Has fast will we get to with cpu's?

[Terminal23]

How fast can cpu's possibly go to?The fastest right now is 1.7 Ghz.They seem to increase at a fast frequency lately.So if my calculations are correct we should see a cpu doing 80 Ghz in ten years.Is this even possible or am I just dreaming?Any thoughts?

[Hellmund]

Cpu's are already past 1.7ghz. That's the fastest commercially avaiable stock standard CPU. There's no set limit to how fast a CPU will go, once solid state is as small as it can get they'll use a different kind of cpu.

[RobRich]

Several technolgies are currently emerging which should allow for near exponential speed increases over the next several years. A few examples include:

-Semiconductor substrate manufacturing using an advanced gaseous nitrogen interface material instead of silicon. The nitrogen gas actually forms a nearly pure substrate solid when raised to extremely high pressures. The nice part? The nitrogen-based material stays solid after the compression process, even at standard atmospheric pressures.

-Silicon on Insulator (SOI) is likely to be the next masive fabrication change within the retail market sector. Here is my recent article on SOI:

http://sysopt.earthweb.com/articles/soi/index.html

-Silicon Germanium is another hybrid manufacturing process which could allow even current production processes technology to scale to well below .5 micron for increased frequency ramping well into the multiple GHz range.

These are just a few emerging technologies with promising potential. For further research, check out some of the latest papers published at IBM or Intel's developer sites.

Catch ya' later,
Robert Richmond

[Recordlord]

Last year sometime a doctor used biological tissue to complete simple arithmetic. Although the math problems took 15 minutes to complete it is still new technology that havent beeeen tapped yet. It makes me wonder how many bits of data can a single strand of DNA possibly hold??

[Hellmund]

That reminds me of a popular mechanics article I read agggeeeeees ago... They theorised that a 100Litre cylinder of DNA could hold all the information on every computer in the world, I'll have a look for it as I'm probaly way off by now, that was atleast a year ago.

[nilknarf]

That's not to far fetched either. If you consider that current digital systems are binary in nature, as far as the actual data storage. Which means that every bit of data has two possible states, either a zero or one.

DNA, on the other hand, is made up of four ammino acids, and therefore is considered quarternary. That is, it technically has four possible states for each bit of data.

The power of organic processing is also extrodinary. Consider for a moment, as I am writing this, my brain is sending commands to both hands, my fingers, arms, and eyes, just to type this information! That doesn't include the involuntary commands being sent to various organs, and the occasional command to change my seating position. On top of all of this, my brain is also processing the information I'm typing, the language (vocabulary, spelling, grammar) information for this information, a multitude of sensory inputs, unknown numbers of sub-conscious thoughts, interpreting those sensory inputs to determine what's going on around me, and other conscious tasks on demand. All this, while keeping focus on my primary tasks at hand. No computer yet built, can match the size to peformance ability of a human brain. The human brain, or any brain, is truly a marvelous piece of organic engineering.

Tim

[LiLRiceBoi]

Lets just see how long moore's law will take us

[Brangwen]

nilknarf: Well thought out and ponderable stuff.

RobRich: Thanks for the link to a very interesting development.

Brangwen http://www.sysopt.com/forum/wink.gif

[HomeYield]

A graduate student that I know has done research on processors that get their information from optical light. He is doing research to try and speed up processors and even harddrives with light. By sending data at different wave lengths of light the processor can determine from that more than it can from 1's and 0's in an even smaller amount of time. If optics could be utilitzed then processors could go as fast or faster than the speed of light (http://www.aip.org/enews/physnews/1999/split/pnu434-1.htm) .

It would be interesting to see a Pentium 5000 1000THz processor.

[MTAtech]

I remember having this discussion over dinner with a well-respected software developer about 10 years ago. He was convinced that we were then reaching the limit of speed [50Mhz] because of the natural limitations such as the speed of light. Boy was he wrong.

I am sure there is some limit but we aren't anywhere close to it.

The real question is what will we do with that speed? My PC is more powerful than my univerity's entire computer system when I went to college. However, on that weak system professors where unlocking the secrets of the universe. Although my computing power is greater, my uses are much more modest.

[DemonKnight]

Who needs more Hz? I say go for higher bit rates. Who would like a 1Ghz 1024Bit processor? hell even a 1Ghz 128bit processor would be awsome. Considering it has 4 times the bandwith that a normal 32 bit processor has.

[nilknarf]

32-bit processors use 32-bit instructions. A 128-bit processor would use 128-bit instructions. This doesn't necessarily mean it would have a higher bandwidth, although you would certainly hope so, it just means it can process a larger instruction.

[Bob The Great]

Could you amagine feeding your DNA powered PC! har har!

No really cool stuff is happining! I can't wait to see what the future will bring! PC's always amaze me. Or actaully the people who make the PC's!

[charmler]

I read about a crystal based storage system somewhere that used lasers of varying frequency and speed to store data. I also remember someone slowing a laser down to about 38 mph recently. The implications were for computing speeds of wavelength proportions and zero heat produced as photons contain no heat.

[samwichse]

"The real question is what will we do with that speed?"

Well, I read an article (in popular science, mechanics, or something like those) that this guy said that at about 20ghz, real time voice instruction could be possible. I'm not talking about just dictation, even though that could be more accurate, but something more on the lines of your computer would be able to distinguish between your speech and background noise, and further, when you were talking to it and not other conversation, and what sort of request or command you were making, or if you wanted to enter some kind of data by things like intonation.

The winner of this big robot competition said that their robot had the intelligence equivilent of a fly, and that a dog or cat was years away.

EDIT: just imagine the star trek computer.

[This message has been edited by samwichse (edited 06-10-2001).]

[SysOpt]

20GHz by 2007 (http://www.sysopt.com/news.html#992204340)

[zskillz]

hmmmm sysopt, curious about that .02 micron setup

as I understand it, the concern isn't really so much how small we can get the transistors, becuase the problem is that the electrons don't obey classical particle mechanics once confined to a small enough box. (the entire theory behing a scanning tunneling microscope STM)

so my question is - what size box is the absolute minimum that we can still treat the electron as a particle?


rob, i read ur article, and was impressed... however, I do have a few questions - it appears to me as if the SOI technology will not really decrease the current size of the technology, but just effectively allow a larger number of Hz to be used (due to lowerd capacitance discharge time) - is this understanding correct?


finally, a question about the direction of processing in general?

I'm curious about the several directions that computer processing could go in the future - (these are jsut a few of the possibilities)

1. quantum computers - no longer process orders consecutively, but all of the orders at the same time. the potential for this area seems to me to be unfathomable (sp?)... it seems as if we stick with the idea of binary, electron spin is set perfectly for that type of processing

2. neural networks - now I won't even pretend to understand how these work or even what the really are to be honest with you, but if my idea is correct, with the possiblities involved with organic/inorganic molecule processors, the ability to create a "brain" of sorts could become a reality where individual transistors form bonds to more than one other transistor

3. and finally the difference between higher frequencies and just a larger bit rate... I believe that we are already seeing this issue come to light (in it's early stages) when comparing the new p4's to the thunderbirds...


wow... I wish I was a genius, that way I would at least know what my own knowldedge limitations were! I want to work in this field, but it's tough to even tell where to begin!

c'est la vie
-Z

[hallam2003]

in light of zskillz's last post, not the asme thing, but it made me thing... this doesn't mean anything about the smallness about the enclosure around the electron, but about the composure of the electron it's self. I haven't had any education here, so correct me where I'm no where close. an electron can be enclosed as small as you like it, at least untill the edges of the elctron are to close to the edges of the enclosure, this is not considering that electrons act differently in this environment. well, if you break down an electron, you are left with waves, right? radiation? wow, nuclear computing baby!!! also, what is stopping developers from using higher level number systems. if a computer was based on the decimal system, wouldn't it run faster. as i see it, it's just that we have to make those computers smart enought to know thing more than 1 and 0...hhhmmm, wow, i just hyperwarped into "AI" anyone going to see that movie BTW?


Hallam

[NDC]

Well, it looks like CPU's wil be doing 20Ghz in 6 years fro now if this prediction is true...


http://www.sysopt.com/news/gws/cdata/Forum1/HTML/000080.html

[RobRich]

Sorry for the late reply, but yes to the question posed to me above by zskills. SOI is a process to increase operational efficiency and ramp MHz ranges without the need for smaller die size manufacturing. However, SOI could prove to be a mechanism to allow smaller die sizes due to increased gate switching effectiveness with lower amounts of power needed for sustaining electrical capacitance of the switch itself.

BTW, Intel just debuted a transistor 80 atoms wide by 3 atoms thick that operates in the THz (Terrehertz!) ranges. The link posted by NDC outlines some of the broad aspects of the technology for those interested.

Robert Richmond

[zskillz]

hallam.. there is a whole lot of physics involved at the molecular level... the point is that at a certain point, an electron acts as a wave and a particle (much like light).... it is at this point that the electron stops obeying the laws of classical mechanics and starts to follow the laws of quantum mechanics... the problem is that the current computer technology only works when the electron follows the rules of classical mechanics

[zskillz]

^bump^