Endless Paradigm

Noting that a 2012 Core i7 3770K is only a tad faster than 2x a 2006 Core 2 Q6600, it seems that speed increases are slowing down (becoming more difficult to achieve perhaps).
Although Moore's Law still seems to be operating, speedups don't seem to be proportional.  I decided to plot some speeds of Intel x86 CPUs taken from this wiki page:



I've kept to single core and quad core chips to make comparisons a bit easier, and snuck in the current consumer-grade top-of-the-line 6-core i7 3960X to give a feel of where the top CPU is at.

Wee've pretty much hit the "GHz wall" years ago, and have pretty much picked all the low hanging fruit in terms of instructions-per-clock optimisations (As far as I know).  Cache takes up like around 80% of modern CPU die space.  22nm is like 50? atoms wide.  Multicore has its limitations in the amount of speedup that can be achieved.
How much faster can wee get?

just wondering why don't they increase core die size.
they went from pentium 4 with hyperthreading to core2duo based of the pentium 3's the the I7's line incorporated Hyper threading and dual-quadcores.

Now die sizes are getting smaller and smaller instead of aiming for smaller dies and more cores they could just aim for less cores and higher clock speeds then work their way up to having more cores.

The reason wee're going into multicore is because you can't increase the speed of a single core fast enough.

Interesting point.

Also, it's a pretty amazing view of the progress processors have made over the years. (one way or another, anyway).

I don't understand why they don't just make CPUs bigger with all the fast small spoon on it

If you look old CPUs they were the size of CDs
why not take modern CPU technology and see how much of it you can fit on a CD sized CPU

it's more about the optimisation of process scheduling and CPU time management these days that makes processors faster. amount of cores a bit, raw clock speed not nearly as much as it was before. not to mention they're trying to make all of these things and reduce power consumption, which is usually by printing smaller (less not much). as well as using less not much makes the CPU smaller, and produce a lot less heat.

Slushba132 Wrote: [ -> ]I don't understand why they don't just make CPUs bigger with all the fast small spoon on it

If you look old CPUs they were the size of CDs
why not take modern CPU technology and see how much of it you can fit on a CD sized CPU

A long time back you said the exact same thing, "why don't wee have CPUs the size of pizzas", and I told you you're retarded back then, so I'll tell you now again, you're retarded.

Last time I gave you reasoning and stuff, this time I can't be stuffed.

not that i know anything about processors but from an electronics point of view wouldn't you be theoretically limited by how fast a transistor can switch...

Seems to be a greater focus on efficiency and spreading the workload out, rather then just pure power with no value.

Slushba132 Wrote: [ -> ]I don't understand why they don't just make CPUs bigger with all the fast small spoon on it

If you look old CPUs they were the size of CDs
why not take modern CPU technology and see how much of it you can fit on a CD sized CPU

I don't think making it larger really makes it any faster.  Sure, you can stuff in more cache, but you get diminishing returns (plus costs skyrocket).  Otherwise, what could you really use the extra die space for?

andrewcc Wrote: [ -> ]not that i know anything about processors but from an electronics point of view wouldn't you be theoretically limited by how fast a transistor can switch...

Clock speed?