First, it would be great to know what "optimal" means.

It's still a very general definition. In Control Theory (which can be a real pain in the butt), the optimal control of a given system is one of the most difficult problems to solve. However, it does introduce a useful concept: Cost.

Not cost in dollars, but a general cost that we want to use as the basis for our controller. Basically:

If a given controller minimizes the cost of the system at any point in time, we have achieved optimal control

.

For instance a "cost" could be time. The less time it takes for a mobile robot to navigate from point A to point B, the more optimal the robot's controller is. One of the most well-known optimal control problems in computer science is the Traveling Salesman problem:

For a given set of destinations and given distances between destinations, what is the path the salesman should take to minimize the time it takes to visit all cities?

In the case of controlling a mobile robot, time certainly would be a factor in optimal control. So might the amount of energy used by the system, or the risk of damaging the robot. A cost function is, in general, a function of time, the inputs to a system, and the state of the system.

So, can we apply this idea to PC building? We could define our cost function, literally, as the cost of the PC. The cheaper, the better. A given PC is more optimal than another if it costs less to build. But then we don't necessarily get good performance.

However, we also might want a high-performance system. More performance is more optimal. However, in the real world, high-performance PCs cost lots of money, and most people aren't willing to drop many thousands of dollars on one.

So clearly our cost function must be a function of performance and money. What about looks? How about noise? Temperatures? Power Consumption? Reliability? Performance over time? Value over time? Ease of part replacement?

The more factors you base your PC on, the more complex finding the optimal solution, or even a good solution, becomes. Maybe that's why some of us spend so much time worrying about small components in our system. Maybe our cost function should include another factor: design time.

Putting aside that though, it's clear that one PC is not going to be good for everything:

You can't have an awesome gaming rig AND one which sips power, because high-end graphics cards draw lots of power.

You can't have an awesome gaming rig which is cheap, cool and silent, because paying for watercooling equipment is expensive and also makes upgrading more of a hassle than with air cooling.

You can't have a professional-grade workstation that is also cheap. Workstation-grade parts cost a lot of money, and for good reason.

You can't have a

professional-grade workstation that is also great at gaming. Workstation-grade GPUs aren't nearly as good as consumer GPUs at gaming, and consumer-grade GPUs don't have the same level of reliability or 10-bit color output, nor have the same level of software support for professional applications.

Computer hardware updates quickly and gets new features and performance boosts. Your standards of performance will not be met by future games, as your top-end GPU will be replaced by new top-end cards. Upgrading is expensive, even after selling your used card.

Waiting because you can't decide if you want a rumored feature of whatever CPUs or GPUs are coming next increases the time it takes to design a new computer. If design time is part of your cost function, you should build sooner rather than later (this is a bit tricky though, since release dates change and you might want to build a new computer at the same time a new component happens to come out).

It's not easy to design the "optimal PC", but hopefully you'll be smart and build a good one, even if it's not perfect. After all, is it really so necessary to be optimal? Go enjoy your new PC.