5.6 ghz?? impossible!!

[BurntKooshie]

I think the funniest thing is that this so-called 3ghz limit doesn't exist - Intel has parts of the ALU's running at 3.6ghz with the release of the 1.8Ghz P4. IBM has demonstrated a 210Ghz (yes, two hundred and ten gigahertz) transistor. Take a loook. So while the 3.0 Ghz limitation is a theoetical limitation, considering that it has already been broken, is there any reason for it to continue to be a theory?

 

[yaethom]

Well the 3ghz limitation arises from simple math actually.
However, you will have to note that this calculation is based on a .25 micron manufacturing process. The smaller the manufacturing process, the higher it will raise the limit.

It's true that it is an old theory, but what I am getting at is not necessarily the how high their GHz can go, but how they are doing it.

Lemme whip out a calculator and I'll get back to answering that post.

...

 

[yaethom]

Okay, it looks like I need to brush up my physics, I'll get back to you guys as soon as i double check this with a couple of colleagues.

And you do have to keep in mind, the 3 GHz limitation is based around some older technologies.... you will see.

 

[rc5]

It's also amazing why Microsoft can always exhaust every single bit of cpu power you throw to it.

 

[BurntKooshie]

I admit I know little EM physics, but you made that statement with blatent ignorance of the fact that Intel is using .18 micron for the P4, and achieving >.18. They never claimed to be getting 3.0ghz on a .25 process....

Why would you then, be skeptical of 5.6Ghz being attainable, considering Intel is moving towards smaller processes, and you yourself know it? While you may have the math 'n physics to back you up, you didn't bring that out, and even though you may, it sounds somewhat troll-like to make statements that 5.6Ghz is impossible. Especially because you appear to know better, given your math knowledge!

 

[Rado]

Obviosly you do not read the news. 3 GHz is a joke. They are talking 20Ghz at IBM...

READ: http://www.electic.com

If you read the news, you would have known that.

 

[Lucky]

"5.6 ghz?? impossible!!"

uh yeah, didnt they say the same thing about hard drives? back when there WAS NO hard drives? just floppies? then we moved to 10 meg hard drives. My first "advanced" computer, a mac SE, has a 40 meg hard drive. I never thought I would need more. Do you really think anyone at that time thought that there would be 1 gig hard drives holding 25x more than that and were scarcly the size of a matchbook? I bet if you introduced that idea, people would have said "matchbook sized gigabyte hard drive?" Impossible!!!

Get over it. I dont see ANY limits in the near future, and i certainly dont see it stopping anywhere near 5.6 ghz. 10-15 years down the road I certainly can envision 100 gig processors, tho Im flabbergasted by it, just as those who has 1-4 gig hard drives 5 years ago were im certainly flabbergasted by the possibility of 100 gig hard drives like we have now...and how we are flabbergsated by single drive 1 gigabyte HD's in the future now, but which will certainly become reality in the near future.

 

[bigd480]

maybe i'm looking for too simplistic an answer (i'm a business major not an engineer)...

but:
the point of raising the speeds is to get more IPS (instrx per second)... suppose you made a P4 that's twice as big as a normal one -like just doubling the amount of everything... and then you used supercooling via peltier or cryogenics or whatever... wouldn't you be close to the performance of a 5ghz chip?

 

[smp]

I think eventually we won't worry about how fast our processors are.. when are 'quantum computers' coming out? Should I start saving?

 

[Sohcan]

<< suppose you made a P4 that's twice as big as a normal one -like just doubling the amount of everything... and then you used supercooling via peltier or cryogenics or whatever... wouldn't you be close to the performance of a 5ghz chip? >>

I'm not exactly clear on what you're saying, but...

If you were to double the number of transistors by doubling everything in the P4 (number of execution units, number of instructions issued/retired per cycle, reorder buffer, size of register file, size of trace, L1, and L2 caches, etc), you would probably lose performance. First of all, it would probably decrease the clock rate, since it would increase the critical path length due to an increase of the size/number of ports of the register file and reorder buffer, assuming the transistor technology and number of pipeline stages remains the same. Secondly, x86 lacks instruction-level parallelism, largely due to its variable instruction size (1 to 15 bytes) and the large number of addressing modes. Beyond 3-way issue superscalar, x86 processors quickly experience diminishing returns, so a 6-way issue CPU would not be much faster than a 3-way issue CPU, despite drastically increasing the number of transistors.

An increase in the number of transistors has to be used efficiently to experience any increase in performance. x86 processors have used as much RISC techniques as possible, such as pipelining, register renaming, and out-of-order superscaler execution. Besides increasing clockspeed, the best bet for increasing performance for x86 CPUs lies in thread-level parallelism, such as on-chip multiprocessing and multi-threading.

 

[nortexoid]

asynchronous achitectures remove the MHz labeling of CPUs since different parts of the core run at different internal clock speeds....this is most likely the future of CPU architectures...

ways to bypass any theoretical limits:

1) quantum computing - adhering to the laws of quantum physics of course...
2) analog processors...non?

 

[UNCjigga]

Hey, do you know anything bout that 2.0GHz Pentium III in the works???
.
.
.
.
.
.
.
.
.
.
.
(laughs hysterically!)

 

[cvlegion]

Look at the P6 core, it went from the Pentium Pro (150MHz) to the Tualatin and who knows how high that will go. It is all a matter of shrinking the die. Increasing the length of the pipeline and being a completely redesigned core, I am not surprised that they expect to get up to or over 10GHz with the core (revised of course). Intel is not going to put billions of dollars and years of effort into a project just to scrap it the next year. They are working on their next project which will come out who knows when (5 years?).