What is the reason for CPUs being so small?

The more transistors a CPU has the faster it is right? We seem to keep trying to make the transistors smaller and smaller yet the size of the CPU remains the same, a few cm squared.

Why isn't the size of the chips ever doubled allowing twice as many tiny transistors on to the chip. Surely this would be a less expensive way to create a faster chip? Or will this create performance and/or manufacturing problems?

Solution 1:

Well, the larger a single die is, the lower the yield of usable processors - ANY flaw would result in a useless or crippled chip.

Multi die processors are possible, but not cost effective - the pentium pro was THREE dies in a package, and had horrid yields.

I also believe heat dissipation would be affected - you'd need bigger heatsinks to dissipate the increased heat from the increased transistor counts.

Solution 2:

a long line from one transistor to another makes the switch time (and clock speed) slower

this is because the long line acts as a capacitor that needs to be charged above a certain level before the connected transistor switches

and many lines in parallel can interfere with each other on high frequencies (the main reason parallel connectors are disused)

and you can only dissipate so much heat with normal heatsinks (the reason why the processors are flat is to avoid heat buildup inside it)