Yes, questions about how to achieve or improve effect X should be on topic, but we should hold them up to certain quality standards:
They should answer the question "What have you tried?"
We should hold them up to the same quality standards as other SE sites like Stack Overflow (or as you say Graphics Design). If we don't require question authors to show some effort, we'll just open ourselves to "Do my work for me" questions. This can include a picture of video of what the author has already achieved but code would be even better.
Such questions should be specific.
What kind of effect should be achieved and which characteristics are particularly important. Are we talking about stills, offline animation or real-time animation? Should the solution fit in with an existing renderer or not? We don't want questions like "How do I render smoke?", because they are way to broad for this Q&A format and you could write an entire book about it. If there are several important components to an effect you want to achieve, consider splitting it up into several questions.
I think these questions (and their answers) can be great and useful content, because studying them should hopefully enable people to come up with their own effects more easily. But that doesn't mean we can't ensure the usual quality standards for those questions.
(Private beta note: I'm willing to give some leeway for point 1, because obviously many of the questions right now are made up or remembered, because there's only so many "real" problems a handful of private beta users will encounter during those two or three weeks. Right now it's more important to cover as much on-topic ground as possible. But in return I'd expect questions to be that much better regarding point 2.)