Wlakiz said:
HappySqurriel said:
| Pemalite said:
Clock rates are the frequency at which the transisters etc' operate at.
We haven't hit a "practical limit" thus far, new technologies and techniques are always being discovered which can improve how quickly a transister switches and the frequencies that they operate at. - For example, I remember reading a few years ago of a 100ghz transister.
Take the Intel Atom. - It is actually paired with low-powered transisters, these don't scale in frequency to well but they do save on power consumption.
The Core i7 series however uses transisters which scale in frequency far more aggressively, however they will and do use that little bit more power to pull it off.
Also, extreme overclockers managed to break the 8ghz barrier on the new AMD FX chips, so that 5ghz wall was effectively smashed, a few more die shrinks and maybe the 3D transisters may even improve that situation for stock clocks. (Global Foundries is also working on 3D Transister tech.)
There is allot to CPU design, more than what most people realise, you can watch how a CPU is made here (Dumbed down of course and not showing any architectural stuff.) -http://www.youtube.com/watch?v=qLGAoGhoOhU
|
What I was actually talking about when I mentioned that we had hit a practical limite of clock speeds is that all major chip manufacturers have hit barriers to making commercial processors that run much faster than 4GHz. Consider that most architectures saw a doubling of their clockspeed every 18 to 24 months for decades and for the past 7 to 8 years desktop CPUs have been stuck in the 3GHz range. The reason that is most often cited for this stalling is that running these processors at higher speeds makes them run far too hot and makes them too unstable for comercial release.
While I have no doubt that single thread performance will continue to increase, the bulk of increased performance over the past decade has come from the increases to multi-threaded performance; and I suspect this will continue for the next decade.
|
No, they simply stopped scaling through Clock speeds is simply because it provides no added benefits. In order to make use of a 4+ghz clock, you need a larger pipeline otherwise many of the cycles are just sitting there doing nothing and creating more pipeline stages are pretty capped out because you will eventually get stalled anyways while waiting for things like memory write to complete.
Heat and Power consumption create different issues where the clock timming gets messed up and create errors but it can be remedied to with additional cooling which again adds to the increased power consumption (Yeah, you need to use energy to remove energy).
SImply, they stopped creating high clock speed is due to poor efficiency, not technical debt.
|
I'm not really sure what you're trying to say ...
My point was that, for some reason that is only really clear to the engineers inside of IBM, AMD and Intel, it has become impractical to produce commercial processors that exceed 4GHz; and have instead focused on increasing performance by increasing the number of threads that can be run. I have seen multiple explainations for this from people with a far better understanding of the physics involved than I have, and I suspect most of the people on this forum have, and often the reason surrounds the RC Time Constant, Settling Times and other issues that are far beyond my understanding; but in most cases the issues boil down to being unable to handle the excessive amount of heat in a way that is viable for the average consumer.
Realistically, it doesn't matter what the issue is all that matters is that chip designers have moved away (at least in the mid term) from pushing clock-speeds to increase single threaded performance. Now the focus is on increased parallism, both within the CPU in the form of pipelining or through increased cores.
Currently Online 
