StarDestroyer.Net BBS

News.com.com reports that a recent Intel study indicates the imminent cessation of chip density increase.

http://news.com.com/2100-7337-5112061.html

So it seems we won't be squeezing many more transistors into a chip, I suppose we need to start working on new architectures and such, perhaps something using neuronal based logic?

Still like 16 yrs or so. Worry bout it then.

I suppose we need to start working on new architectures and such

Whattaya mean, "start"? Researchers have been delving into the ideas of DNA-chips and quantum processors for years.

This may be a really stupid question but...

If we are running into a limit for how dense we can make the chips can't we simply increase the size to increase processing power? Processors right now are bloody small, doubling or even trippling the size would seem to me to be fairly trivial. Is there any reason not to do this?

Luke Starkiller wrote:This may be a really stupid question but...

If we are running into a limit for how dense we can make the chips can't we simply increase the size to increase processing power? Processors right now are bloody small, doubling or even trippling the size would seem to me to be fairly trivial. Is there any reason not to do this?

Well, in the article itself:

researchers are exploring a variety of ideas, such as more efficient use of electrons or simply making bigger chips, to surpass any looming barriers.

Luke Starkiller wrote:This may be a really stupid question but...

There's no such thing as a stupid question. . . only stupid people.

If we are running into a limit for how dense we can make the chips can't we simply increase the size to increase processing power? Processors right now are bloody small, doubling or even trippling the size would seem to me to be fairly trivial. Is there any reason not to do this?

Increasing the size of a chip allows you to put more transistors on the chip yes, but it won't make them operate faster. The speed of chips is greatly limited by both transistor size, and the length of interconnects between parts on the chip. As the length increases, resistance goes up, which slows the performance as current has to be dumped into the transistors to make them switch. They act as capacitors, so you get an RC time constant that has to be dealt with. Smaller transistors reduce the capacitance to ground at their inputs. Smaller transistors also tend to switch more quickly than larger transistors, and at lower voltages.

With longer interconnects you need more power to get high speeds, and next generation chips are already looking like they'll need more than 100 Amps at 1 Volt.

In short, to make computers faster, you have to make the parts smaller. Making them bigger lets you put more on the chip, but they all run slower.

Huge chips also produce far too much heat and consume far too much power, due to the aforementioned electrical resistance issues.

Ah, that makes sense, thank you.

So in other words, baring other break-throughs, the only way to make a computer run faster is to run chips in tandem?

BlkbrryTheGreat wrote:So in other words, baring other break-throughs, the only way to make a computer run faster is to run chips in tandem?

You can run chips in SMP or NUMA mode, but that costs more money, nevermind the heat issues. You can also put multiple cores on a chip, but that also is expensive. There's various other practical limitations as well (volume constraints for a case).

Graeme Dice wrote:
There's no such thing as a stupid question. . . only stupid people.

"There are no stupid questions, just lots of inquisitive idiots" sounds much better in person.