July 19th, 2002 01:15 PM
Great paper Xmadness, wanted to suggest something about it though if you push it out into a formal tutorial. You started to hint around at what I am about to say but then moved on into other things. Some stuff that would be nice to see in the article are a couple of other things that can make or break chips/chipsets (and is a large part of why AMD can hang in there with lower processor speeds). You have other variables in there besides how many things can be done with one cycle of the clock (although they are somewhat tied to a cycle of the clock). The biggest bottleneck on a PC right now is not the processor, it is the speed of the memory and the bus. Unless you have programs that are specifically dedicated to calculations that take advantage of the processors onchip memory and cache, the fastest processor in the world will still have to sit and wait for writes/reads/scans of memory and will still have to do this by accessing it across a bus (usually of the same speed as the memory). This also leads to another thing that can make or break a processor and that is its onboard memory/cache. The bigger and more efficient it is, the less often it will have to slow down by going out and accessing main memory. Intel released a version of the celeron with a very tiny L2 cache and it wound up being a dog on performance because it constantly had to slow down to go out and access main memory. You did mention this roughly by adding there are other things in the picture like memory, graphics gard, etc, but it would be nice to see a brief explanation of why that is so and to have it tied together.
Also, although it has been a very long time since my VLSI classes and I don't remember the exact formula, the single biggest factor in the power consumption of an integrated circuit is that frequency at which it operates. The equation is something like P = nf^2, where P is power, n is some other part I forget, and f is frequency. So as you increase frequency, you increase power consumption by the square of frequency (as frequency becomes very high, n becomes roughly irrelevant). And of course, as you increase power consumption you increase heat. There are some neat tricks you can do with integrated circuits like changing the interconnects, lowering the voltage (effects power, many circuits are moving from 5V to 3.3V, if they haven't already, but you have problems with interrpeting digital 1's/0's cause the 3.3V pulse is more effected by various types of noise), etc that can lower the power consumption, but it is still as far as I know the overriding problem in integrated circuit design. And on a final note, many manufacturers are now looking into materials alternative to pure silicon in circuit design that can withstand higher tolerances of heat, a couple of which are gallium-arsenide and silicon-carbide.
Excellent post Xmadness, greenies for you.
There is only one constant, one universal, it is the only real truth: causality. Action. Reaction. Cause and effect...There is no escape from it, we are forever slaves to it. Our only hope, our only peace is to understand it, to understand the 'why'. 'Why' is what separates us from them, you from me. 'Why' is the only real social power, without it you are powerless.
(Merovingian - Matrix Reloaded)