Home

News

Forums

Hardware

CPUs

Mainboards

Video

Guides

CPU Prices

Memory Prices

Shop



Sharky Extreme :


Latest News


- Acer Fires Up Two New Ferrari Notebooks
- Belkin Debuts Docking Station for ExpressCard-Equipped Notebooks
- Logitech 5.1 Speaker System Puts Your Ears At Eye Level
- Dell, Nvidia, and Intel Fire Up Overclockable Gaming Notebook
- Gateway Puts Premium Components Into Affordable Home Desktop
News Archives

Features

- SharkyExtreme.com: Interview with ATI's Terry Makedon
- SharkyExtreme.com: Interview with Seagate's Joni Clark
- Half-Life 2 Review
- DOOM 3 Review
- Unreal Tournament 2004 Review

Buyer's Guides

- September High-end Gaming PC Buyer's Guide
- September Value Gaming PC Buyer's Guide
- October Extreme Gaming PC Buyer's Guide

HARDWARE

  • CPUs


  • Motherboards

    - DFI LANPARTY UT nF4 Ultra-D Motherboard Review

  • Video Cards

    - Gigabyte GeForce 7600 GT 256MB Review
    - ASUS EN7900GT TOP 256MB Review
    - ASUS EN7600GT Silent 256MB Review
    - Biostar GeForce 7900 GT 256MB Review





  • SharkyForums.Com - Print: Moore's Law

    Moore's Law
    By Moridin December 08, 2000, 01:36 PM

    Moore's law says that processor complexity will double every 2 years (due to process shrinks/improvements). The effect has been that performance has also doubled every 2 years. This has held up very well since the early 70's, but it cannot go on forever.

    At some point it will be physically impossible to make a transistor any smaller and still have it work. Even before this happens we are going to run into other roadblocks like the resistance of the wires increasing as their size shrinks.

    What I want to know is how much longer you think Moore's law will continue to hold and what you think will happen when it starts to break down?

    By Arcadian December 08, 2000, 01:45 PM

    Well, Moore's law states that performance will double every 18 months. I think this will continue for the next few years. After that, microprocessors will start to integrate more parallel functions on the die. I also think that Intel's IA-64 architecture will allow for more headroom 3 or 4 years down the road when it becomes mainstream.

    After that, I'll agree that you can't push CMOS any farther. However, there is more research using bipolar junction transisters (BJTs), which can scale smaller than CMOS, and provide faster switching speeds at a given size. The biggest problem with BJTs is that they yield very low, but I think this will change by the time the industry is desperate for it. I'm thinking this will be 5-10 years down the road.

    After that? Possibly quantum computers that use electron oribtals to see what state each atom is in. There are already several theories that can be applied to quantum computing. However, it requires that the programmer think a lot differently, and allow for different ways to parellelize the data. But I think that practically, this is 20-30 years off.

    By nukefault December 08, 2000, 01:53 PM

    Thats easy enough, at least in theory. CPUs won't always be little chips full of wires, and will therefore bypass the size problems entirely. There is already research being done on holographic, biological (DNA-type), and a bazillion other ideas. Most of them sound like bad '70s sci-fi right now, but eventually one of them will yield results.

    The two examples I gave are actually new storage ideas to replace the usual hard drive platters, but you get the idea. Heck, even MacOS Rumors (http://mosr.com) posted an article a couple years back about quantum computing and another about plasma-based CPUs. They both sound like sci-fi to me right now, but I'm sure someone (can we say "IBM"?) said the same thing about the idea of a fully functional home computer a couple decades back

    By Moridin December 08, 2000, 01:58 PM

    quote:Originally posted by Arcadian:

    After that, I'll agree that you can't push CMOS any farther. However, there is more research using bipolar junction transisters (BJTs), which can scale smaller than CMOS, and provide faster switching speeds at a given size. The biggest problem with BJTs is that they yield very low, but I think this will change by the time the industry is desperate for it. I'm thinking this will be 5-10 years down the road.


    I wasn't aware that BJT's scaled better then CMOS. To me their function always seemed more random.
    Still I would think that you would have a lot of problems with them. They tend to be much more sensitive to heat and have much higher leakage currents and therefore would require a lot more power.

    By Moridin December 08, 2000, 02:10 PM

    quote:Originally posted by nukefault:
    Thats easy enough, at least in theory. CPUs won't always be little chips full of wires, and will therefore bypass the size problems entirely. There is already research being done on holographic, biological (DNA-type), and a bazillion other ideas. Most of them sound like bad '70s sci-fi right now, but eventually one of them will yield results.

    The two examples I gave are actually new storage ideas to replace the usual hard drive platters, but you get the idea. Heck, even MacOS Rumors (http://mosr.com) posted an article a couple years back about quantum computing and another about plasma-based CPUs. They both sound like sci-fi to me right now, but I'm sure someone (can we say "IBM"?) said the same thing about the idea of a fully functional home computer a couple decades back


    Yes there are a lot of ideas out there, but are any of them really practical? Current techniques have been so successful because the production problems are relatively straightforward. Do the same thing you did before using a higher frequency laser and you have a smaller, faster transistor. (very roughly to say the least) The problem is that at some point that smaller transistor will no longer function.

    The basic problem as I see it isn't coming up with new ways to perform logic operations it is figuring out how to mass-produce the things.

    (BTW IBM didn't start believing in the PC until well after they lost control of it)

    By Arcadian December 08, 2000, 02:18 PM

    quote:Originally posted by Moridin:

    I wasn't aware that BJT's scaled better then CMOS. To me their function always seemed more random.
    Still I would think that you would have a lot of problems with them. They tend to be much more sensitive to heat and have much higher leakage currents and therefore would require a lot more power.

    You're wrong on two counts. First, BJTs are very sturdy against heat. Much more so than CMOS. They also happen to have far less in the way of leakage currents. You are correct that they require more power, though, but that is simply because they have a larger surface that needs to switch.

    Modern BJT circuits are very different than those of old. It's hard to explain, but the modern BJT integrated transistor looks like a wedding cake. It is created using layers. The old BJT looked more like a quicksand pit, with different sunken layers for the emitter, base, and tranceiver.

    I had a college professor when I was in college (not that long ago) that used BJT technology on Gallium Arsenide (instead of silicon), and was able to reach 80.0GHz on .5u. It was a network chip with limited functionality, and 4/5 dies per wafer were bad, but you get the idea. It's a very powerful technology. It just needs to be applied towards microprocessors (which have 1000's of times more transistors than my professor's proof of concept machine, and need much higher yields). Needless to say, some BJT processes can reach .25u. Just think of the speeds you can get on that.

    By Sketch December 08, 2000, 02:22 PM

    I was told in college electronics course that BJT could go faster because the trasistor would actually switch faster. We didn't do much with BJT's though. It was a CMOS-centric course.

    Edit: Also, I've read where we still have a few more iterations to go with CMOS technology. I've read that we'll hit the theoretical limit around the year 2020, but again, I seen no substatiating proof of that.

    By Moridin December 08, 2000, 03:07 PM

    quote:Originally posted by Arcadian:
    You're wrong on two counts. First, BJTs are very sturdy against heat. Much more so than CMOS. They also happen to have far less in the way of leakage currents. You are correct that they require more power, though, but that is simply because they have a larger surface that needs to switch.

    Modern BJT circuits are very different than those of old. It's hard to explain, but the modern BJT integrated transistor looks like a wedding cake. It is created using layers. The old BJT looked more like a quicksand pit, with different sunken layers for the emitter, base, and tranceiver.

    I had a college professor when I was in college (not that long ago) that used BJT technology on Gallium Arsenide (instead of silicon), and was able to reach 80.0GHz on .5u. It was a network chip with limited functionality, and 4/5 dies per wafer were bad, but you get the idea. It's a very powerful technology. It just needs to be applied towards microprocessors (which have 1000's of times more transistors than my professor's proof of concept machine, and need much higher yields). Needless to say, some BJT processes can reach .25u. Just think of the speeds you can get on that.

    Are you referring to the actual construction of the BJT or the energy level representation?

    The analysis of the BJT that I am familiar with has the energy levels for the emitter and collector higher then the level for the base. The electrons can freely travel from the collector to the base since it is traveling to a lower energy level. To go from the base to the emitter the electron must gain energy from some external source (heat, photoelectric) and be carried onward due to the electric field. Therefor the higher the temperature the greater the chance that an electron can be promoted and contribute to current (I).

    The current can be controlled by applying a voltage to the base-emitter junction which raises or lowers the energy level for the base, thereby increasing or decreasing the chance of electron to moving from the base to emitter. All this is for a PNP transistor it is reversed for NPN, i.e. your analysis is done on holes, and the base energy level is higher, etc.

    If you are saying that there is a BJT with a newer model I would like to see some details if you can point me to any. If it is functionally very different is it really a BJT or is it really a new type of device altogether?

    I have worked with Gallium Arsenide field effect transistors in (GAs FET's) in communication circuits (up to 10-20 GHz), so I know Gallium Arsenide allows you to reach much higher frequencies.

    Traditional BJT's have very low input impedance compared to any FET let alone something like a CMOS FET. How is this problem avoided?


    By ilsie December 08, 2000, 04:09 PM

    I'd be curious to see this wedding cake design also. I would think this sort of tiered design (am I understanding this right?) would be extremely hard to implement using a typical PR/etch process. Plus, it seems that it would be a complete nightmare adding a metal layer to that.

    Again, I may be visualizing this wrong. I would like to see a cross section of this if you've got a link handy.

    By Marsolin December 08, 2000, 05:09 PM

    I've seen projections that show Moore's Law to continue for at least another 10 years. Who knows if that will be correct, but I personally think we'll see Moore's Law continue for another 5 years at minimum. It wasn't too long ago that people were saying we wouldn't go below .1u. Now that's only two generations away and there are reports of much smaller than that being possible.

    It seems as if we'll see changes in the way integrated circuits are built, but the pace remaining fairly constant.

    Arcadian: What do you do for a living? I see your posts everywhere around here with details on many different subjects.

    By 100%TotallyNude December 08, 2000, 06:30 PM

    quote:Originally posted by nukefault:
    Thats easy enough, at least in theory. CPUs won't always be little chips full of wires...

    Eh? They never were full of wires... a chip full of wires has never been constructed. At least by anyone sane... Silica subtratem are not "little wires". A best comparison would be little semi-conducting channels I guess. Infact, I see a comparison of a cpu to a tiny model of Venice with its channels full of electrons rather than water a little more the ticket. But rather than charming Duomo-type buildings its going to look more like the surface of the Death Star. But the parallels to wires... I don't see it.

    By Arcadian December 08, 2000, 11:05 PM

    Moridin:

    You are correct on all accounts. That is the way that a Bipolar Junction Transistor works. The new way I was refering to does not change this behavior. It simply allows for better ways of making large scale integrated circuits out of BJTs. The wedding cake analogy is what my professor in college used, and by his diagram, it did sort of look like a wedding cake, with the Collector, Base, and Emitter on different layers. It actually makes the design much more stable, especially at smaller processes. Hope this clears some stuff up.

    By zarei December 09, 2000, 06:53 AM

    Moore's law says nothing about the speed of the processors on the market. If one processor in the world runs twice as fast as the last one the law has been fulfilled...neh?

    So it may be that the speed increase in PC processor may well slow down.

    By 100%TotallyNude December 09, 2000, 10:46 AM

    quote:Originally posted by zarei:
    ...neh?

    Wakari masu nihongo desu ka?

    I probably didn't say that quite right...


    Contact Us | www.SharkyForums.com

    Copyright © 1999, 2000 internet.com Corporation. All Rights Reserved.


    Ultimate Bulletin Board 5.46

    previous page
    next page




    HardwareCentral
    Compare products, prices, and stores at Hardware Central!


    Copyright © 2002 INT Media Group, Incorporated. All Rights Reserved. About INT Media Group | Press Releases | Privacy Policy | Career Opportunities