CPU Timeline to 1 Thz

Discussion in 'Processors and motherboards Intel' started by PureStoke, Jul 27, 2002.

  1. PureStoke

    PureStoke Guest

    Alright, Im bored ok... I just read the origonal document which has become known as Moore's Law. It states that the number of transistors in any givin circut should double within the time of a year due to increased lithogrophy tech and other technical advances. However, CPU speeds double every 18 months so I put together a timeline based on a perfect world in which Moores Law is held up exactly on schedule, pending we dont make the switch to clockless computing. So it sould, in theory, go something like this.
    I didnt feel like putting in transistor count, not that bored.

    CPU Timeline (hopefully)
    July.............2002.....................2,500 Mhz........2.5 Ghz
    January......2004.....................5,000 Mhz...........5 Ghz
    July............2005...................10,000 Mhz.........10 Ghz
    January.....2007...................20,000 Mhz.........20 Ghz
    July...........2008...................40,000 Mhz.........40 Ghz
    January....2010...................80,000 Mhz.........80 Ghz
    July..........2011.................160,000 Mhz.......160 Ghz
    January...2013.................320,000 Mhz.......320 Ghz
    July.........2014.................640,000 Mhz.......640 Ghz
    January..2016..............1,280,000 Mhz......1.28 Thz

    (damn non-aligning fonts...)
    5 Ghz at the start of 2004 looks about right to me?
    I know these numbers look way to high at unreasonable dates but according to Intels promise to uphold Moores Law for at least the next 10 years, it should go something like that. Hopefully. I know it wont. If we actually have cpu's clocked at these speeds within such a short time that would be simply amazing. I mean, what kind of game runs on a 1 Thz cpu. It would be the same jump from a game built for a 1 Mhz machine to a game built for a 1 Ghz machine. Thats a hell of a jump. An atari 7600 was about 16 mhz I think... Doom III will require about 1 Ghz for reasobable play. So think of the jump from pong to Doom III, thats the jump we'll have with future 1 Thz games.

    Just to check my accuracy I did the same calculations going backward in 18 month incriments.

    My Timeline Reversed to check for accuracy
    January.....1995......................75 Mhz..........07 Ghz(.07)
    July...........1996....................150 Mhz..........15 Ghz(.15)
    January....1998....................300 Mhz............3 Ghz(.3)
    July..........1999....................625 Mhz............6 Ghz(.6)
    January...2001.................1,250 Mhz.........1.2 Ghz
    July.........2002................ 2,500 Mhz.........2.5 Ghz

    Now I dont really remember what cpu speeds were back in the day but I do remember cpu's being faster than 75mhz in 95. So I went to Intels Website and the closest date proc was the Pentium Processor which was released October 10th 1994 and was clocked at exactly 75mhz; so maybe this list isn't as flawed as it appears. 1999 had processors clocked at about 600 mhz as well. Perhaps Moores Law is actually slowing down.
    Did I do something wrong or is Moores Law just not held as precisely as CPU manufacturors claim. I've read a press release from Intel in which they mention putting 20 Ghz Chips out in 2010. Now, my calculation based on Intels own Moores Law at 80 GHz is a great deal faster. Even if Moors Law accelerated at 1/2 speed we should be faster than 20 GHz in 8 years.

    Did I do something wrong? Moores paper on the subject is very clear about this and its hard to misinterpret. Do any of you all have any more accurate timelines? Where did I slip up in my math here? I mean, im prety good at multiplication by 2's.
    Last edited by a moderator: Jul 27, 2002
  2. Actually speed beyond 20GHz my not be possible with silicon because it would require transistor size below 10nm and electron waves are about that size in silicon! So anything smaller then 10nm in silicon can't move electrons because its smaller then then a moving electron!!!

    Moleclertronics that use a single molecule as a transistor mite be abile to perform higher because they could support sizes of about 1nm also with neural network architecture that could support thousands if not millions (like are brain) of simulations operation per cycle would greatly improve performance so cycle speed will become less important in the next 20 years

    As for 1Thz operation I think we can only expect that from Quantum computing using Ion in a magnetic trap or in some way a matrix of atom to make calculations at using qubits or 1,0,Undifind at below 1nm die space but Quatum computers will work by using a uniform group of atoms so die space could be as large as a cm its just each atom is do a calculation!

    Also I should mention that we have liquid helium cooled Super computers running at 450GHz at this time! But Room temperature super conductors mite not be possible so don’t expect having one of these fit on top of your desk!

    As for games: look at Cgi movies now and think by 2010 that’s the way all computer games will look! Also so day there will be direct connection using cybernetics to directly connect your brain to a computer then all game will be like the matrix only better. Also you will have the freedom to pull the cored out anytime you want and go for a walk in the then super polluted, globally warmed, overcrowded earth were people can only see real beauty in a computer simulation!
    Last edited by a moderator: Jul 27, 2002
  3. PureStoke

    PureStoke Guest

    Ya i know you're exactly right about the atomic limit of current silicon chipsets. Quantom Computing will deiliver unpresidented gains in performance but conventional silicon chip design is still in its infancy. I believe it was IBM which demonstrated a silicon circut with transistors built on .02 micron lithogrophy operating at 210 Ghz. I might be wrong but i remember hearing something about the creation of a switch operating at over 1 thz. Many experts also have mentioned creating 3 dimensional chips which could potentialy pose a tremendous performance gain if heat issues can be tackled. Paralelle processing will also become more practical as many of these 20-50ghz silicon chips can be linked via an optical circut. Our brain is in fact composed of millions of paralelle computers. A nueron itself can do little computing, but when paired with billions, you get the fastest computing device known to man. As far as I know, quantom computing is defined as any operations performed at a subatomic level, so one form I know is called optical circutry. Lasers have many pro's such as being capable of forming switches which can intersect without interfering with each other. There is also no current limit to how small laser energy can be transmitted. I suppose all you need is one quanta packet to send a signal and quanta are fractions of the size of an atom. Information can be sent on different wavelengths of the laser allowing massive performance gains over standard 1,0 operations.

    I imagine silicon will take us up to at least 50 ghz before retiring. This limit will focus attention on Quantom computing which will probably be obtainable with the technology of the times. However, the switch to quantum computing will render Moores Law obsolete thus making my timeline faulty. We could very well jump from a 50 ghz silicon chip to a 1 Thz quantom chip in the course of about 10 years. Either way, computing speeds are sure to skyrocket in the next 50 years. To the point where I dont imagine it will matter anymore.

    As far as direct brain link goes I couldnt be more anxious for that. I've read a few articles on the emerging possibilites. Computer enhanced dreaming is only one piece of the pie though. Network based computer dreaming will take dreaming to a new level. Imagine sharing a dream with other people who are sleeping as well. You very well could go on a "Final Fantasy" type adventure with your friends in your sleep. (with the gfx being far superior to the films)

    Super real computer simulation will pose another social problem. If you can live in a dream world where you are the King of some virtual country and it looks and feels as real as this world, many people simply won't want to ever leave the simulation.
  4. Quantum computing is limited to fields of pure number crunching for more varying activities I’m putting my bet on molecular electronics constructed in a neural network like fashion to improve manufacture and cost. A Molecular electronic chip can be constructed of individual neurons that are manufacture in a liquid and then place together in a crystal made of neurons forming a neural network of limitless size. So in the future the power of your process will be measured by the quality of each molecular electronic neuron and how many neurons you have in the CPU!

    As for speed of the Artificial neuron, are Biological neuron run at about 30hz!, but because the network is so huge we can out computer 10 super computers in learning operations. Computer already have a pure speed advantage over us and logic that would take a human years to compute and they can do it in a millisecond. With Artificial neural networks computer will final be better then us at everything!

  5. Also we could do over 20Ghz with silicon no problem but it requires super conductivity and very cold temps. which I don't think we could ever make fit on a desktop!
  6. PureStoke

    PureStoke Guest

    ya but computers which were no more powerful than a gameboy used to be the size of buildings. Of course, as long as there is money to be made, a breakthrough will be will be made as well. Intel has got billions invested in making sure they wont be out of jobs in 10 years time. Even if we just cant make desktop computing a reality in the future, there will be supercomputing buildings which computers will connect to via super broadband networks. Computing would end up being distributed like electricty or water. Heh, can you imagine paying a data bill. I think 20ghz will be no problem, even without superconductors. I've also heard something about carbon nanotubes which behave like superconductors at room temperature, but I don't really know enough about it.
  7. intersting idea sounds a little old though... Bakc in the 50-60's they though that was the way all computing was going to be done and totally missed the personal computer revolution!
  8. PureStoke

    PureStoke Guest

    go to btexact.com. Or more precisely,


    open the .pdf file on the bottom of the page.
    This is a timeline they put out and its a cool read. Don't get to excited though, remember they said we'd have hotels on the moon by the 80's.
  9. Some of the thing I see i agree with some I think there @#$%#$%^ use...

    Orgasm by E-mail by 2010 SWEET!!
  10. WeaZel

    WeaZel Guest

    what you're saying is true about the "speed" of a processor but not necisarilly<--"totally spelt wrong" the performance of one. there's the ****load of optimizations SSE SSE2 3dnow! yadda yadda, cache size, data prefetching, data predictions, fsb speeds, a whole **** load of things.

    so in honesty you could say that your 5GHz cpu could actually perform not 5 times faster than a 1ghz machine but maybe 6 or 7 beause of it's other enhancements

    in other words there's more to cpu's than shear clock speed this isn't an AMD speech either, this goes both ways.

  11. PureStoke

    PureStoke Guest

    Ya thats true, I ran some tests just to prove that. SiSoftware Sandra benchmarked my new 2.53 Ghz cpu at 3148 MFlops and an older PIII 500 at 670 MFlops in the CPU arithmetic Test.
    2533/500 = 5.066 (5 times mhz)
    3148/670= 4.7 (only 4 times performance)
    This result doesnt concur with what I expected, but thats probably becuase the pIII gets more bang per mhz than the p4.

    I ran the multimedia Benchmark to see if the same thing happend.
    The 2.53 iSSE2 got 12243 floating point it/s. The 500 iSSE got 3320 fp it/s. So...
    2533/500 = 5.066 (5 times mhz)
    12243/3320 = 3.7 (not even 4 times the performance gain)

    Maybe the P4-P3 comparison just isn't fair. Its seems like the p4 does a lot less work per mhz than the p3, allowing it to climb to obsurd clock frequencies. Oh well, maybe the results got thrown off some by me having this Iexplorer window open or something.
    And I imagine the score listed for the 500 was a best possible. Im gonna run this benchmark on my PIII downstairs and see if it does as good as the listed P3 500 in Sandra.
  12. WeaZel

    WeaZel Guest

    Actually the P3 does have a better IPC(instructions per clock cycle) than a P4...look at the tualatin's thye are quite the performer guarantee'd if they got a 133 fsb or at least some kind of support from Intel that they'd gt more adopted. but instead intel asks way to much coin for them and crippled them one way or another. the p4 was/is a rushed cpu, intel knows that ever idiot in the world sees mhz as the performance standard so they did a few "tweaks" for the p4 which actualy did more harm than good.
  13. Longer pipline is killing the P4 in performance but it makes it easy to overclock it!
    Speed is the game and Intel knows were all idiots:
    the fast it goes the better right! Wrong but most people just don't know that!
    Its like as race between a sports car and a truck, ya the car will win but if you add a trailer (workload)... well you get the picture.
  14. GeekBoyJOEY

    GeekBoyJOEY Guest

    Consider the XP core as a twice bigger wheelbarrow moving 40% slower than Intel's smaller wheelbarrow. Efficeint is the key.

    Consider the OPS penalties of the 20 stage pipeline that the P4 has to deal with. It takes longer to complete the cycle even though the frequency is higher, regardless of the size of the data. It's being pushed through the pipeline at a fast rate but it has a long way to go before executing the next operation.

    MHz = higher
    OPS = lower

    I would say that's a step backwards.

    Thunderbird or Mad Dog is fortified wine with a high alcohol content. High in alcohol (MHz) but very low in quality of buzz (OPS). I don't think you're going to be too smooth with the ladies after drinking that crap.....Hehehe..... ;)
  15. GeekBoyJOEY

    GeekBoyJOEY Guest

    It kind of seems like Intel did this to buy time while keeping the high margin revenue coming in with MHz labling?

    You got to give Intel credit, they're marketing geniuses. They're selling a performance inferior CPU while still claiming to be the performance leader.

    Non optimised SSE/SSE2 benchmarks are a disaster for the P4. Wasn't 3DNow! made for the same reason on the original K6? The SIMD intructions are the only thing holding up the P4's FPU now when compared to the XP's FPU. It's crazy to even think about Intel being beat in the FPU area. You would of never thought that day would come but it did....Remember that the AthlonXP has SSE but not SSE2 as a implemented SIMD. The P4 doesn't put 3DNow!+ on their chips? Arrogance?

    Hoooray for TOUGH competition! Intel and AMD people rejoice at the low cost, high end computing we now enjoy at home. :D

  16. ---TK---

    ---TK--- Ancient Guru

    Likes Received:
    2x 980Ti Gaming 1430/7296
    nice 8 year old dead thread. the review for the commodore 64 still here
  17. XeIDy

    XeIDy New Member

    Likes Received:
    HIS ATi Radeon HD 5770

    Wow, how untrue that is for the year 2010 and all the way to all the years after 2002. We're still stuck at 4 Ghz max 6-7 Ghz if crazily overclocked. Only thing that has been heightening since then is the RAM that's all..
  18. killerxr290

    killerxr290 Banned

    Likes Received:
    Im Out
    Mores law is still holding in the fact that CPU Computing power is almost doubling every 18 months.

    Witch it is.

    For example.. lets compare my 2008 cpu to a new 2010 cpu.

    A Core i7 980X can do alsmot twice as much work as my current CPU.

    But as for speed, it may be 2.50Ghz (well oc to 3.75GHz) but thats 4 cores, total of 10Ghz (15Ghz OC) of computing power.. the I7 980X on the other hand...

    6 cores of 3.2Ghz each with a total of 19.2Ghz of computing power! almost double my CPU ;)

    And if we OC the 980X to 4ghz, thats 24Ghz of power in total and dont forget hyper threading also adds a little bit of a boost as well.

    I think we are doing fine keeping up with mores law :)
  19. F1refly

    F1refly Ancient Guru

    Likes Received:
    970GTX-oc edition
    bet those ancient posters never thought that thread would ever revive. wow 8 years. i prolly had XP, a Pentium 3 and still rockin my voodoo card back then, long before ever knowing of this site.

    but to topic i think we'll just keep seeing more cores and ghz will eventually be a marketing tool of the past. it pretty much already is. Cores are the bit marketing bling now, couple that with a blu ray player and a shiny case with a bright monitor showing a video of some family editing photos and you have some happy Best Buy employees and thankful Dell/HP CEO's.
    but if anything, we seem to be heading towards power efficiency, smaller products to do all of our basic tasks, Fusion on the horizon..etc and portable power may take over predominance among enthusiasts. the age of the power hungry as we know it may dwindle a bit over the comming years.

    so i'll wait for 8 more yrs to go by and someone revive this thread so i can see how close my prophecy is..muahahh
  20. scoter man1

    scoter man1 Ancient Guru

    Likes Received:
    MSI GTX 1070ti
    Haha, Its cool to look back on this though. It is pretty much true though. We have been double processing power... just not single core speed. Well, we arent too far off. If you take the i7 980x, and count all 12 logical processors as a core running at 3.6ghz, thats 43.2ghz of power. With a 5ghz overclock on that cpu its kinda like 60ghz. So, I guess we left behind one of the doublings.

    @f1refly, I think you have the right idea there. Seems like the intel atom D510 and things like that are being praised right now.
    Last edited: Jun 3, 2010

Share This Page