close, but not correct. the number of transistors on a similar sized chip is expected to double every 18 months or so, that may not necessary mean more mhz. in fact, mhz has fallen since the p4 3.4ghz days (not its 2.xghz)

and this rule is close to limit on current silicon tech. come new breakthrough is needed to continue.

another limiting factor is human. if any of you have a 9" netbook, you can see that the keyboard is more or less unusable. for human ergonomics i noticed that 12" seems to be the minimum. same issue with screen size. so unless you come up with new human machine interface method, the physical size of computer wont go down much.

although iphones and other smartphones may evolve to replace portable computers (provided you dont need to type much)

can i crap here? LOL its easy. Find a way

sometimes the processor are getting smaller adn smaller.... newscientist megazine once report that scientist are using spin of electron as proccsor

Hi, you are referring to Moore's Law. The law is named after Intel co-founder Gordon E. Moore.

Foundemental of transistor based CPU is silicon (sand) as it exhibits conductance depends on thresholding of the gate, hence the name semi-conductor.

As mentioned by our friend forumer in post above, it has limitation.

When the gap between the collector and emitter is too short (to make it smaller), election tunneling occurs which is uncontrolable independent from the Gate. We certainly do not want a processor giving different values/results as it act randomly.

Second consideration is the cost. The closer the gap of the C and E, the production yield (the success rate on producing a good CPU) would be more challenging/difficult. Cost increase as a result.

The reason why PDA can be made such small in size as the CPU inside is generally of lower spec CPU as this device is not as power hungry as those engineering research workstation.

Off Topic a bit. Which company owns the most powerful and smallest CPU?

The answer is not a company, it is the Pentagon (Military Research Center).  All the "consumer" grade CPU are a few generation behind what is being research and used in military.

The internet and advanced telecomunication system were the branch out of militory project. Of couse a few generation behind....

I hope am not stating the obvious 

take a look at mac book air, they managed to stuff cpu, 2 gig of ram, built-in graphic chip, etc etc, on a board thats about the length of a pencil. most of the space in modern laptop is taken up by the battery. the weight by battery and disc drives.

and yet the mac book air is still 13" diagonal. thats due to human interface requirements. until you have other form of interactions that is practical and can be used easily (hud built in glasses, voice control, 3d motion sensor, etc) you cant go much lower than 12" diagonal.

so while components will of course get smaller and faster. laptop size is mroe or less a plateau now. desktop may shrink to shuttle/all-in-one size.

Lemme put my 2 cents in. The the EE business there are a few ways to make things smaller and in this case, computer devices smaller. Granted digital electronics and microtronics isn't my subfield, but i do know how things work. Although you can PM Ikanayam if you want to know more lol since his field is dedicated to microprocessors.

Either way, the first way to make things smaller is markedly to cram more shite into a smaller area, which means more processors into a singular microprocessor. You don't use IGBTs for switches, FETS with individual logic cells are used. Anyways, but making the production process smaller and multileveled, you're essentially cramming more transistors into the same area, thus you can make smaller chips.

The limits of this are dependent on the materials used, the gate construction, the routing of the microwires and so on. But generally you get the picture lah.

Secondly, to cram more shit into stuff, you need to integrate more discrete components into the system, or increase route density so that more options exist when it comes to component placement. The easiest way is to stuff more conductive layers into a PCB (circuit board), so that routing becomes easier and the physical layout can have it's density increased.

Thirdly, you always have the problem of heat, cramming so much into such small areas always gives you a problem with heat. By adopting either lower powered components or adopting better cooling, technically, the component density around "hot spots" can be increased, giving you a smaller package by default.

That is only the time difference isn't it ? Tiny container will heat up faster but given a sufficient time, heat emitted by big candles will surpass small candles when thermal equilibrium is achieved.

What really matters is how the chips are constructed and designed to work. A big chip that is designed to run maybe 10000 tasks should be hotter than one small chip which is just running a single process. Of course, a chip in a small computer is going to run more processes than the bigger computer since you can only have limited chips on a small board of a small computer ... maybe this is the answer to my question 

Just my rough guess