G4-power said:
This major technological innovation might be just behind the door. I read about it at our local library, it wan in a Finnish computer magazine (Finns are quite big on technology, don't mean to brag, but Linux is made by a Finn, and Nokia phones have started from Finland). Anyway, it had an article about some new laser-technology, that would allow for 38 nanometer process. There was some comparing to the current 90 nm process, and the new technology was much simpler, and more accurate.
They couldn't give any accurate dates when this new technology will be usable, some 2006 probably or later.
That's a pretty interesting thing, but I think we'll be seeing the 65 nm process way before that.
For a good read of the briefs of die shrinks and physics behind it refer to :
http://www.sudhian.com/showdocs.cfm?aid=610&pid=2294
As well, i'll quote for you to save you some time one important part:
Anatomy of a Die Shrink:
When we refer to a CPU as 90nm (.09 micron), were referring to the space between CPU traces. To put this in perspective, human hair varies between 40 microns and 120 microns in thickness. To calculate how much of a difference this is from existing 130 nm, we take the square of each, arriving at a shrinkage of 47%. In other words, moving from 130nm to 90nm more than halves the size of the gap between CPU traces.
The reason this does not translate linearly to a direct CPU die shrink is because not all components in a CPU shrink by this amount. In this case, Winchesters approximately 84 mm sq die is roughly 61% the size of Newcastles 144 mm sq die. This is obviously still a significant reduction.
Shrinking the gap between the CPU trace lengths, however, has several effects.
Increased Thermal Leakage: As the gaps between the CPU traces shrink, the amount of current that leaks out of the transistors increases. This translates into heat, which translates into a hotter-running CPU.
Increased Thermal Density: This is a key factor that cant be overlooked. As the surface area of a CPU shrinks, the amount of heat that has to radiate out of that area does not. This means that, all else being equal, a 130nm CPU has a lower thermal density than a 90nm CPU. It also means that a smaller chip runs hotterall else being equal. Up until now, all else hasnt been equal, which is why weve seen the improvements that we have.
Decreased Operating Voltage: Die shrinks typically allow for lower operating voltages because less power is required to bridge the smaller gap. Note, however, that thermal leakage can work against thisif power is leaking out of the transistor, obviously voltage cant be lowered by as much as if the transistor leaked less.
The answer (in very broad terms) to why we havent seen the 90nm problem before now is because the positive effect of being able to lower operating voltages has outweighed the negative effects of increased thermal leakage and thermal density. Other technology upgrades (such as improved substrate technologies) have also helped. The question is, have they helped enough?
SOI and Efficiency vs. Netburst vs. Physics:
For all that reviewers have compared them to each other, AMD and Intel both are finding their approaches to computing tested by a third foe more implacable than either corporation could ever bethe laws of physics. Unfortunately for both companies, Physics is not impressed by marketing terms, does not care about full color ads, and is uninterested in either corporations bottom line. The question, in this case, is whether or not AMDs decision to bet both on a more-efficient approach to computing as well as IBMs SOI technology has paid off.
