First of all, going from 45nm to 32nm means that every transistor takes up half the space it used to. The choice then is between the same number of transistors per chip resulting in lower per unit cost or twice as many transistors per chip resulting in better performance. As usual, there will be some of both.

Some people need better single-core performance, some people need more cores, and some people just need lower power consumption. Not everyone needs the same thing, which is why there are different product lines (Server, Desktop, Mobile, and Netbook) each with different models.

Anyway, I'll go back to waiting for my program to finish; it's been almost 10 hours so I obviously fall into the needing better performance category.

The problem is that every large company (like banks) don't sign contracts with their customers. They have policies (which are impossible to find online by the way) but they change them and then claim they told you and then you're stuck paying $10 or something that's a pain in the butt but not worth fighting about.

And no, the free market obviously hasn't fixed it.

I've always wondered why bothering writing over a disk 5 times with zeroes. Generally, the read out from a bit is something like 0.980 or 0.020, which is obviously a 1 or a 0 (respectively), but look like they used to be a 0 or a 1 (respectively) before the last time they were overwritten or else they would look like a 0.995 or 0.005 or something. So wouldn't it make more sense to alternating writing over with 0's with 1's so that it would even harder to recover.

On a similar note, as a grad student in physics all the computer code I generate belongs to the university. However, most of the code I've written is either too short for anyone to care, or an expansion of something I GPLed before I became a student. So as an employee I can modify that code a whole bunch, but in the end it's still GPLed.

At least, I think so since IANAL.