Sign in with
Sign up | Sign in
Your question

Why haven't we moved beyond silicon yet?

Tags:
Last response: in CPUs
Share
August 2, 2009 6:54:44 AM

I'm not sure if this is the correct forum for this discussion, but it's the closest match I could find. I apologize in advance.

Anyway, over the past few years I 've kept hearing about amazing new CPU technoligies that never see the light of day.

We seem to be a plateau of computing speed at the moment as there's no cost effective way to break the 4 ghz barrier... *if* we stick with old fashioned silicon cpu's, so we're moving to multi-cores instead.

Silicon-germanium chips could run at 90 ghz and could be cheaply made due to to the fact that they'd use the same manufacturing process as silicon chips.

I've also heard of Ultra-Thin Wafers, Optical Computers, Crystal Lattice Quantum Computers, etc... I'd provide references but It's almost 3am here and I'm tired >.>

Long story short, It sounds like there's ways out there to make a chip that would run at 300 ghz, or have the equivalent computing power, and could be made cheaply enough to sell to the home user.

So where are they?

More about : moved silicon

a c 83 à CPUs
August 2, 2009 7:09:39 AM

Perhaps you've heard of something I haven't, but the techs you mentioned aren't commercially ready yet. Just because you've got something working in the lab doesn't mean its ready for prime time. For example, the optical computers I've heard about work, but they are room sized and so large you wouldn't be able to run one in your home.
a b à CPUs
August 2, 2009 7:14:07 AM

Or quantum superconductors which have to run at 4 degrees point Kelvin.
Related resources
Can't find your answer ? Ask !
August 2, 2009 7:22:26 AM

Or, the original Univac. Vac? Yes, vacuum tubes, The transistor we use today was suplnted slowly in AM/FM pocket sized radios in the early 60s, before that, vacuun tubes, being used over 30 years.
Once we actually see a few of these at the enterprise level in use, itll go commercial, soon after.
Its why we didnt see 15000 rpm platter speeds in DT use for awhile. The failure rate when dealing with new tech is very high, as we just havnt perfected the things needed to acquire consistency, and is why its so expensive, having unique tools creating unigue archectectures using new materials with little known faults/improvements creating huge yield failure for mass production, so eventually the work with it til its good enough for someone (usually almost always enterprise, or sometimes government exclusive) to pay these high unique failure added costs
August 2, 2009 7:53:59 AM

Silicon is so cheap and well understood chemically, physically, and business-wise that it probably won't make sense to supplant it for high volume manufacturing of CPUs until the fab development folks can't squeeze any more performance out of it. People are looking into all the technologies you mentioned, but until there's a clear return on investment in terms of either performance or price, they'll remain in the labs.
August 2, 2009 8:08:06 AM

But even silicon has its end. The tools used are already coming into question as to its return inxestment, as we go to smaller nodes. The wider approach? It may or may not pan out, depending on how much is found to be serial vs MT in common apps, and accross the board. We can see the transitional need coming, but just not yet
August 2, 2009 8:12:11 AM

I learned this in Materials course for my electrical engineering program. Basically we still use silicon for several reasons:

1. Cheap material, silicon is dirt cheap because it is very abundant in Earth's crust. Manufacturing facilities can be anywhere because you can pretty much get silicon anywhere. Other materials are located at certain places or needs to be synthesized.

2. Silicon has all the right properties, high melting point, non-toxic, physically durable, forms oxides well to insulate the PCB.

3. Current materials such as Gallium Arsenide or Germanium are not environmentally friendly. GA is toxic, does not make for a good insulator and is more power intensive. Germanium has low melting point and is sensitive to light.

4. Electronics manufacturing is highly specialized, it will take billions of dollars of investment to restructure existing capacity.

So here are the main reasons for still keeping silicon. Silicon is just so cheap, its not possible for other materials to compete against it. Until any promising materials can be produced at the price of silicon, silicon will always dominate electronics. No one definitely wants to pay 10 times more for the same comp they have now.
August 2, 2009 8:25:55 AM

It would require a promising (available node shrinkage) material at higher costs using a much easier manufacturing method with similar performance, while also showing future scalability to smaller processes.
Nothing probably will ever "out cheap" silicon, but its the overall performance/future performance and manufacturing costs that will most likely drive that next material.
Its why laser is so popular, lights cheap, found alot of places heheh, but it points to the fact it may be the only alternative when it comes to material usgae/costs vs silicon
August 2, 2009 8:35:00 AM

Yes you are right, for now because of the economics involved we will stick to silicon. There are few promising materials out there but only in the labs. Only time will tell, and you only need to wait 10 more years or so. I'm personally betting on some sort of carbon-metallic compound.
August 2, 2009 9:07:39 AM

If we dont have a good idea, and this question is still being asked 6 years from now, we will have reached a stalemate, at least for a period of time, as bringing it to market, and ramping production would take another 2 years, whatever the replacement
a b à CPUs
August 2, 2009 9:16:56 AM

One of the drawbacks to moving to new materials is backwards compatibility. Much like the 86x architecture is now. In others words, in technology you have to be able to go backwards as well as forwards at the same time. That and the powers that be control the market (Tesla vs Edison at the turn of the century comes to mind) plus the material that it's made of has to be abundant and cheap to process.

You may want to have a look at the properties of Graphene
http://www.sciencedaily.com/releases/2009/07/0907310900...

Here's an article on a first run with Graphene
http://www.physorg.com/news148916104.html
a b à CPUs
August 2, 2009 9:33:26 AM

Investment in new technology costs alot more money than trying to milk current technology. If it ain't broke, don't fix it. Why can't we teleport? Because we have planes to move us around (ok, not entirely accurate :D ).
a b à CPUs
August 2, 2009 10:17:51 AM

I can teleport.

How else to I end up on all your screens?
August 2, 2009 10:23:00 AM

amdfangirl said:
I can teleport.

How else is my hair blowing all over?

Fixed
a b à CPUs
August 2, 2009 2:31:36 PM

JAYDEEJOHN said:
It would require a promising (available node shrinkage) material at higher costs using a much easier manufacturing method with similar performance, while also showing future scalability to smaller processes.
Nothing probably will ever "out cheap" silicon, but its the overall performance/future performance and manufacturing costs that will most likely drive that next material.
Its why laser is so popular, lights cheap, found alot of places heheh, but it points to the fact it may be the only alternative when it comes to material usgae/costs vs silicon


Actually, nano-self-assembly might hold the promise of significantly lowering manufacturing costs in the future. This would probably be limited to some regular array structure (memory for example), at least initially. IIRC some researchers at MIT built a self-assembling nano-car, that would roll around when lit up with a light beam.

Of course, next step would be nano-flat-tire change :D .

August 2, 2009 2:47:27 PM

Sounds outta this world, nano nano
August 2, 2009 2:51:14 PM

Perhaps I'm just a wild crazy guy, or I don't push my system hard enough(not really a gamer, i play freelancer,freespace 2, and emulators whenerver i have the time), but my system runs at 3.2ghz and it runs quite nicely,so the speed is fine for me, would love to have one of the 8 or 12 core amd chips coming out. On a side note, i used 64 bit windows vista business and windows 7 ultimate, and vista wasnt as bad as i thought, windows 7 however was spectacular, now i feel comfortable to replace xp pro and i was just using the release canidate, the final product should be awesome
August 2, 2009 3:05:48 PM

Need more speed, tho everyone may not agree with me
a b à CPUs
August 2, 2009 4:32:28 PM

JAYDEEJOHN said:
Sounds outta this world, nano nano


LOL, actually it's nanoo-nanoo :D .

IMHO: Silicone > silicon > silly cons :whistle: 

But then I missed my daily fix of Robin Meade this AM...
August 2, 2009 4:34:31 PM

One thing I can deduce from this is, Robin Meade >Robin Williams
a b à CPUs
August 2, 2009 4:49:52 PM

JAYDEEJOHN said:
One thing I can deduce from this is, Robin Meade >Robin Williams


LOL - there we are in 110% agreement! :D 

August 2, 2009 6:23:37 PM

Evolution_13 said:
I'm not sure if this is the correct forum for this discussion, but it's the closest match I could find. I apologize in advance.

Anyway, over the past few years I 've kept hearing about amazing new CPU technoligies that never see the light of day.

We seem to be a plateau of computing speed at the moment as there's no cost effective way to break the 4 ghz barrier... *if* we stick with old fashioned silicon cpu's, so we're moving to multi-cores instead.

Silicon-germanium chips could run at 90 ghz and could be cheaply made due to to the fact that they'd use the same manufacturing process as silicon chips.

I've also heard of Ultra-Thin Wafers, Optical Computers, Crystal Lattice Quantum Computers, etc... I'd provide references but It's almost 3am here and I'm tired >.>

Long story short, It sounds like there's ways out there to make a chip that would run at 300 ghz, or have the equivalent computing power, and could be made cheaply enough to sell to the home user.

So where are they?


because none of the tech you mentioned can be found in a consumer-level stable form, e.g Quantum computing is only stable for 1 microsecond.

also because of Intel, software, oh and did I mention Intel?

However Graphite processors can be created in a stable form and can reach 1THz, there is also a type of processor that releases no heat.

but because of software and Intel, these won't be implemented because extreme overclocking levels on bottom-end CPU = less sales on the top-end.
August 2, 2009 7:41:36 PM

I'm in complete aggrement with so much of what has been said. Better, Faster, Smarter, Cheaper. A point that no one has really brought up is the software side of things. Software is lagging way behind. Now we have 4 cores everywhere and a 6 core opteron and a 80 core cell. But most software is only running on one core and one thread. How about two i7s stuck together a-la core 2, running at say 3.6ghz, 8 cores of silicon running on software coded for 8 cores and 16 threads even 32 threads so much the better. They could call it i14. At this point and for many years to come software will be the holdup in computing speed. Is there even one game out now that can take full advantage, and I mean full advantage of a four core cpu? Windows 7 may finally make 64bit work. Yes I know Unix but why not 128bit by now? And if it has to be incriments of 8 why not 400bit. The software guys need to catch up. To be fair to the software guys one problem is the crybabies who want 8 bit and 16 bit software to work on any x86 system. There will come a time that we will have to say goodbye to old software. Crysis is a good example, anyone seen any example benchmarks of OCed i7 965 systems with 3 295 gpus in sli that cannot sustain 60 fps min at a 1920 x 1080 res.? Why is that because the silicon or the hardware is at fault or becasuse the software is written like crap? We'll all have uber expensive multi GHZ, multi core, multi gpu, machines built of exotic materials but no one will have coded the driver for the printer. Oh wait we have that right now. Welcome to the furture.

Is anyone having a moment of ZEN with Robin Meade?
a b à CPUs
August 3, 2009 9:40:32 AM

The moment they figure how to mass produce graphine based microprocessors, will be a great day. Graphine conducts electricity better than graphite whilst still remaining an isotope of carbon, so you do find trace particles in common pencil 'lead'.

That'd be cool.
a b à CPUs
August 3, 2009 1:47:19 PM

Diamond is a much better semiconductor than Silicon, it can operate at much higher tempratures and frequencys, and is probably the ultimate semiconductor, the problem is of course the price.
August 3, 2009 2:19:15 PM

Not if we get all those exes to contribute their old rings heheh
a b à CPUs
August 5, 2009 5:23:14 AM

I don't know why most girls insist on getting diamond rings.

I prefer to have the title "Miss"
a b à CPUs
August 5, 2009 10:29:22 PM

^ IMO, you can thank de Beers. Both the company and the beverage.

Also, after arduously researching lots of Hollywood fanzines, I have discovered that celebrity diamonds can usually be found near large deposits of silicone. :D 

Clearly this is a much more valuable resource than the aforesaid "exes"...


August 5, 2009 10:50:43 PM

Need to keep those sharp objects away from the silicones tho heheh
Maybe itll become a trend, and more diamond will be left over, and all that extra silicon can still be put to use
a b à CPUs
August 6, 2009 6:47:41 AM

Let's try alchemy
!