The way I see application of Intelligence to Magic in D&D, where wizards need to re-memorize spells every day before being able to cast them is that Intelligence is closely tied to learning ability and retention. All spells have this magic(duh) property about them that they're volatile, i.e., it's hard to keep them in memory, they constantly try to escape it and escape it completely, full bit overwrite, particularly when they are actually being recalled -- during the moment of casting. So, individuals who have a better ability to absorb and retain knowledge are better at resisting this effect. This ties in neatly with the notion that wizards are scholars who constantly exercise their mind.
Other sentiment I've seen here is that the intellectual nature of magic often is not reflected in gameplay, mitigating spell preparation and casting to almost purely manual labour. In some games I've seen this being mitigated a bit by allowing to combine spells for a different effect. This reminds me of shell scripting, where you usually chain together simple programs made by others that do specific stuff, having learned about how to use them and what effects can be achieved with them, but usually not delving into how exactly they are implemented.
I've also been entertaining in my head the idea of advanced magic mechanics (basically, reverse Clarke's law [http://tvtropes.org/pmwiki/pmwiki.php/Main/SufficientlyAnalyzedMagic]) where you design spells yourself and fine-tune them to the resources available and situations where you will use them. Something in the direction of Spellcraft: Aspects of Valor, where you create spells by combining alchemycal ingredients in specific proportions and a magic domain, but even deeper. I'm thinking of wizards being able to create and manipulate atomic 'elements' that have specific properties about them, but don't do much by themselves and building spells out of them.
For example, to create an energy bolt you first need an element that will receive the mana and store it for consumption. Then you need an element that converts mana to explosive force and wire it to the previous one to receive mana from it. Since the resulting spell would just immediately explode in caster's hands, you need to add a mana-impeding element between the two previous ones, which can control the flow of mana. To this one you need to wire up a two-element combination: a shell that envelops the spell and a sensor element wired to it so that when the shell breaks, the mana impedance gate is opened. Now the spell explodes on sufficient impact.
Now what's left is to put it to add another output to the mana storage element -- an element that will keeps converts mana to momentum, so that it's launched off the caster's hand. But the spell's not yet ready. One last step is to add another mana impedance gate between the mana storage element and the momentum element, so that the mana flow is limited to specific amount per time unit, limiting the maximum momentum and leaving some mana in storage. Now the spell is ready! The wizard creates all the elements, which naturally requires time and mana. Upon finishing, the caster starts pooling mana into the mana storage element, which in turn supplies it to the momentum element and the spell takes off, presumably towards the target. As the spell hits the target, the shell breaks and mana impedance gate to the explosive element is released, causing an explosion of magical energy at point blank range!
Since each element requires mana just to create it and its attributes(i.e., shell hardness, maximum storage) depend on supplied mana and after some amount of time the element will just dissolve if it not supplied additional mana to sustain it, designing spells would require careful consideration. This of course wouldn't be the past-time for everyone, but you could make it possible to trade spell blueprints and enchant items with them, allowing those not interested in intellectual exercises to enjoy their effects, while rewarding those who are with money and combat advantage.
A far as the platform for implementation of the mechanic goes, I think it's too complicated for a pen&paper game, so a computer is an only option. Computation of spell behaviour with these mechanics sounds to me like a large overhead, so most likely visual would have to be greatly sacrificed, possibly even other aspects of gameplay, even if some hard limits on spell complexity are imposed. One idea for optimisation I have is to heuristically calculate spell effects from some number of 'training' casts and after that apply this 'pre-calculated' effect in future instances of casting, without computing the whole mechanics again. Not sure how well this would work with non-trivial spells.
