when do you think processor power will end?

[Nocturnal]

like when do you think theyll be able to never get a cpu any higher?

what comes after 999ghz?

 

[Gyrene]

what comes after 999ghz?

1 terahertz

 

[Nocturnal]

Originally posted by: dtyn

what comes after 999ghz?

1 terahertz

ahh, i can see it now... in the year 2183 pentium 69839 1 terahertz breakthrough processor!

 

[tenoc]

Armageddon, Gotterdammerung, or later!

 

[MasterHoss]

They'll go well above 1 terahertz

 

[glugglug]

1 MHz processors were in use until the early 1980s.

Would 1kHz be around the 60s? not sure... but 1000x clock speed increase every 20 years seems about right.

So 1 THz will be sometime around 2020.

1 Petahertz (1x10^15) will never be reached for 3 reasons:

1) C/Unix time rolls over on January 18 or 19, 2038. This is a FAR bigger issue than Y2K, but not understood by most journalists.

2) 1 clock cycle at 1Phz = 1 ferrito second. Light travels only 0.3 micrometers (300 nm) in that time. Good luck installing memory modules somehow small enough that both ends are within 300nm of the CPU (what will happen instead is the burst reads/writes will be bigger and latencies will me measured in hundred thousands of clock cycles... but at that point ramping up the clock rate is pretty pointless).

3) Electromagnetic interference from a CPU at that clock speed would be UV. The FSB would most likely give off visible light. Gives all new meaning to "screen-tan".

 

[nortexoid]

i doubt processing power would need to be that high anyway...i'm sure distributed computing will be taken to an entirely different level...that is, u need processing power? - log into a distributed computing network that handles it for u for a small fee...perhaps no fee at all.

multi-proc systems will also become more commonplace in the future, i'm certain...

and then u'll have different components of a "processor" running at different frequences, and so rating processors by MHz will be utterly impossible...i imagine this isn't too far off.

 

[microAmp]

1 ferrito second.

ferrito, never heard of that, just made me hungy for some Fritos

 

[glugglug]

I think within the next 5-10 years we will start seeing ALL CPU lines become massively hyperthreaded, and rather than multi-proc servers, there will be a more heavily hyperthreaded single CPU.

Multiple separate CPUs would require higher latency to safely share a single system bus.

A 33 GHz CPU with registers in arrays of 32 (for 32 hyperthreads) will probably be marketed as "1.1 THz".

 

[Agent004]

Electromagnetic interference from a CPU at that clock speed would be UV. The FSB would most likely give off visible light. Gives all new meaning to "screen-tan".

lol, I guess it would be our next 'sun' bundle in some food and we have (or rather in ) a microwave oven

 

[nortexoid]

aside from the fact that SMP configurations perform faster than hyperthreaded software utilizing logical CPUs and a single actual CPU, the problem of latency can still be hidden, by, i.e., [larger] cache.

the only reason for hyperthreading to gain popularity is because of its cheapness...but w/ the price of procs nowadays, SMP configs are pretty cheap.

also, u could hyperthread & SMProcess at the same time...

EDIT:

also, hyperthreading relies much more on the programmers' end, and i imagine is far harder to code for

 

[MisterBombastic]

Originally posted by: glugglug


1 Petahertz (1x10^15) will never be reached for 3 reasons:

1) C/Unix time rolls over on January 18 or 19, 2038. This is a FAR bigger issue than Y2K, but not understood by most journalists.

2) 1 clock cycle at 1Phz = 1 ferrito second. Light travels only 0.3 micrometers (300 nm) in that time. Good luck installing memory modules somehow small enough that both ends are within 300nm of the CPU (what will happen instead is the burst reads/writes will be bigger and latencies will me measured in hundred thousands of clock cycles... but at that point ramping up the clock rate is pretty pointless).

3) Electromagnetic interference from a CPU at that clock speed would be UV. The FSB would most likely give off visible light. Gives all new meaning to "screen-tan".

And the Earth is flat, the sound barrier will never be broken, and man will never walk on the moon.

Never say never.

 

[McCarthy]

Sometime after Deep Thought proclaims the answer to the to life, the universe and everything is 42.

--Mc

 

[Utterman]

Originally posted by: McCarthy
Sometime after Deep Thought proclaims the answer to the to life, the universe and everything is 42. --Mc

But then we will have to compute the question to life, the universe and everything

 

[Daxxax]

1) C/Unix time rolls over on January 18 or 19, 2038. This is a FAR bigger issue than Y2K, but not understood by most journalists


uhhh whats that got to do with cpu speed??

 

[Scipionix]

But then we will have to compute the question to life, the universe and everything

Or else we shall become an interstellar highway bypass.

Even a frequency of 1000 GHz means one cycle every trillionth of a second, during which light travels 0.3 mm. So would you need a processor with no traces longer than 0.3 mm? You could have a processer that is a 0.3 mm cube, leaving you with a volume of 0.027 mm^3. Even if you could manufacture such a thing, how small would the gates have to be? It would probably have to be liquid nitrogen cooled too. If I'm not mistaken, optical lithography has lasted far longer than anyone expected 10 years ago. Intel says it will last until 2005, but after that what are the physical limits for silicon? Unless we have working quantum CPUs by 2025, it looks like we're screwed.

 

[staticfly]

screw that... organic cpu's are the way of the future

 

[Scipionix]

Yeah, if you want to do a 256 city traveling salesman problem right quick, but try playing Quake III on it. Well, at least you'll have someone else's nucleotides to blame.

 

[Maggotry]

At some point a newer technology will come along that'll make cpu speed ratings irrelevant. Something like a quantum computer. A new rating system will probably have to be used.

 

[amdskip]

never

 

[CrazySaint]

I don't know what speed CPUs will top out at, if they ever do top out, but I bet you AMD will still be using a 266MHz FSB

 

[docgonzo]

By the time CPU speeds get close to light speed, computers will have taken control of us, and we will all be living in the Matrix and not care...hehe Or perhaps there is another dimension that forgoes physical space making connections possible at faster-than-light speed? Or perhaps India will bomb Pakistan, and America will bomb the Arabs, and the Russians will bomb the Chineese, and then all of this non-sense will be irrelevent? OR, perhaps comets will come from the sky and destroy all rhetorical questions? GOD BLESS ZERO FRICTION!!!!

 

[yee]

Originally posted by: glugglug
1 MHz processors were in use until the early 1980s.

Would 1kHz be around the 60s? not sure... but 1000x clock speed increase every 20 years seems about right.

So 1 THz will be sometime around 2020.

1 Petahertz (1x10^15) will never be reached for 3 reasons:

1) C/Unix time rolls over on January 18 or 19, 2038. This is a FAR bigger issue than Y2K, but not understood by most journalists.

2) 1 clock cycle at 1Phz = 1 ferrito second. Light travels only 0.3 micrometers (300 nm) in that time. Good luck installing memory modules somehow small enough that both ends are within 300nm of the CPU (what will happen instead is the burst reads/writes will be bigger and latencies will me measured in hundred thousands of clock cycles... but at that point ramping up the clock rate is pretty pointless).

3) Electromagnetic interference from a CPU at that clock speed would be UV. The FSB would most likely give off visible light. Gives all new meaning to "screen-tan".

Its actually called a femto second. link

Anyways, by the time we reach 1 petahertz, quantum computers would have been available.

 

[Vernor]

Originally posted by: Nocturnal
like when do you think theyll be able to never get a cpu any higher?

what comes after 999ghz?

The Singularity.

 

[Scipionix]

actually, all you have to do is set up a static warp field in the 1500 millicochrane range. Then your data can move superluminally with ease.