Having multiple graphics or sound chips in a system has exactly the effects you would expect. It would allow the composite system to perform multiple processing tasks at once, with the results of each contributing processor being combined into the final output (or spread across multiple outputs), producing more sophisticated visuals or sounds.
Graphics:
Multiple video processors provide the ability to compose an image with multiple independent display layers, whether they be bitmaps composed of bitplanes or chunky layers, sprites or tilemaps.
An example: The 32X attachment for the Mega Drive provides a separate video chip independent of the one in the base system. They both draw their own set of tilemaps and sprites simultaneously, with the output from the Mega Drive being routed into the 32X with a video cable, and then output to the TV afterwards. The result is additional independent layers of graphics, and additional video RAM for storing tiles and sprites (in that both units are independently storing their own resources).
A similar system is used when a 2D graphics adapter is used with an early 3D graphics accelerator in a Win 98-era PC. Like the 32X, you plug the output of one into the other. The 3dfx Voodoo accelerator provides 3D graphics rendering, which is mixed with the 2D image buffer provided by the 2D card.
Later, a similar card-linking system for the PC would be developed (with various names, such as like SLI) where two cards work in unison to provide a single output image by processing partial parts of the output simultaneously.
Sound:
A system comprised of multiple independent sound chips can use all the different voices and modes of the base chips mixed in any way the system designer has allowed for. If the outputs of the chips are mixed together, this allows all the independent waveform/PSG/FM voices to be used at once, or assigned to different stereo channels. It could also be used to provide a higher bit-resolution output.
Examples:
The original Sound Blaster PC sound card included a single OPL2 FM synthesizer chip for AdLib music card compatibility. This chip provides mono music output. The Sound Blaster Pro 1 improves on this by including a second OPL2 alongside the first. The two chips are linked to the left and right channels to provide stereo musical output; as each chip is programmed independently, complex panning and echo effects are possible. Many Sierra DOS games provided special support for the Dual OPL2 version of the Sound Blaster Pro in their music driver.
(For backwards compatibility, attempted direct hardware access to the original single OPL2 affects both the chips in the Pro 1 equally, resulting in a mono sound. Later SB Pros used an OPL3 chip instead which uses a different interface - still single OPL2 compatible, but not with dual OPL2.)
Later Sound Blaster cards and clones combine even more multiple independent sound interfaces together! Final generation ISA Sound Blaster AWE cards contain:
- OPL3 FM synthesized sound (backwards compatible with OPL2),
- a typical SB series DMA-driven PCM stereo waveform sound stream,
- an EMU8000 wavetable sound processor,
- a feature connector for an add-on wavetable sound processor with its
own ROM sounds, (distinct from the EMU)
- CD-ROM interface Redbook audio input,
- and a line-in input (which may be connected to an
external MIDI device through the MPU-401-compatible gameport for
high-quality sampler MIDI music)!
DOS and Windows games may use any combination of these to provide sound and music output.
And, yes, I believe that if you put multiple sound cards in a PC and there were no resource conflicts (manual settings for DOS, maybe a bit of tweaking for Windows), you could combine multiple sound cards together in that fashion as well. It's possible to install a Gravis Ultrasound wavetable-based card alongside a Sound Blaster PCM-waveform-based card, though very very few pieces of software would attempt to use them both simultaneously!
The Mega Drive contains both the YM2612 FM chip and the earlier SN76489 PSG chip from the Master System. Games use both together in various ways to provide different musical timbres - the FM chip produces bassy sounds, and complex effects; the PSG chip produces square waves. Instrument patches can use any number of voices from either chips to create complex output sounds one chip alone could not produce.
The CPU overhead of controlling a synthesizer chip is not high. The whole purpose of using chips like these is to offload waveform calculation to the chips themselves rather than the CPU. In the simplest case, a small routine updates the register state of each chip at regular intervals. The most complex thing a synthesizer chip control routine would do is to constantly alter the parameters of the synthesized tone by calculating or stepping through an ADSR envelope to provide a more pleasing sound than the straight square-wave tone of a PSG chip (such as vibrato effects in Game Boy music), and even that is very low-overhead.
