Many say that the limits of electronic circutry is being reached in todays microchips and scientists are looking at alternative methods of micocircutry, even at the molecular/atomic level. Whatever way is found, Ive wondered at how future computers will spec. The current microchip is only about 3 or 4 layers deep, but what if future micochips become a thousand layers deep. Things like cache and RAM will be intergrated as part of the processor itself, even much of current HD memory will be stored in the chip. Things like buss speed will no longer be a factor as the processing width of the processor itself will be killobytes wide rather than the current 32 bits. What do you speculate computers will be capable of if you can break wide open the avenues of bandwidth travel?
By: ageorge - 12th May 2003 at 20:16
Re: that’s very true…
Originally posted by Vortex
there’s a lot of ‘technicians’ in fab houses, but of course many of the IC technicians are also at least college graduates or even post graduates. Anotherthing is that software and hardware is very different, you can design some cmos circuits to do some ASIC work, but that doesn’t at all mean you know how to program your remote control written by some wierd cs geek…
That is very true Vortex , towards the end of the Livingston Fabrication Plant 1’s life NEC started fabricating third party design u-com and EPROM devices , I don’t think ANYBODY knew what these were for , we just made them and shipped them – what the customer programs into EPROM’s is up to them , and away over my head !! , the small u-com IC’s were just for pocket calculators , car air-bag control circuits , PLC systems etc
By: Arabella-Cox - 12th May 2003 at 20:09
that’s very true…
there’s a lot of ‘technicians’ in fab houses, but of course many of the IC technicians are also at least college graduates or even post graduates. Anotherthing is that software and hardware is very different, you can design some cmos circuits to do some ASIC work, but that doesn’t at all mean you know how to program your remote control written by some wierd cs geek…
By: ageorge - 12th May 2003 at 18:59
Originally posted by mixtec
ageorge- I recall explaining how to upload pics on this forum to you because you said you were hardly able to operate a tv remote. And now you say you design computer chips for one of Japans largest electronics companys? So I quess you could design a tv remote and you wouldnt be able to work it. Im curious, did you work in Japan with Japaneese engineers or does NEC have some subsidiary in Scotland?
Hi Mixtec , NEC had 2 wafer fabrication plants in Scotland and an assembly plant , the first fab only made up to 4meg chips then was mothballed when Fab 2 opened , this made the 64 meg chips , I worked there from June 1991 till Jan 2000 , at the start there was a lot of Japanese engineers but as people were trained the Japanese were allowed to rotate back home . I would like to point out that I did not say that I designed IC’s – I worked in a lot of the manufacturing processes though – especially towards the assembly end , wafer dice , wafer mount , gold wire bonding etc . I also spent some time in the PYW area , it was a good job but as it was becoming easier to manufacture reliable IC’s the labour costs in Scotland made the plant unviable and I left before it closed , it’s still mothballed and could , technically , be restarted but the cheap labour costs in the Far East pretty much rule this out , the NEC Wafer Fab in Shanghai is a carbon copy of the Livingston plant and it is still running.
By: mixtec - 10th May 2003 at 00:53
ageorge- I recall explaining how to upload pics on this forum to you because you said you were hardly able to operate a tv remote. And now you say you design computer chips for one of Japans largest electronics companys? So I quess you could design a tv remote and you wouldnt be able to work it. Im curious, did you work in Japan with Japaneese engineers or does NEC have some subsidiary in Scotland?
By: Arabella-Cox - 9th May 2003 at 20:30
heat sinks.
actually, you’ll find out that the thermal conductivity between aluminum and copper isn’t that big of a deal for consumer computers…that’s because our cpus don’t really get that hot relative to high performance ones. Keep in mind that temperature is what kills(which also include side effects like what george is talking about, thermal expansion mismatch). As to dipping in liquid nitrogen….that’s when heat flux management is really cruicial because your silicon based CPUs will not operate at that temperature….what’s needed is to maintain the chip temperature at optimum temperature (forgot what that is for silicon…not really too low actually). Watch out for condensation also, will kill your computer instantly.
By: ageorge - 9th May 2003 at 19:48
The physical amount of layers is not a problem , just as I left NEC we were producing 64 meg chips with more than 120 layers all using the relatively old technology of I/I , sputter , dry etch ,wet etch and epitaxial growth ( spun layers) . The only real problem we had was that when the chips were resin enclosed the fact that the silicon substrate expanded at a different rate than the resin itself , this dissimilar expansion caused chip cracks over very small temperature differences – however this is only with memory , when it comes to processors the heat created by the different substrates – dissimilar expansion again – causes problems which is why they need heatsinks , fans etc.
By: Hand87_5 - 9th May 2003 at 19:20
You’re both correct.
Let’s come back to basics.
CPU’s are mainly based on transistors. In oprder to make it simple , there is 3 main kinds of transistors :FET , BJT , MOS.
Microprossesors are based on Cmos technologies. This technologies allows fast switch and has a main advantage : consumption. A MOS transistor doesn’t consume any energy at steady state. However , the interesting behevior is happening when a CMOS change of state and delivers 0’s and 1’s.
There the CMOS consume energy during the transition time. As you all know , the modern CPU’s go faster and faster . The state of art right now is around 4 GHz.
A transistor is supposed to change is state 4e9 times/s.
It will consume energy 4 billion times/s and therefore ……. generate heat.
This is one of the issues that the ingeneers have to deal with.
By: mixtec - 9th May 2003 at 19:10
Originally posted by Vortex
Can you explain that a bit? EVERYTHING generates heat….bio-computing, DNA-computing, optical-computing….The problem with electronics at higher “bandwidth” (actually a stupid name because it’s not really a “width” in this case) is the loss of electron conduction and other “things” may have a higher “bandwidth” but eventually it’ll start to fall off too….so, heat is always always an issue.
This is starting to go over my head, I was hoping someone with more of a background in this would jump in. But yes, I agree that everything creates heat to some point, but some alot more than others. For example where I live, I have to have a room fan blowing on my computer or the thing will get hot as an oven. The temp on my Athlon 1.5mhz would regularly hit 170deg F before I used the fan. Why? Because AMD uses aluminum instead of copper like intel, which has higher resistance. I supose if I ran my processor in liquid nitrogen it might superconduct and then I could really overclock the thing. Working with light, Id think the heat gernerated would be minute. I dont know enough about DNA, biocuputing, etc to comment, but I dont think theres a big problem with heat and therefore “layers”.
By: Arabella-Cox - 9th May 2003 at 02:40
Originally posted by mixtec
Heat generation is a very major roadblock to electromic circutry. It effects not only the “layers” you can work with but mhz speed. Your not held back by that with all the other emerging alternatives.
Can you explain that a bit? EVERYTHING generates heat….bio-computing, DNA-computing, optical-computing….The problem with electronics at higher “bandwidth” (actually a stupid name because it’s not really a “width” in this case) is the loss of electron conduction and other “things” may have a higher “bandwidth” but eventually it’ll start to fall off too….so, heat is always always an issue.
By: mixtec - 9th May 2003 at 00:22
Votex-
“even today Mixtec you can have multiple “layers”, as much as you want physically via wafer bonding, thining, or epitaxial growths. The problem isn’t “layers” but the heat generation. “
Heat generation is a very major roadblock to electromic circutry. It effects not only the “layers” you can work with but mhz speed. Your not held back by that with all the other emerging alternatives. A thousand layer computer chip would have about 250 times the processing power of coventional chip much the same ways computer chips are connected in parralel to make cheap supercomputers.
“There’s an ideal bus width “ie, 64, 128 bit etc” based on what the application is for. If i only need a 8bit address…what am i going to do with the rest?”
Your going to make a faster computer is what your going to do with the rest. It can take up to 150 instructions for a computer to access a simple 8 bit address. All those pipelines are going to get everything done at once instead of cycleing around 150 times.
Interestingly, high speed serial with multiple pipelines or channels can be quite fast too. With current silicon based ICs, i’ve heard predictions that the technology can be stretched to at least 2020…so, that’s still quite some time”
I think things are already being stretched thin with conventional circuit technology. Although that doesnt bother me as Id like technology to remain as it is for I while and give me time to catch up.
mongu-
“I think that by making chips more and more complex there is a risk of high production losses due to flawed chips.”
Thats the other limiting factor, the “wires” inbedded in chips are reaching the limit of how thin they can be made which is why molecular technology is being sought to take things even farther.
“Perhaps the future will bring lots of smaller chipsworking together, like brain cells.”
Its being done now, you can buy mother boards with dual processors, servers have been doing that for a long time. But theirs a severe speed loss (measured in nanoseconds) which is another reason that chips just keep getting smaller
By: mongu - 8th May 2003 at 22:50
I think that by making chips more and more complex there is a risk of high production losses due to flawed chips.
Perhaps the future will bring lots of smaller chipsworking together, like brain cells.
By: Arabella-Cox - 8th May 2003 at 21:02
..
even today Mixtec you can have multiple “layers”, as much as you want physically via wafer bonding, thining, or epitaxial growths. The problem isn’t “layers” but the heat generation. There’s an ideal bus width “ie, 64, 128 bit etc” based on what the application is for. If i only need a 8bit address…what am i going to do with the rest? Interestingly, high speed serial with multiple pipelines or channels can be quite fast too. With current silicon based ICs, i’ve heard predictions that the technology can be stretched to at least 2020…so, that’s still quite some time.
By: Arabella-Cox - 6th May 2003 at 01:38
I like the potential for light computers better… currently we have gone from roadways 8,16,32 and soon to step into 64 lanes wide. Unfortunately down each road the vehicle (data) must arrive before another can be sent, though they travel fast so the delay is nowhere near as bad as if we were using real cars. Each vehicle carries a 1 or a 0. With a light computer simultaneous vehicles can be sent depending upon the speed of the transmitters and receivers, and each vehicle can carry a whole page of more of info…
Will take some time to make it small however, though some light computers are working.. they are just too big right now. (And probably not very efficient).
By: Ja Worsley - 5th May 2003 at 10:34
Neural net interfaces, Bio-organic terchnology and data crystals may have seemed something out of sci-fi series of not long ago, but all to often these days what is science fiction soon becomes science fact.
Becareful about what you wish for, soon it may come true. Don’t forget about the down side of all of this, machines that will soon run man, not man run machines!
Sign In
May 5, 2003 at 8:23 am