Slashdot Mirror
Intel Devises Chip Speed Breakthrough
Chad Wood writes "According to the New York Times (free reg. req.), Intel has demonstrated a research breakthrough, making silicon chips that can switch light like electricity. The article explains:''This opens up whole new areas for Intel,' said Mario Paniccia, a an Intel physicist, who started the previously secret Intel research program to explore the possibility of using standard semiconductor parts to build optical networks. 'We're trying to siliconize photonics.' The invention demonstrates for the first time, Intel researchers said, that ultrahigh-speed fiberoptic equipment can be produced at personal computer industry prices. As the costs of communicating between computers and chips falls, the barrier to building fundamentally new kinds of computers not limited by physical distance should become a reality, experts say.'"
So when do I get my new high-speed fiber line? :D
Does this mean that we may be able to overclack without regard to temperature? Will optical technology make the processors run cool? I really hope so.
... is the coolest technology you've never heard of.
For some reason, buried among a zillion dog-eared back issues of "People" and "Sports Illustrated" at the Seattle's Best Coffee shop at the corner of Central and Kirkland Way in Kirkland, Washington, somebody left a copy of Photonics Spectra in the magazine rack. I'm an electronics geek who had never heard of the field, and I probably spent three hours and two quad-damage lattes poring over that magazine. Fucking amazing stuff. Spend some time at the photonics.com website if you don't believe me.
Seriously, photonics looks like it might be the Next Big Thing.
Great now we'll only have to buy from two companies in the future Intel and Microsoft.
Seriously though, when I hear some chip news, and how it's the 'next best thing' I kind of wonder how much is just marketing hype. So far I heard of terabyte chips... Coming Soon!!!... Faster chipset will do... and so on. Yet on the market you see none. According to most companies capabilities (providing it's not just hype), from what I gather, they have a chipset in the works that can fly you to the moon, wash your car, bone your partner, and have you back in time for work the next morning. However, these companies have to make as much money as they possibly can selling you their fourth, third, and second generation chips for the next few years.
MoFscker
I have a feeling this will one day be seen as a development with the same order of importance as say, the development of the first semiconductor. However, it will probably take at least a decade to sort out all of the implications.
And all our yesterdays have lighted fools The way to dusty death. --Will
I love generalization.
Three states have been around awhile it's called Tri-state Logic. Gordon Moore gave an interview in PC Magazine. He discussed multi-state logic, but said it was a non issue. He said that neural networks were much more important breakthrough.
Reliability?
I've had three AMD chips. The 500 mhz K6-2 is still running strong several years later, now in a younger sister's hand-me-down PC. My first 2000+ withstood the power supply exploding and lasted another 6 months before finally giving out. I'm currently using a new 2000+ with no problems. On both 2000+ chips, I have yet to even a STOP error that I can't attribute to something else.
How much more reliable do you want?
Cogito ergo sum in Slashdot.
Well, while it would be tits, it would not be because of the word "tertiary". "Tertiary" is to 3 as "secondary" is to 2.
The word you are looking for is either "trinary", or "ternary".
Either way, if you look at how the word "bit" is formed, you can think of two ways:
1. B-inary dig-IT
2. BI-nary digi-T
If it is the first case, then either "trinary" or "ternary" would still yield "tit":
T-rinary dig-IT
T-ernary dig-IT
However, if it is the second case, we could have a problem:
TR-inary digi-T = TRT
TE-rnary digi-T = TET
But, I agree with your original intent because neither of these are us fun as "TIT".
Actually Intel's behavior in this regard is far worse than AMD's.
With AMD, the bullshit is just a thin (and obvious) marketing layer, which is easy enough to ignore. Intel, on the other hand, release slow chips with high clock speeds because they know the vast majority of morons out there will only pay attention to the MHz rating.
As a case in point, the infamous P4 Celeron. High-ish clock speed, crap performance, completely destroyed by similarly priced AMD processors.
I think AMD's naming makes a lot of clueful people a bit uncomfortable, but seems justifiable in a market dominated by a world-class bullshit artist like Intel.
We live, as we dream -- alone....
Sorry, accidentally posted anonymously the first time:
The limitation on physical distance in an electrical medium is dictated by its impedance, which dissipates the electrical energy in the form of heat. This creates an enormous problem of power loss, which increases linearly with the distance of the transmission line.
An optical waveguide, such as fiber or the silicon waveguides mentioned in the article, see no such losses due to electrical impedance.
Theoretically, as long as the parameters are met for photonic propagation, light will stay in the waveguide indefinitely. However, there are still losses due to imperfections and impurities in the medium itself, caused by microscopic deformities, bubbles, splices in the fiber, etc. There are also some losses dues to quantum effects, which we see in the form of 'evanescent' waves that tunnel outside of the boundaries of the waveguide.
What you really want to be asking is what is the transmissive and absorbtive properties for the silicon medium they use for the particular wavelength(s) of light that they are developing the technology with. If you know that, then combined with the effects above you can get a decent estimate of the power dissipation of the system for a given photon source.
My feeling, without performing the calculations, is that you will be pleasantly surprised at how little energy will be dissipated in the form of heat.
~Loren
Incidentally, this was the trick used on that nifty TI graphing calculator hack which played "music" out the link port. The link port is just a digital I/O on the bottom of the calc, but someone figured out how to toggle it on and off in machine code to use it just like a "1-bit D/A". Plug some headphones in (with appropriate adapter) and you heard some really poor quality Green Day song. The frequency wasn't nearly high to be more than a proof of concept, but it was cool nevertheless.
I should know better, but I just have to comment on the fact that you're one hell of a troll. If you supposedly own an A64 and were going to bitch about the true clock speed, I'd expect you to know that the 3000+ clocks at an even 2GHz, the same as the 3200+, only difference between the two being cache.
The last line of that post was pure brilliance/troll as well. Show me ANY benchmark were a P4 2.2GHz outperforms an A64 3000+ and you might be onto something. Until then, please stop speaking out of your ass.
I was thinking, if they use light, than the limitation on the size of the chip will disappear (or become less important, rather) and you could have a chip big enough so that you can actually see how it works. Wouldn't that be cool?
Agreed. I don't see how this changes the problem of having an expensive (III-IV) device BEHIND the modulator. If you want to send data quickly, you either need a fast (2.5Gb/s is industry standard now) vertical cavity surface emitting laser (VCSEL), or you need a cheap CD laser that can be electro-optically modulated quickly. The other issue is that the cost of the standard 850nm laser inside a short-range fiber-optic module is only a small fraction of the total cost of the module. Decreasing the cost of one single component doesn't "change the world"...
Easiest way to see this is to imagine A and B have an instantaneous communication device. They synchronize their clocks and then separate at velocity v. Some time later (t1), A sends an instant message ("lol d00d") to B. Due to time dilation, A knows B will receive this message when his clock says t2, where t2 < t1. In B's frame, he receives this message when his clock says t2, and he instantly responds ("r0x0r!"). In B's frame, A is moving away at speed v, so the time that B knows is on A's clock when he receives his instant message is t3 < t2. But that means that A receives a response to his IM at t3 < t1, which is before he sent it!
So that rules out instant communication. If you redo this argument mathematically, but allow the speed of the communication to be a parameter, you can find a constraint on the speed of information exchange to preserve causality. It's not immediately obvious to me that it will come out to be the speed of light, though. I suspect that it should, or I'v made an error in setting up this thought experiment.
Having not read the paper, it's hard to say how great this works, but it's worth mentioning that optical microchip clocking may be a major development over the coming decade. As clock speeds get faster (4GHz anyone?), small variations called clock skew and jitter become critical difficulties. Basically, because the clock signal doesn't propagate in an exactly predictable amount of time, different chip parts end up out of sync. Because optical clocking would rely on waveguides, with faster transmission and using uncharged particles that don't pick up random electrical signals, sending clock signals via light waves could be very beneficial. Of course, this development only speaks of the sending end - the modulator - not the receiving end, but we can be sure that Intel and many others are hard at work developing this technology.
OK here is the formula:
cost: number of digits * number of states for a digit
base is b
number is n
cost is c
c = (1+log_b(n))*b
where log_b is logarithm base b
If (d c) / (d b) = 0 and n approches infinity b approches Pi.
In other terms: to store big numbers you better off using Pi based numeric system. 3 is the closest integer, hence the tertiary storage promises to be more effective.
Code poet, espresso fiend, starter upper.
Indeed. Remember that electricity moves at the speed of light. (pause) Yes, the speed of light. (pause) Yes. Not the speed of light in a vacuum, the speed of light in the transmission medium in question. When this is wires on a PC board or traces on a chip, the capacitance and inductance of the wires -- which form the transmission medium -- slow down the photons which mediate the field propagation (at least that's one way of looking at it). For example, the speed of light in a coax cable is slower than the speed of light in a vacuum (although it's usually a fair fraction of it, typically well over 90%).
That said, and understanding that signals on chips are already propagating at fractional-light speeds, you can tell that the original statement is bunk. Why? Because we're already at the physical limitations. It is already true that it takes an appreciable fraction of a clock cycle for signals to propagate from one side of a chip to another. Remember, light -- in vacuum -- travels about 11 inches per nanosecond. Slow that down to 0.3c and suddenly your 3 GHz processor clock means you can get about 1 cm between clock cycles, or, from one edge to another of the big, modern chips.
So, the important question is: how fast do infrared photons travel in doped silicon? Anyone know?
Put my fist through my alarm clock with its ding-dong death inside my ear. - The Blackjacks.
This is a much simpler question that I too am curious about. When Intel does their HDTV demo will the 5 mile coil of fibre get warm -- however imperceptibly -- because photons are "flowing" through it? If so, would that show up dramatically at small scales?
Because 51% of R&D dollars are spent in the U.S. Just like we were the center of industrial capacity in the early 20th century so are we the leaders in the idea capacity in the early part of the 21st. Near universal post-secondary eduction along with programs that encourage the brightest from around the world to flock here are what is keeping America afloat in the world economy. That's why Republican's desire to defund education is so scary, if we lose this edge we will fall as the worlds leading power.
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
Not to get involved in a flame war about CPU's, but the worse experience I ever had was with an intel chip. I overclocked it and it fried the cache. That was my fault.
I buy AMD now because it is cheaper, and outperforms intel chips in most areas. I honestly feel I am getting the best for the least.
Now I own a 64 and am extreemly happy with it. I was worried about buying one so soon after they where released, but I haven't had a single issue. The only issue I had was actually with my soundcard not working in linux...so I bought another and it does.
I am also rather impressed by the new heatsink chassis for the 64. I don't know what P4 is using, but the 64 has something better than I have seen on any previous CPU model.
Anyway, putting it bluntly...in my several years of experience using AMD processor I have never had an issue. I know several other AMD users that also have had wonderful experiences. I honestly don't know what you are basing your opinion on, but from this end it doesn't hold water.
NR
...barrier to building fundamentally new kinds of computers not limited by physical distance should become a reality, experts say...
I was under the impression that physical distance was always a limitation...? Which "experts" are saying this?
in girum imus nocte et consumimur igni
To make a gate that can handle more than one state, you need more than units than states. I mean, I can implement a binary gate really simply, just a single transistor. How would you implement a trinary gate? TRy and design something more simple (taht can be designed on silicon). Also remember that it needs to be usable in the end. This means that:
1) It needs to be usable in the end. Binary is simple, when voltage is present, it causes something to happen, another gate to flip, a value in a memory circut to be set or unset, etc. With a larger set of states you again need more circutry to be able to differentiate one state from another which again increases complexity more than gain.
2) Be able to keep the states consistent. IT's easy with binary, on or off, voltage present or absent. With more states it gets hard, how is one defined from teh next, and what happens if the input voltage changes (which does happen) and changes the amount flowing through. I mean if the voltage sas for a second, does that throw off all calculations? Computers are imperitive devices. It is necessiary that one stage be able to rely on the fact that the result of the prior stage was correct.
3) As I mentioned, you need to be able to implement it on a silicon chip. YOu might be able to get some complex device that daels with a bunch of potentiometres and count those as "gates" but you'd be forgetting that they aren't implementable on silicon as a transistor is. Thus you get nothing workable in teh end.
Look, you're welcome to try and design a higher state chip, but I'll give good odds that you don't get anything even near working. IF you like, I'll run the idea past the EEs at work, but I already know what they are going to say.
Now quantum computers are entirely different. They solve problems in a whole different way and, indeed, work on a different level than conventional computers. But for the normal silicon chips, you are stuck with binary. Nothing else can be made workable.
So stupid to argue over this.. guess what. I have both AMD and Intel chips on various systems. None of them have broken down or burned out. Why? Because I assembled the heatsink and fan properly. They serve their purpose and theres nothing much else to complain about. But here's a question.. why did this stray so far off topic.
blah, blah, blah, blah.
The U.S. post-secondary system is the best in the world without question. Not only by quality (9/10 of the worlds best institutions in any subject area will be in the U.S.) but by quantity. In the state I live in no student is more than 20 miles from a college, university, or branch location. This makes it easy for anyone who wants to get an eduction to get one. As a good example of how the worlds best come here one of our state schools that I wouldn't have even considered as a backup school has students from 157 nations!!!! Our secondary eduction may be lacking in some regards but we make up for it. Besides most comparisons are not on level grounds, a large percentage of the nations we are compared against do not have universally guarenteed secondary eduction. For instance both Japan and Germany have a system where only the top percentage of students will enter the college track eduction, these are the students that take the standardized tests, not the entirity of the population where in the U.S. every student who has not dropped out takes them.
Finally I would point out that the U.S. has largest percentage of the population in postsecondary education:
Per 100K population:
Korea 4,955
Japan 3,139
U.S. 5,398
U.K 3,126
France 3,617
Source
In fact the U.S. has nearly as many students in postsecondary education as the rest of the first world combined at over 15 million!
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
Hi Stan!
:)
I didn't know you had a 3 digit ID. lol nice
Just thought I'd say hi.
Loth
-=Lothsahn=-
The thing is if you look at the complexity of the circuit/number of transistors in a delta-sigma DAC, it is orders of magnitude more complex than the traditional current-stearing DAC.
The reason that the "digital" system, the delta sigma DAC, comes out ahead is the real strength of digital systems- noise immunity.
With the traditional DAC's, there are only 16 current sources/resistors in the whole chip, but each one needed to be trimmed by a laser to a very precise value. This costs a lot of money, which is why you need to spend 10x more on the chip to produce the same performance as a delta-sigma DAC. (most industrial DAC's and high-end audio still use this system)
With the delta-sigma DAC, you can put in several thousand transistors, but each one only needs to respond to two voltage levels. Each transistor can be poorly made, it just needs to be good enough to turn on when the voltage goes high and turn off when the voltage goes low. So digital chips require much less precision and can be made at much lower cost.
Basically, it is much cheaper to make tons of cheap transistors rather than one precise/accurate one.
That is the same reason binary is better than trinary- Trinary systems would probably require less transistors than a comparable binary system, but each one would need to be much more precise.
IMHO, funding isn't the problem with education(before college). Discipline is the problem. Teachers are tied up the whole day dealing with kids who should be kicked out of class but aren't because it would hurt their feelings. Smart kids aren't allowed to go to more advanced classes because then the less advanced kids feel left out.
Add to the fact that parents are on the kids side and not the teachers side. When a child fails a subject the parents first blame the teacher instead of themselves or the child. I have a few family memebers who are teachers and they work entirely too much trying to help every student learn, but if the parents are not involved it becomes nearly impossible.
No, the problem today is not lack of funding, but that America as a whole doesn't care about education anymore. Sure people pay a lot of lip service to helping the children and fixing the education system, but then no one wants to do anything about it. In order to fix the system the two main things that need to happen are 1)discipline needs to be restored and 2) parents need to become part of the solution.
It all depends. I managed to snag 12 college credits hours and many of my friends also received credits for college and AP classes in high school. There are intelligent kids who are open minded and when they feel that the regular classes are cake, they find ways challenge themselves further. Most colleges around this area allow concurrent admissions to high school students. One of my classmates took advantage of that to finish Calculus 3 (also counts as a math credit for high school as well) during his first half of his senior year. Sliding by is just pure laziness.
1f u c4n r34d th1s u r34lly n33d t0 g37 l41d
While 30 day paid vacations seem commonplace in a number of countries, I'd say 2 to 3 weeks is the rule of thumb in the US. There's also the dreaded "sick days" that force people to make lame excuses to use, or else they'll lose them.
They made you think they work too little, but in fact it is you (and I) that work too much.