samples for PC Engine-style wavetable synthesis (which to my ears sounds better than standard SNES music), but AFAIK no commercially released game did this.
>Until you realize that all instruction execution times are multiples of four cycles. Yes, the CPU is 7.6 MHz, not 3.58 MHz, but effectively SNES CPU is slightly faster for FastROM games
The Genesis not only had a faster clock, but also a wider data bus, more registers, wider registers, and higher memory bandwidth. You're right that "megahertz myth" comparisons are invalid, but you're wrong that this lets the SNES win. https://segaretro.org/Blast_processing has some detailed technical comparisons.