25GHZ Processors in 10 Years???? Yeah Right! [Archive]

GnomeProject

07-08-2003, 04:47 PM

Is this article for real or is the register a bunch of hogwash like many have told me before??? I can't imagine Apple will be able to obtain such huge speed increases, but then again, it might be a blessing in disguise depending upon your point of view....

http://www.theregister.co.uk/content/39/31594.html

-Adam

nerrollus

07-08-2003, 04:55 PM

Just picture all the people sitting around with 386's 10 years ago saying, "3ghz processors! Yeah right!"

The industry is moving fast, and it seems like it just keeps moving faster. Honestly I would think a 25 ghz processor would be slow in 10 years...

RWiggum

07-08-2003, 04:57 PM

Considering that we have seen a roughly 100-fold increase in processor speed in the past 10 years, a 7-8 fold increase in the next ten should not be that difficult to imagine, even if Moore's law starts slowing down. I would think we should expect at least 100-200GHz processors in ten years.

nerrollus

07-08-2003, 05:02 PM

I'm going to have a dual terahertz machine with 3 terabites of RAM, and a 500 terabite harddrive, running Linux. Yeah, it will be sweet..

Icarus

07-08-2003, 05:15 PM

Moores-law states that processor speed doubles every 18 months, this gives (roughly) 6.5 cycles in 10 years. Using 3Ghz as a base...

288000Mhz CPUs
or
288Ghz

But the top limit of "Current CPU technology" limits it to 10Ghz before heat disentigrates the chip. Unless another technology is developed we're not going very far from here...Quantium Computing is looking very promising :)

they did say this about the heat 10 years ago and we'd never get to or over 1Ghz, so...:rolleyes:

DMR

07-08-2003, 05:17 PM

Originally posted by nerrollus
Just picture all the people sitting around with 386's 10 years ago saying, "3ghz processors! Yeah right!"

The industry is moving fast, and it seems like it just keeps moving faster. Honestly I would think a 25 ghz processor would be slow in 10 years... Agreed- 10 years is a loooonnnggg time in term of computer/processor evolution. I wouldn't be at all surprised to see 25GHz machines in 10 years, and I would be surprised if they were still were still silicon-based.

hlrguy

07-08-2003, 05:24 PM

I would be skeptical as well. Electrical impulses in a pure substrate would move 1.2 cm between clock cycles. You have to allow for impure substrate AND you have to have at least a 5% stable time before the next cycle, so I seriously doubt that chips will be reduced to ~1.0 cm square. It doesn't do a whole lot of good to run a clock at 5 times the rate that an impulse takes to travel from one edge to the other. You simply have 5 'waits' before you can perform the next useful operation. At best, I think they will find a way to multi-thread multiple CPUs in a single chip (eventually) that give equivalent of, say the GXYZ4 (future chip running a 5 gig) running at 25 gigahertz.

hlrguy

nerrollus

07-08-2003, 05:35 PM

Man on the Moon?! HAHAHAHA!

Shoot.. I'll never be able to fill up this new 520 meg hard drive!

What the hell would I need more than 2 megs of RAM for?!

Automobiles? HAH! They'll never catch on!

Sorry, too many lines like that have been said in the past for me to say them now. I seriously would never put a limit on technology.. Given time, I honestly think people can build technology for just about anything.

RWiggum

07-08-2003, 05:35 PM

I've even read that the next "big thing" in CPU's may be asynchronous chips - no clock at all. The analogy is that a clocked CPU is like a bucket-brigade putting out a fire. At each beat of the clock, everyone in the bucket-brigade moves from facing left to facing right, and if there is a new bucket coming, it changes hands. The buckets are like a bitstream - if the memory or bus or whatever can't keep up with the clock speed of the CPU, the bucket brigade keeps working, but water isn't moving down the chain. An asynchronous clock would be more like everyone working at their own pace to get their own water. That way everyone is always working and water is always moving, even if some of the workers (transistors, calc engines, whatever) are slower than others.

At that point, the clock speed of the chip is no longer a measure of performance. Of course, it's a poor measure these days anyway - even between AMD and Intel's 32-bit x86 chips, much less a POWER, UltraSparc, or G4. The marketroids will have to make the masses swallow some other spec, like GigaFLOPs or MIPS (or Tera, Peta, etc. - add any power of ten you like), which are equally as uncomparable between architectures.

Icarus

07-08-2003, 05:51 PM

You know they will never let the "almighty MegaHertz" die...Intel will still tout that they are the best and fastest because they have more MegaHertz, and everyone will believe them because the will continue to shove it to people in their mass-marketing campaign.

In 10 years...

AMD "Our processors can run 600 processes per clock cycle"
thepeople "huh?"

Intel "Our processors run at 10 Billion GIGAWATTS!!!"
the people "ooOOooo"

DMR

07-08-2003, 05:52 PM

Originally posted by hlrguy
I would be skeptical as well. Electrical impulses in a pure substrate would move 1.2 cm between clock cycles. You have to allow for impure substrate AND you have to have at least a 5% stable time before the next cycle, so I seriously doubt that chips will be reduced to ~1.0 cm square.That's exactly why I said I would be surprised if "chips" are still silicon-based in 10 years. Eventually we'll just run up against the immutable physical limitations of current semiconductor materials.

GnomeProject

07-08-2003, 05:55 PM

I love sparking great discussion like this..it's not that I don't believe it will happen, but I think Moore's Law is a pretty good guideline of what's to come and forsee a major slowdown (ie. Intel can't even keep their new 3.25ghz. P4's from melting...so do you think Apple and IBM will keep a 25hgz. cool enough????) All I know is it will take a lot of work and more patents before we get past Moore's Law....

-Adam

nerrollus

07-08-2003, 05:57 PM

Originally posted by mahdi
You know they will never let the "almighty MegaHertz" die...Intel will still tout that they are the best and fastest because they have more MegaHertz, and everyone will believe them because the will continue to shove it to people in their mass-marketing campaign.

In 10 years...

AMD "Our processors can run 600 processes per clock cycle"
thepeople "huh?"

Intel "Our processors run at 10 Billion GIGAWATTS!!!"
the people "ooOOooo"

Haha! Too bad that's true..

AlexPlank

07-08-2003, 06:01 PM

I wonder how many bits the processors will be. The G5 is now 64bits as is the Itanium. Will processors ever go up to 256bits or is 64bit good enough? I dont really understand how bits influence the speed. Would 384bit processors just be wasting energy without much benifit?

glussier

07-08-2003, 06:20 PM

Moore's law as nothing to do with megaherts or gigahertz. Moore's law is the doubling of transistors every 2 years. Check http://www.intel.com/research/silicon/mooreslaw.htm .

And, no the 3.2ghz P4 doesn't melt. I just got one and it works pretty good, faster than a 3200+ anyway.

hlrguy

07-08-2003, 06:25 PM

Originally posted by DMR
That's exactly why I said I would be surprised if "chips" are still silicon-based in 10 years. Eventually we'll just run up against the immutable physical limitations of current semiconductor materials.

1.2 cm was in a vaccum, you can't get better than that. I would like to think asynchronous will/would take off in 10 years, but Hz are too entrenched. I think 20 years for truly asynchronous computer's to be around. I still think that the brute force approach of Hz increase won't be cost effective. Multithreaded mulit CPU chips is where I think it will happen. Think like supercomputer clusters. 1 superstupendous machine that costs 48 million or 60 PIIIs with parallel programming for 100K.

As mentioned already...sure there are 3 GHz machines for 1K at best buy, and there are 16 wait states on the 70ns ram, 400 on the HD. Most everyone's CPU is waiting for the data to get there so Hz is not a real good indicator of performance anyway.

hlrguy

GnomeProject

07-08-2003, 06:43 PM

What do you think allows you to increase the processor clock so high??? The number of transistors is obviously directly related...anyway, I thought I read an article about the 3.25's melting, and one that cited a person's house burning down b/c of their machine catching on fire! Either way, I stand corrected.

-Gnome

glussier

07-08-2003, 06:51 PM

Originally posted by GnomeProject
What do you think allows you to increase the processor clock so high??? The number of transistors is obviously directly related...anyway, I thought I read an article about the 3.25's melting, and one that cited a person's house burning down b/c of their machine catching on fire! Either way, I stand corrected.

-Gnome

Processor clock and number of transistors have nothing in common. And, Intel never made a 3.25ghz processor. Mine is a 3.2ghz which happens to run at a frequency of 3204.8mhz. The processor melting and the house on fire are probably urban legends, thought the melting processor could be an AMD XP which has lost it's heat sink.

Stween

07-08-2003, 07:01 PM

Originally posted by RWiggum
I've even read that the next "big thing" in CPU's may be asynchronous chips - no clock at all.

Asynchronous chips have been in the research field for many, many years, and are a very interesting concept. The problem is, they're very difficult to get to work, considering every instruction surely must wait for some other to have completed fully before it can start.

The simple way to achieve this is with a clock controlling the show. It'll be a very long time before asynchronous chips are capable of anything resembling what synchronous consumer chips are capable of, if ever.

bwkaz

07-08-2003, 07:27 PM

Originally posted by glussier
thought the melting processor could be an AMD XP which has lost it's heat sink. A year ago, or two, maybe. Not anymore.

AMD won't put a motherboard on its "recommended motherboards" list unless it supports the thermal diode on the XP's silicon die, and shuts the clock down (AKA, locks up the system) when the temperature being reported from that diode gets too high. That's the same setup that Intel has used for years (and, AFAIK, all Intel-accepting motherboards have to have this diode monitor circuitry on them -- though maybe that's just because I haven't ever looked at non-Intel-chipset motherboards for Intel processors).

AMD's approach doesn't forcibly stop people from making motherboards that'll toast your XP CPU if you pull the heatsink off, but they just won't put that board on the recommended list, which should be just as good -- unfortunately, there are probably enough DIY'ers that don't pay attention to that anyway, that it may not matter.

As for the thread's subject, not only are there limits like the speed of light, there are also limits like transistor switching frequencies. The things can't switch faster than some rate, unless they get smaller -- and once individual transistors are only an atom or two apart, it gets EXTREMELY difficult to separate signals. Personally, I think quantum (or perhaps DNA) computing is the future, but we'll see. Maybe optical will become accepted, too -- after all, it can be configured to emulate quantum, DNA, silicon, and whatever else, or work natively too.

DMR

07-08-2003, 09:27 PM

Originally posted by hlrguy
As mentioned already...sure there are 3 GHz machines for 1K at best buy, and there are 16 wait states on the 70ns ram, 400 on the HD. Most everyone's CPU is waiting for the data to get there so Hz is not a real good indicator of performance anyway.

hlrguy Agreed- I wasn't arguing that in any way.

:)

glussier

07-08-2003, 11:25 PM

shuts the clock down (AKA, locks up the system) when the temperature being reported from that diode gets too high. That's the same setup that Intel has used for years (and, AFAIK, all Intel-accepting motherboards have to have this diode monitor circuitry on them -- though maybe that's just because I haven't ever looked at non-Intel-chipset motherboards for Intel processors).

The P4 as greatly improved cpu protection over the P3 and then the XP. The P4 can send itself into cpu throttling and even shut itself if it is overheating. The P4 doesn't need any help from the motherboard bios or the chipset to protect itself.

Vm.

07-08-2003, 11:36 PM

Heard about Rombus? They already say that they can get you about 6 gigs of BUS speed. That was unimaginable too (for many).

glussier

07-08-2003, 11:43 PM

Originally posted by Vm.
Heard about Rombus? They already say that they can get you about 6 gigs of BUS speed. That was unimaginable too (for many).

You can get 4.3gB/sec with Rambus and 6gB/sec with dual channel ddr with the new Intel 865p and 875p chipsets.

seabass55

07-08-2003, 11:46 PM

Who would have thought back then we'd be where we are now? (http://members.tccoa.com/seabass55/tandy.jpg) <--Click Me

trilarian

07-08-2003, 11:53 PM

I think async will play a part, but something I haven't seen anybody say is sync. parallel processing. As we approach the physical limits of silicon even a new and better material will be hard to push because the industry is so dedicated to silicon right now. So once the limit is close by, run the top of the line in parallel. Really, you can use a lot of cheaper machines. Recently I read an article from the University of Illinois were they linked together playstation 2s in a linux cluster. They were able to build a supercomputer by linking the ps2's together. There have already been machine's like the Cray that used 2D mesh topology for processors. So on a smaller scale, and to keep the Hz tag, use 4 processors at 5 GHz and they have an "effective" frequency of 20GHz. Kind of like the AMD + Hz scheme. As nanotechnology matures, more and more processors could be on one chip. Another possibility is changing mediums. Quantam is a possiblity, but probally not too soon. I think light will come before that. Optical networks are already in use. Give 10 years of research and you may see completely optical chips, or optical chips with and electronic control plane.

trilarian

07-08-2003, 11:54 PM

Originally posted by seabass55
Who would have thought back then we'd be where we are now? (http://members.tccoa.com/seabass55/tandy.jpg) <--Click Me

nice

Artimus

07-08-2003, 11:58 PM

Am I the only one wondering what the purpose of a 200ghz or even 10ghz chip would be?

I mean, I'm sure that they'll be able to come up with videogames with such good graphics that you will need to upgrade to the latest chip.

I can see that you would want to have a computer capable of playing movies without skipping, but what would the average person(Not too much of a gamer) need 10ghz for?

I'll take my 400mhz computer, thank you very much.

seabass55

07-09-2003, 12:06 AM

Artimus: Just think of it the same way we though 10+ years ago. Who knows what will change where that speed of processor will be needed. Maybe we won't be using monitors but holograms. My first computer was a Vic20 and playing Gorf was the greatest thing since sliced bread and I never thought I could be playing games in surround sound (another thing that wasn't even heard of) at over 100fps in full 3d worlds.

Icarus

07-09-2003, 12:16 AM

Originally posted by Artimus
Am I the only one wondering what the purpose of a 200ghz or even 10ghz chip would be?

I mean, I'm sure that they'll be able to come up with videogames with such good graphics that you will need to upgrade to the latest chip.

I can see that you would want to have a computer capable of playing movies without skipping, but what would the average person(Not too much of a gamer) need 10ghz for?

I'll take my 400mhz computer, thank you very much. It's very simple what they can do with processing power this fast...

Artificial Intellegence that can think as fast as a human. That would be the #2 use of these...

#1 would be to run the lastest release of Windows
and
#3 is Gaming :D

Nuada Storm

07-09-2003, 12:21 AM

Originally posted by Artimus
Am I the only one wondering what the purpose of a 200ghz or even 10ghz chip would be?

Visual Computing Interfaces - The next step beyond a graphical computer interface where you work in 3D just like on your real desktop.

Natural Voice Command Recognition - Similar along the lines of what is used in SCi-Fi movies and shows.

HighSpeed Heuristic-based searching that takes into account misspellings and accents.

Centralized computing systems in the home with multiple consoles for the entire family.

Further blending of computing, entertainment and communications.

Full Environment Immersion for gaming and simulations.

Biometric input devices

I am sure there are more ideas in the works. Some of these things are basically available today but the implementations are woefully inadequate. Well except for the BioMetric Weapons systems in the Apache Helicopter.

Stween

07-09-2003, 08:23 AM

Originally posted by Artimus
Am I the only one wondering what the purpose of a 200ghz or even 10ghz chip would be?

People were saying the same when we had mainstream 8MHz chips - "Why would we *need* 66MHz? 100MHz?" ;)

User interface development pushes up the minimm spec for home systems. Advances in compression algorithms might also push processing power forward - fractal compression is a particularly interesting field which requires a fair amount of power behind it.

C'baoth

07-09-2003, 10:02 AM

Quantum computing. The absolute pinnacle of computing using the laws of physics as we currently understand them. I wrote a research paper on it for a physics class that I took. It involves actually encoding data by using the spin of electrons to represent binary digits. So, theoretically, a single electron represents a transistor in the current scheme. Unfortunately, electron movements are rather hard to predict. Actually, accoring to Heisenberg's uncertainty principle, it's impossible. So we've got some physics to conquer before we get there.

One can dream though...