Slashdot Mirror
Cringley: Chip Manufacturing To Radically Change
eshefer writes "
This week cringely talks about a company called rolltronics which he claims will make the current microprocessor fabrication on silicon wafers technology defunct in five years. The company uses roll-to-roll printing on plastic (somewhat like newspaper printing presses) making the process much cheaper to produce then current technologies. "
Why do you need 1 CPU clocked at 1 GHz... if you can have 1000 CPUs clocked at 1 megahertz?
Think about it. Ever heard of the 'connection machine'? Web printing of CPUs at unprecedentedly low costs is a situation that _begs_ for massively parallel computing.
Suppose you could print up a big screen and behind every pixel is a 1 megahertz computer? Let's see- let's imagine the screen is three feet high and... well, any length, right? Make it as wide as your room. Call it 30 dpi and assume you're viewing it from a fair distance. That's 1080x4320 pixels (or so) for 3x12 feet. If each pixel's running a computer at about 1 megahertz that's more than four thousand gigahertz of CPU ;) _and_ you can paper your wall with it.
Yes, getting _information_ to those clever pixels would be the trick, but it's not an insuperable problem. The CPUs can be pretty smart at one megahertz. Give them a few K of ram too, think of them as like a peculiar sort of 'accelerator card' only massively parallel. It's absolutely trivial to do hacks like Asimov's 'Prime Radiant': you'd just have each 'pixel' an alphanumeric generator able to consult an overall RAM location, a pointer to where in RAM to look, and with a brightness control connected to some gaze direction sensor. Presto, wall full of text that scrolls and becomes clearer and follows your gaze- and that is one of the _easiest_ things to do with this stuff.
Better to ask whether you could have a suitcase-sized block with several terabytes of _really_ _slow_ RAM (read: way faster than a HD, and impact proof!), for the cost of a floppy drive- after all, it's only a matter of printing more pages, right? Just keep piling up the 'slow RAM' pages. The same thing could be done as read-only: if you're OK with text, you could have a 'book' (i.e. an object with a screen on it) that is ANY book you like, with terabytes of data in it, printed so cheaply as to be virtually disposable. Look around at your books and ask yourself how much it'd cost you to have a _scribe_ write those out for you ;)
Ask whether the massively parallel 'screen' concept could be used for video game systems (imagine unrolling a wall-sized Quake, or flight sim- even if the quality is not ultra refined, some aspects of it would be as advanced over the current state of the art as the current state of the art is over software rendering- for instance, perhaps you'd have only solid colored triangles BUT the system would build them itself, being fed only very high level object information- and you'd get 100X or 100,000X the model geometric detail you can have today, because each pixel only 'sees' a couple triangles directly behind it and decides only what color it's going to be. So you wouldn't get 3Dfx motion blur but you'd get every leaf on every tree in the forest, as each pixel handles its own (emergent) geometry. Even if you only had Atari 2600 level pixel sophistication- think of the resolution! You could have new games designed to take advantage of that- like, I dunno, a Pac-Man that has just the one maze but it's the size of the wall :)
Or, each pixel knows how to do MPEG transforms- so you unroll a videophone or movie screen the size of your wall. Power's an issue, but if you can get wide enough traces... like a foot wide... a lot of power dissipation issues become less of a problem :) really, the idea has loads of possibilities that don't require heavy centralised processing.
Absolutely. 300 dot _line_ _screen_ is high for mass printing: maybe that's what Shotgun's thinking. Besides, I bet a lot of the action's going to be in large format- doing stuff with wall-sized screens, or medium-res displays that are 2' tall and 6' wide and curve all around the back of your desk. With surface area like that who needs super high resolution? Big screens are cool :)
I could see similar uses in things like scanning, where you're dealing with a lot of parallel data anyhow. You could print some sort of processor that gets rolled up and stuck to the scanner's sensor bar, that interpolates and sharpens and does all sorts of nice image enhancements very cheaply- and have the scan go much faster than current models do. Output bandwidth would be the bottleneck- unless the rolled-up processor also encoded the result into JPEG for you :)
1-Why is the crack smoking junkey who put that website together?
2- Anita Borg, Ph.D. -- Product Innovation and Social Responsibility. Hehe. I just think it's too funny.
Palin...
It would be worth considering that should this scenario come true, it would have an interesting impact on the usages of free-vs-commercial operating systems. If the computer costs $15 to make, people are not going to be spending $80-$100 to put Windows Whatever on it.
.NET and similiar network-based hosting/application providers as a place to store data and applications on, with the $15 computers being treated as more of an access device than a computer -- the catch would be whether or not the monthly service charges or what have you over the long term were cheaper than buying a 'real'/non-disposable computer with software or not.
On the flip side, if the computers are 'disposable', then this might drive up interest in MS
At least you clicked on the link and learnt something, which is more than the dumbass moderator who thinks the post is ''offtopic''
Ball Semiconductor have at least as interesting a plan to deposit semiconductors on small spherical surfaces. They have some small gates working already.
And it just may BE your shirt pocket. That's what Cringley probably knows but isn't saying - When it becomes that cheap to just "print" a computer, they'll be integrated into everything: refrigerators, automobiles, clothing, furniture, you name it. Sure, there will still be information appliances, but their purpose will evolve into enabling your coordination of all the other computers you will interact with throughout your day, from your own household accoutrements to public infrastructure to your employer and the internet at large.
It is going to make the world unrecognizeable.
Again.
And the amount of information that will need to be exchanged is going to make today's bandwidths look like trickles. Right now, we are at the knee of the exponential growth curve of the telecommunications market, and technology will keep up with demand as improvments in optical swithcing continue. Communication service is going to become more important than banking - hell, banking and finance has already become little more than information flowing around a network.
You want to be a part of it? Forget putting your money in the people who make computers. Invest in telecommunications, and the hardware that supports it. That's where the fortunes are going to be made.
I can see the fnords!
These are just areas where we will see the incremental improvement that Cringley described.
Hell, in 1988, when I bought my first CD player, the hard drive couldn't store even one track from a CD. Now my hard drive holds dozens of ripped CDs, in many cases uncompressed.
I can see the fnords!
Well, aside from trying to poke holes in your math, there are (at least) two other reasons why these "plastic" computers are going to be a big hit, despite the fact that they're lower on the curve than ones made out of polluted sand: - Cost. The organic materials are much less expensive than the ultrapure materials required to make semiconductors and hard drives. That's why CDs are so cheap. (At least as long as the oil holds out.) - Ubiquity. When you can print a computer on any old visibly clean surface, not just plastic sheets, then you've turned an important corner. You can now put computers into eyeglasses, furniture, windows, coffemakers, even underwear (just imagine!) It doesn't matter how fast the machine once you pass certain computational thresholds: the ability to support a graphic interface, the speed to reproduce audio,and another milestone for video. Each of these thresholds open up yet more applications for embedded computing.
I can see the fnords!
Linux BeOS FreeBSD MacOS X QNX
SUB-20000 USER ID FOR FREE!
-Joe
Yes, because the is a larger lower bound of feature size they are not going to be cranking out 1GHz CPUs on this stuff, especially when the chip is the size of a sheet of paper - it takes electrons time to get from place to place.
So this limits the power of individual processors using this process, but you can go massively parallel, just add another processor page, or 3 or 10...
This may never produce a barnstormer of a computer, but it sounds promising for consumer electronics and web appliances.
-josh
You're thinking in terms of silicon. It's not a straight scale-up when you're changing the basic materials. It's my understanding that a polymer-based CPU of that size would generate less heat and use less power than a silicon-based CPU sized as they currently are.
Rechargeable batteries.
What if someone accidentally rips my laptop? :)
Have you painted a shed today?
Yeah, but you yanks have no honour.
Well, they tried marketing DVD's that expire (DIVX) and look how long that lasted!
I don't doubt that there would be an easy and economical way to recharge/reuse/replace the power cell.
You are in a maze of twisty little passages, all alike.
There was an interesting Wired Magazine article that discussed the work being done by Paper Computer to make cheap flat computers.
There was a Slashdot article about these guys over a year ago.
-----
nuclear iraq bioweapon encryption cocaine korea terrorist
Considering how reluctant paper is to disappear down the hole with our lousy 1.5 GPF toilets, I seriously doubt the computer would go anywhere.
GPF=Gallons per Flush
Jesus was all right but his disciples were thick and ordinary. -John Lennon
Great. Now you can have the dog fetch your laptop. I hope that they find a way to prevent it getting soaked at the same time
----------------------------------------------
the pun is mightier than the sword
Actually, there are at least two companies. Dieceland Tech Corp is promising a $10 phone, and this Register story says a $20 laptop also. The DTC "future" page has an image of a laptop but no details.
So you're saying that we're not going to see any new technologies because it is rare for a new technology to become mainstream? That's ridiculous - of course we'll be seeing new technologies appear, sooner or later. Sure, most fail, but not all of them.
"Than" and "then" sound very different when spoken in a good old South African accent. In general though I don't have that problem at all, I've never had much of a problem with spelling. I think its a genetic thing or something, some (otherwise intelligent) people seem to struggle with spelling. Or maybe it has do with how much a person reads. Reads books, that is, not websites like slashdot.
I don't have all the numbers to fill this in, but here's a try.
The technology is aiming to be available 5 yrs from now (notice how long shots are always 5 yrs out)
The latest generation of microprocessors have around ?20?million transistors. 5yrs will give us 3 more doublings (assuming Moore's Law holds), so will be looking at replicationg 160mil transistors for the processor. 1Gig of Ram, add another billion. All the other circuitry, lets just make it simple and say that an average computer will have 1.5billion transistors in 2005.
A magazine has around 200 pages. So each page of this computer will have to hold something like 7.5million transistors (assuming an even distribution).
Assuming they can print at 300dpi (which I believe is high for mass printing) on 8inx11in media gives
300x8x300x11 = 7.92 million
This may look like it will pass until you consider that a transistor will take more than a pixel and then consider inter-transistor wiring. If this is enough they will barely be cutting the edge unless:
-they can print at higher resolution
-they can print more pages
I don't see the need anyway. Computers are cheap now. You can get one for $100. What's expensive are the latest processors, and they're not expensive because of production cost. It's recapturing the engineering cost that drives up the price. This will only produce $15 computers if someone is willing to pay $500million for the first one.
Aah, change is good. -- Rafiki
Yeah, but it ain't easy. -- Simba
Advantages:
Drawbacks
A often-quoted great app is the head-up display for cars: a transparent set of electonic circuits that you glue on your windshield and contains its own display. UV protection films are mandatory for keeping the circuits from burning in the summer, but it looks feasable and cheaper than the usual optical projection solutions.
Don't sell that $12 million 193-nm optical stepper in your silicon fab, though. We're not there yet, especially for medium or high speed circuits.
--
Mad science! Robots! Underwear! Cute girls! Full comic online! http://www.girlgeniusonline.com/
If this is true, then that's a lot of cheap computers, that are going to need an operating system. One that's already demonstrated an ability to be easily ported to lots of architectures. This could be a big win for Linux.
Key to financial independence: Spend less than you earn. Save and invest the difference. Do it for a long time.
Think more along the lines of magazine presses which are less likely to exhibit flaws (in my experience). The production of newspapers is done with less concern about details - as long as the thing is basically readable, no one's going to complain about a $.50 paper and the newspaper presses know this. As the circulation goes down and cost of issue goes up, people are more likely to complain. In this case, the circulation is going to be extremely low - everyone isn't going to be buying a computer every day, week, or month. The company would naturally have to make sure that this roll process is accurate enough that they can limit the number of misprinted computers to an amount that can be recovered by profits without a problem.
I rent DVDs for $5, when with DIVX I could have "purchased" the physical item. Besides the convience of having it closer for more replays (paying the $5 again), or being able to buy it for life ($20), I have nothing when I return it to Blockbuster. And late fees? Pfft.
And what would have made DIVX even more tempting if it caught on was the opportunity to hack the player to play "expired" disks.
So what I'm saying here is.. I don't know what I'm saying. In DIVX it's a good idea, because a video disk is an item you'd perhaps rent. But for a computer? I like my comptuers big, grey, and power hungry. I'm barely sold on the concept of batteries, let alone in computers.
-bugg
On the other hand there's a whole range of electronics out there where this sort of density is not an issue and this could make a lot of older fabs that are building this stuff redundant.
I could imagine a cool disk drive replacement with this technology - basicly a pile of mylar sheets - I bet you could get comparable densities at similar prices .... and you wouldn't have to spin them ....
Nowhere in Cringley's article is there any discussion of the performance penalty that this process would entail. Let's assume we want to duplicate the equivalent of 50 million transistors clocked at 1 GHz. Right now Intel can squeeze that many components on a 200 square millimeter die by using a CMOS process with a 0.18 micron feature size.
Now assume that your printing process needs transistors with 10 micron feature sizes to ensure proper registration and a high enough yield to be manufacturable. That increases your effective "die" to 956 square inches. (Area increases with the square of feature size.) That's equivalent to 10 sheets of single-sided paper.
For a multi-layer printing process, 10 layers of plastic sandwiched together would definitely be possible. HOWEVER - you are not going to be able to clock your circuit at 1 GHz! Because of the much larger size (and capacitance) of your circuit, you'll do well to get a 1 MHz clock speed (1000X slower).
While this process may be very useful for e-books, displays, etc., I don't see how any high-performance computing could be done with a microprocessor constructed with this technique. Your only alternative to slower clock speeds would be massive parallelism to achieve higher computational throughput. Assuming a direct tradeoff of speed versus number of transistors, you would need 10000 layers instead of 10 layers in your process. There goes your low manufacturing cost.
It's not just enough for a computer to be cheap. It's got to be fast, or it's no good to anyone.
-Daniel
Sounds like we've been through this path before. Unfortunately, developing new technologies rarely works and just because there is a company dedicated to it doesn't mean much more. Remember 3d protein memory based on lasers and rhodopsin? I'll believe it when I see it.
-Moondog
Here's a Google cache from the rolltronics site: http://www.google.com/search?q=cache:www.rolltroni cs.com/Roll2roll.htm. I'm sure Google can handle the traffic much better than the original seems to be holding up...
Think outside the... Hey, where'd the friggin' box go?
Wouldn't the length of traces eventually represent a problem? We're already running into 'speed of electrons' problems with current designs, wouldn't an 8" CPU only magnify the problem, and create a speed limitation?
Think outside the... Hey, where'd the friggin' box go?
"Where to start? The circuit size will drive up power usage and heat generation."
Where to start? Decreasing density leads to better heat dissipation. Changing fabrication materials could mean less heat generation. Size doesn't mean anything so long as no space is wasted; moving outwards in the x axis, instead of adding gates upwards to the y-axis are equivilants. ie: building out instead of up.
Ace
What he was saying (with the whole Longbow/rifle analogy) was that, while the technology may seem to be inadequate/kludgy now, it may not be in the future.
AFA all the 'it won't compete with a X Ghz processor' has it occured to anybody that you print the pages for the supportable stuff (battery/display) and all 100 pages have a notch dead center where a real processor is dropped JUUUST before you laminate on the keyboard?
So then you HAVE your 1 Ghz, $40 laptop.
Enough is enough I say! Bring back the 'a' in than!
------
And the old-line companies like Intel and AMD, which are currently fighting over which is the superior obsolete semiconductor company, well, those outfits go out of business
Bzzzt Wrong.
Intel or AMD BUYS our friendly RollTronics and maintain their positions in the new era --or-- they get involved enough in the technology and prove that it, in fact, does not work (in order to protect their $XXX Trillion dollar fab investments)
If you're ever stuck in a bathroom stall that's "empty" you'll have to severely double check the piece of scrap paper you use... you might be tossing/flushing your laptop by mistake.
What a concept.
Imagine sending THAT back to tech support for repair: "Reason for repair:" Euhh..... "Curry Related Emergency?"
--- Jump!! Fire!! Bullet time!! - Lego version of the Matrix
While I'm with Cringely on most of what he says in the article, the 15th Century longbow/arquebus (they didn't have muskets then, damn it!) comparison is a poor one. Essentially, firearms underperformed the longbow right up until the middle of the nineteenth century when breech-loading and mass-produced rifles made them faster and more accurate (hand-built muzzleloaders were slow and inconsistent) than the bow.
The reason that firearms replaced the bow some three hundred years before they were its technical equal was economic: an archer required years of training to have the accuracy and muscle development required to be any use at all in combat (archers can be identified from their skeletons, having asymmetric bone thickening in their arm bones) and have to be fit and well on the day as the physical effort required to discharge thirty arrows rapidly is huge.
A muzzle-loading musket, despite having a 2-metre circle of probable hit at fifty metres range and a rate of fire perhaps a tenth that of a longbow, has a *way* lower ammo cost, a training overhead of perhaps a fortnight (and you can teach it in an afternoon if your student is bright and you don't care if he hits anything), and no great strength or stamina required since the kinetic energy that does the damage doesn't come from the soldier's own muscle.
The solution to the technical difficulties is to use them en masse, mix them with pikes or give them bayonets for close work and fight on the defensive if at all you can.
Basically, the longbow was betamax to the musket's VHS...
-- AndrewD
A Maze of Twisty Little Laws, All Different.
I suspect that the most important use of a technology like this will be to produce flat screen displays - driving the current high cost of HDTV and flat screen monitors way down.
Yes, developing new technologies rarely works. That is why we are still in the Stone Age.
Last time I checked, most plastics have a large thermal expansion factor. Most modern CPU's have 5 or 6 metal layers of lines to connect things together. Materials are carefully chosen so the silicon, Inter Layer Dielectric, interconnects, and packaging all expand about the same. Can you say thermal cycling and stress crack failure? Actualy I see this technology being useful in something needing less than 3 connect layers that are not metal (limiting speed due to resistance) like an Active LCD or full color LED display. I wonder if they can make color LED's with this stuff. A large bright color display you could roll out on the wall in the confrence room would be neat.
The truth shall set you free!
Why the f$ck would anyone think computers are going this way? Smaller is a trend. Wearable is a trend. Remote processing is a trend. All of which can be pushed to utterly ridiculous limits within the next decade! This Rolltronics seems more like a scam, especially with lines like "This is a multi-billion dollar opportunity."
My wristwatch has more processing power than the first computers. /. Ran an article on a wristwatch that runs Linux. The ultimate CyberGeek I know had LCD glasses (prototype), Nintendo gloves, and a book sized unit that made Xbernaut look archaic. Not to mention full time wireless hookup to the net. While Cringely discounts 'incremental' changes, in ten years that's going to be reduced down to contacts, a wrist wrap (nerve sensors), and something the size of a pager. Hopefully running on an ethanol fuel cell.
For my $.03 CDN, the reversible switch is probably a better bet, as it allows 3D 'chips', without the heat problems. Quantum is still a ways off. And Rolltronics is going nowhere.
Cringley says that the battery will be intergrated into the stamping procedure. This could be _really_ bad in my opinion becuase once the battery runs out so would the "computer".
Lets say you pay for this month's Wired and comes via a wafer-computer. You read it and enjoy the interactive articles and eyecandy. Life seems that much cooler.
Next month you want to re-read that artice. Too bad the battery is dead. Now you gotta pay for last months issue again.
This seem like too much control over content I paid for. We are already bitching about DVD region encoding. At least DVD's don't expire.
Leknor