Another thing I failed to mention was oversampling. Higher sample rates prevents aliasing and the so called artifacts noticeable more in high frequencies, but that's also more demanding on the cpu. It's become common to have an option for the amount of oversampling we want to use in software synthesizers, and virtual analogues also include that in their detailed specs. That's also another feature that requires a lot of processing power. The technology is there and it's easy to implement, I think, but isn't pushed to its limits because the developers have to compromise between quality and cpu hit. So as our processors got better, we get a better taste of what can be done with software. The other feature I find to be particularly cpu intensive is unisson. Then there's always multiplying all of this by playing many voices at once, of course.
Read somewhere that our gpus are still working on 32 bit and that 64 would be ideal for high end audio applications. Again, can't even approach if this is true or not.
But do you think it would be possible to have a real high end synthesizer, multi-timbral, with a very high internal sample rate, state of the art filters and all of that running from our graphics card?
|