Sun Turns to Lasers to Speed Up Computer Chips

alphadogg writes to mention that Sun is attempting to move from the typical design of multiple small chips back to a unified single-wafer design. "The company is announcing today a $44 million contract from the Pentagon to explore replacing the wires between computer chips with laser beams. The technology, part of a field of computer science known as silicon photonics, would eradicate the most daunting bottleneck facing today's supercomputer designers: moving information rapidly to solve problems that require hundreds or thousands of processors."

Comments

Great idea!

[peipas]

I assume these systems will be water-cooled so the miniaturized sharks have somewhere to swim.

Re:Great idea!

[aarku]

Yarrrrrrrrrrrrrrrrrr I'm sick of shark/laser jokes. No offense personally intended, just to the whole meme.

Re:Great idea!

[Anonymous Coward]

How do you feel about Sea Bass?

Re:Great idea!

[utopianfiat]

Depends, are they ill-tempered?

Re:Great idea!

[tbcpp]

LOL, if I had points I'd mod the parent up

Re:Great idea!

[jo42]

And "In Soviet Russia" blabs are getting uber lame. Halibut, even "Beowulf Cluster of ..." didn't hang around this long. The stench of olde and rotten is getting rather thick me smells.

Re:Great idea!

[monsted]

In Soviet Russia, beowulf clusters of halibut with friggin laser beams on their heads smell you.

This intention to reseach the use of lasers

[Timesprout]

has been hailed as a positive step by leading members of the shark community.

Are actuators faster than direct connections?

[BadAnalogyGuy]

I wonder if the time saved transmitting information via light is offset by the transition time used to translate that back into electric signals. On a single board, the distance travelled is on the order of decimeters. On a chip, micrometers. Are the time savings *that* significant? Even between peripherals, the time saved seems negligble.

Re:Are actuators faster than direct connections?

[TheRealMindChild]

I am not an expert in electricity by no means, but I have a fundamental understanding of it (or so I think). Energy is energy. With no resistance (don't overlook this point), light traveling via laser or via electrons flowing over a wire, the speed would be the same. Now, in reality, there IS resistance... there is always a "friction" or resistance (ohm) when energy is passing over a wire. In a vacuum, a laser will move as fast as energy can possibly travel. At least on paper.

Re:Are actuators faster than direct connections?

[isomeme]

Electrons in a superconductor (a material with zero resistance) do not travel at the speed of light.

Re:Are actuators faster than direct connections?

[ILuvRamen]

I think the photons strike a really small sort of solar panel where the burst of light turns instantly into a burst of electricity. So there's no digital translation by a chip necessary. Of course you lose a lot of power converting it like that cuz let's say the solar sensor is 50% energy efficient, well you have to use 2x the electricity in the first place to get the desired 1x electricity at the end. So these chips are gonna be fast but they'll suck up energy faster than me eating 50% Walgreens Cocoa Peeps the day after Easter :D

Re:Are actuators faster than direct connections?

[Gabest]

I guess both of you meant electromagnetic weave, electrons do not move too fast.

Re:Are actuators faster than direct connections?

[bartosek]

In fact electrons in your typical electrical wire don't move anywhere near the speed of light.

http://www.eskimo.com/~billb/miscon/speed.html

Re:Are actuators faster than direct connections?

[Anonymous Coward]

... but at which speed?

Re:Are actuators faster than direct connections?

[Anonymous Coward]

Photons are not electrons. Also, conductivity / electrical resistance is not really the same as friction, nor does friction necessarily reduce potential maximum speed, but transfers energy and reduces actual speed as a side effect.

Re:Are actuators faster than direct connections?

[Anonymous Coward]

Neither does the EM wave their motion represents. http://en.wikipedia.org/wiki/Speed_of_electricity

My question is, assuming wikipedia is right and that the EM wave goes 2c/3 through copper, and given that modern chips aready have switchbacks to ensure some wires are the same lengths as each other so the signals sync up, how much time will this 50% signal speed increase actually buy us? Will we have the same signal speed bottleneck (assuming that Sun is trying to solve a real problem, I'm not really a hardware guy so for all I know maybe 2c/3 is fast enough) with these chips in 10 years? If so, we should probably start thinking of workarounds, because something tells me we're not going to find anything that offers a 50% improvement over c.

Re:Are actuators faster than direct connections?

[JustinOpinion]

The article doesn't make it clear whether using optical communications is intended to reduce latency or increase bandwidth.

With respect to latency: the electrical signals travel at ~30% the speed of light, whereas the optical signals travel at ~70% the speed of light (it depends on refractive index, etc.). Over the distances we're talking about (as you said, mm to dm), that's only fractions of a nanoseconds delay savings. This is on the order of a modern computer's switching time. All this complexity to get rid of a one or two processor cycles of latency?

I suspect instead they are looking to increase bandwidth. An optical fiber can carry very high data rates. Moreover a single physical fiber can carry multiple simultaneous channels (e.g. different wavelengths of light). So the intention may instead be to create high-bandwidth links between various processors. Using on-chip lasers can make the entire assembly smaller and faster than the equivalent for electrical wires.

Really what they want, I think, is to implement the same kind of high-speed optical switching we use for transcontinental fiber-optics into a single computer or computer cluster. If you can put all the switching and multiplexing components directly onto the silicon chips, then you can have the best of both worlds: well-established silicon microchips that interface directly into well-understood high-speed optical switching systems.

Re:Are actuators faster than direct connections?

[arjay-tea]

It's not so much transit time, as parallelization where the big advantage is. Many frequencies of light can share the same medium without interfering with each other. Imagine many processors and memory chips streaming data to each other simultaneously, over the same backplane.

Re:Are actuators faster than direct connections?

[rbanffy]

I don't think it's about the time it takes to transfer a single bit but the amount of bits that can be transmitted at once with light rather than wires. If we can talk line-of-sight transmission between boards, it's easy to line up an array of about a million emitters with an array of a million detectors and send back and forth the same amount of data you would need a couple thousand wires (taking translation times into account) to do.

Sun is a very entertaining company to watch. Even when their gizmos never end up in products, they are always cool.

Re:Are actuators faster than direct connections?

[imgod2u]

No, but it depends on whether or not the receiver is current-steered or voltage-steered. If it's voltage steered then it's the propagation of the electric field that carries the signal. In which case, it can be near the speed of light.

Also, future chip-to-chip interconnects seem to be moving towards transmission lines rather than treating circuit paths like bulk interconnects. Wave-pipelining the signal will mean that data transfer rates will not be hindered by the time it takes a voltage swing from transmitter to reach the receiver. Latency is still a problem, however but I imagine the electro-optical conversion process already adds plenty of that.

Re:Are actuators faster than direct connections?

[ChrisA90278]

When you look at a wire, or printed trace on a PCB it is not the resistance that limits how fast you can send a signal. It is inductance and capasitance that act like a low pass filter. We don't care how fact eletrons travel in wire what we care about is how fact we can change the voltage in the wire. We send data by changing voltages not by sending electrons.

Re:Are actuators faster than direct connections?

[TheRaven64]

And when you look at a PCB, it's not just the speed of the signal that determines the time it takes, it's also the distance it travels. Wires on a PCB can only cross by being at different heights (expensive) so it is common to route signals indirectly, which increases their distance quite a lot. When you have 64 wires coming from your RAM chips, and needing to get to your CPU, this sort of thing adds up quickly. Beams of light, in contrast, can cross without interfering with each other.

Re:Are actuators faster than direct connections?

[mikael]

There are several major issues:

The first is the size of the packaging of the chip - the actual silicon might only occupy the space a quarter the size of the whole unit. All that extra space is just used to manage the 500+ copper connections between the silicon and the rest of the circuit board.

The second problem is that as the clock speed of these connections becomes faster, synchronisation becomes a problem. While CPU's are running in the GHz frequencies, the system bus is still running in the hundreds of MHz.

If the chip could connect to the circuit board through optical connections, then all this could be simplified. You would eliminate the need for all the copper connections while simultaneously speeding up the external clock speed.

Re:Are actuators faster than direct connections?

[warmflatsprite]

You're on the right track, but you're not quite there. Solar panels are more or less arrays of photodiodes. AFAIK most fiber system use PIN photodiodes to convert the light intensity over a specific band of wavelengths in a fiber to electrical current. Note that I said current, not voltage. Typically a transimpedance amplifier and some kind of comparator circuit is then used to measure the intensity of the signal. The PIN diodes can convert very small quantities of light to very small currents, and transimpedance amplifiers can deal with very small currents as well. Generally the limiting factor for low-light intensity systems like is the "dark current" of the diode you're using. If the current generated due to your light source is within the noise of the dark current you won't be able to detect any change in the system. Fiber systems operate at light intensities that generate currents well above this dark current, and they do so without a high power demand.

Power issues can be born from speed issues, though. Since photodiodes need a fairly large surface area to be able to generate enough current from light signals, the PN (or PIN) junctions act like a capacitor. Capacitors act like low-pass filters and this limits the switching frequency of the signal you can transmit. This effectively limits the data rate of the system. If you make the surface area smaller, you'll need to increase the intensity and focus of your light beam in order to make up for the change. This could cause high speed systems to have high bus power requirements and higher manufacturing costs.

Re:Are actuators faster than direct connections?

[Anonymous Coward]

I remember writing a report on this stuff in high school more than 10 years ago (a summary of the current tech - I don't mean I was doing this stuff!) At the time the concept was not that you'd generate light via an actuator but rather that light would emit from the chip itself by the process of exciting the silicate via a charge and then as electrons dropped to lower orbits photons were emitted.

The major problem was the purity of silicate required. Gallium arsenide was a better alternative but the cost was prohibitive - would this be an issue now?

Re:Are actuators faster than direct connections?

[cheater512]

Resistance isnt the problem. Its a few cm of copper.

The problem is inductance and cross talk causing interference.
One solution is to shield every wire in a bus but its not really practical. ;)

Re:Are actuators faster than direct connections?

[Anonymous Coward]

http://en.wikipedia.org/wiki/Velocity_of_propagation

70% of c for riser ethernet.

I think Sun is just banking on the "frickin' laser beams" factor.

Re:Are actuators faster than direct connections?

[recharged95]

Then again, I'd take the photoelectric effect over heat anyday.

Re:Are actuators faster than direct connections?

[Anonymous Coward]

The travel speed of electrical signals on conductors is not (directly) related with resistance. A wire can have high resistance and be fast, or have really low resistance and be really slow. What determines the propagation speed is 1/SQRT( inductance X capacitance) which relates to the relative sizes of the conductor wire and the distance to the ground plane.

Re:Are actuators faster than direct connections?

[Anonymous Coward]

This idea absolutely correct. It is all about bandwidth. If you have several chips on the same board and want to send data between them, you either use board traces, or you build a custom package, but either way you have to use metal and you hit a wall. Even if you cover the entire surface of your chip in solder bumps you will never get as much bandwidth as you would like.

Think about where the bottlenecks are in your computer... memory and IO. You want a faster supercomputer, well you need more processors and more memory, but you always have that communication overheard. But what if your memory had a direct optical link to your processor?

Re:Are actuators faster than direct connections?

[kestasjk]

The electricity-water analogy saves the day again: If you have a pipe filled with water you could push on one side of the pipe, and at the other end of the pipe a person would very quickly notice the increase in pressure, as water would start flowing out of his side. This doesn't mean that a molecule of water in the pipe would move far at all.

(It's actually to do with electric fields, which do travel at the speed of light, but the water analogy works well)

Re:Are actuators faster than direct connections?

[Sir Holo]

Cross-chip, it's latency. Timing signals take too long to cross the chip.

Inter-chip, it's probably a little of both. Bandwidth in some cases, and timing for complex circuits in other cases.

Re:Are actuators faster than direct connections?

[dwye]

> In a vacuum, a laser will move as fast as energy can possibly travel. At least on paper.

Of course, the light will be going through air, where the spped of light is only 99% of C

Anyway, the speed of conduction (i.e., signal propagation, as opposed to that of the actual electrons) in copper wire is about 1/3 C, and in a coax about 95% C.

For the distances involved, the difference in speed of signal propagation is not that important. OTOH, light gates are supposed to be capable of faster switching than silicon (even in ECL mode) by orders of magnitude (once they have a decade or so of engineering behind them). Maybe this is the justification for trying lasers?

Re:Are actuators faster than direct connections?

[real gumby]

Very good. The other major issues are interference (e.g. capacitive interaction between lines etc) and the sheer bandwidth -- you can modulate your carrier with plenty of other frequencies (referred to as wavelength in the optical domain and frequency in the audio/radio domain -- go figure).

Re:Are actuators faster than direct connections?

[Just Some Guy]

Wildly appropriate signature link.

Re:Are actuators faster than direct connections?

[meatmanek]

Since the article quotes a bandwidth, "billions of bits of data a second," rather than a latency, I think it's fairly obvious that Sun is attempting to increase bandwidth between sections of the processor.

Re:Are actuators faster than direct connections?

[rbanffy]

Indeed. Didn't realize that... :-)

Re:Are actuators faster than direct connections?

[owndao]

IBM beat Sun at the press release game. http://www.networkworld.com/community/node/25514 from 2-22-2008, has a good explanation of the benefits of optical. The main things that I see you gain are:
1. Tremendous increases in bandwidth (chip to chip, bus, and peripherals)
2. Less dependance on physical layout to handle high "clocking" rates (busses become like transmission lines at today's clock rates)
3. Independence from chip voltage requirements (you no longer have to have the electronics necessary to drive the bus at a different voltage level than what is optimal for chip low-power consumption)
4. Circuits are electrically isolated (fewer cascade failures of chips/boards/computers/etc due to power/lightning/static strikes. Easier to integrate different chip technologies like CMOS, TTL, etc.)
5. Ease of interfacing to pure-optical "special purpose" components (some operations are almost instantaneous in the frequency domain. A Fast Fourier Transform is a good example. A device as simple as a prism does the "calculations" almost instantly.)
6. No crosstalk and no stray EMF (for TEMPEST types out there)
7. Unaffected by nearby high voltages or currents
There are many more advantages that I, as a EE, love. I, personally, have been waiting for this transition to optical for over 30 years. Now if we can make that transition to optical memory (holographic perhaps), and eliminate the keyboard -- or better yet, all the mechanical parts... (sorry, all you MEs!) Photonics seems to be the future for awhile anyway. ;)

Commentary on this?

[Anonymous Coward]

Commentary on this, from an actual EE, not the pretend ones on Slashdot (you know who you are)?

Sounds sweet, but is it expensive in terms of energy/time/money? Does EMI become less of a problem on circuit boards? Will this make designer's lives easier?

Re:Commentary on this?

[ergo98]

Commentary on this, from an actual EE, not the pretend ones on Slashdot (you know who you are)?

Just look up any of the countless other "use light instead of wires" stories that have been widely reported over the past decade(s). I'm not saying it's not going to happen — I'm sure at some point it will — but barring additional information, preferably actual accomplishments, this is just more of the same.

Re:Commentary on this?

[ergo98]

To get you started, here's a search for you. It looks like IBM is only promising a 100-fold performance increase, but Sun got the contract (despite the possibly inaccurate story, it doesn't sound like they actually figured out anything thus far, besides "how to get some government loot") by promising a 1000x increase.

Hey DARPA — I'll give you a 1,000,000x improvement! Email and I'll tell you where to send the cash.

Re:Commentary on this?

[Timothy Brownawell]

Sounds sweet, but is it expensive in terms of energy/time/money? The article claims it will reduce energy usage. It's much faster, so it saves time. And because time is money, it also saves money. I'm going to make a wild guess that it'll be more expensive to manufacture, because wires and solder and very very easy to put down.

Does EMI become less of a problem on circuit boards? Yes, because you're no longer trying to send lots of high-frequency signals thru arrays of tiny antennas.

Will this make designer's lives easier? That would probably depend on what they're designing.

Re:Commentary on this?

[sexconker]

Technically possible, financially infeasible.
What else is new?

It's been so long since SUN was relevant, and this story changes nothing.

-1 : redundant

[UdoKeir]

To quote Scott McNealy:

You know, I have one simple request. And that is to have SPARCS with frickin' laser beams attached to their heads!

Re:-1 : redundant

[ndevice]

Watch out for sun to buy out, or merge with Analog devices soon. If they get their lasers going, they could put them on those Analog Devices DSP parts too.

Re:-1 : redundant

[ioshhdflwuegfh]

and some lazer beans, real loud...

Re:-1 : redundant

[Anonymous Coward]

now that was funny. my mod points ran out yesterday or I'd help you out.

Why not...

[weaponx86]

If the "lasers" require an electrical signal to be generated, isn't this just adding a step? Also you need an optical sensor somewhere which converts the light back into an electrical signal, no? Sounds like building a tunnel where there is already a bridge.

Re:Why not...

[sdpuppy]

In that case the light could be used :

to connect parts in the chip that are furthest away

or

some of the computing / logic is performed in the light domain before it is translated back to electron domain.

Re:Why not...

[sdpuppy]

Also light behaves in a non-linear fashion which opens the possibilities for speeding up certain types of calculations (logs etc)

Re:Why not...

[Anonymous Coward]

Or building a light-bridge where there is already a tube

Re:Why not...

[JustinOpinion]

To use the beloved transportation analogy: it's like moving your cargo off of trucks and onto a high-speed train. Yes it takes time to move cargo, but it's worth it if the time savings of the high-speed train are big enough (for long enough distances, the savings can be significant).

In this case, there may be a delay associated with signal processing, but if the optical transmission is sufficiently faster than an equivalent electrical one, then it's worth it. Considering that electrical signals themselves need to undergo various kinds of switching and processing anyway (data written or read from a bus), I don't know that converting to laser signals will add much of a delay.

Re:Why not...

[Arthur B.]

So do transistors. What's your point ? Analog computation ? Yurk.

New warning stickers...

[fahrbot-bot]

From TFA: Each chip would be able to communicate directly with every other chip via a beam of laser that could carry billions of bits of data a second.

Do not look at chip with remaining good eye.

Don't Shake the computer!

[CubeRootOf]

I wonder what will happen to thier investments if someone shakes the table, or knocks the computer on its side, or even if there is an earthquake.

What happens when the computer gets dusty, or mold starts to grow on one of the lenses?

how will dust be solved? Water? Bugs (of the insect variety)?

Re:Don't Shake the computer!

[Belial6]

Don't worry, someone will ask it a question that is a paradox before then, and the whole thing will destroy itself with sparks and slowed audio.

Re:Don't Shake the computer!

[Kadin2048]

I don't know if this is a serious question or not, but one assumes that the lasers will operate in completely sealed environments (e.g. inside an IC package) or over optical fibers if they need to traverse free space. I think the intra-package situation is probably more common; you could communicate from one core to another on the same die using a laser rather than a wired interconnect and hopefully have less interference/RF/capacitance issues to deal with. This also makes sense given what I know about modern types of laser diodes (especially Vertical Cavity ones) -- they can be created on silicon wafers through similar processes to the way transistors are laid down.

I can't think of any good reason why you'd just be aiming a laser through the empty space inside a PC's case.

Re:Don't Shake the computer!

[vertinox]

how will dust be solved?

Why don't you crack open your 3.5" hard disk drive and find out why dust doesn't bother those sensitive platters? ;)

Re:Don't Shake the computer!

[WhoBeDaPlaya]

Free space would be quite a pain. You'd need to collimate the beam, worry about acceptance angles, mode field diameters, etc.

Re:Don't Shake the computer!

[Anonymous Coward]

Why don't you crack open your 3.5" hard disk drive and find out why dust doesn't bother those sensitive platters? ;)

You jest, but I had a hard drive that was giving me problems just last week. Turns out a thick layer of white dust had covered the front of the drive, due to sitting right in front of a cooling fan. Wiping it off made the drive work problem-free again.

My current working theory: You know that little hole on drives that they tell you not to cover? That's to allow air to enter the drive so the pressure can be stabilized (because the heads ride on an air cushion). Obviously, they need to filter the air to keep dust from getting inside the drive. I suspect that filter got clogged, and blowing away the dust caked on the drive cleared that away.

Or the drive could have just been overheating. But temperature monitoring didn't show up anything unusual.

What impresses me most is that the drive was able to shut itself down when it detected a problem. I didn't suffer any data loss, except due the drive abruptly disabling itself after a random number of minutes. (Of course, it could have degraded a little more gracefully.)

One weakness... toothpaste

[Anonymous Coward]

Just don't let anyone with a mastery of the "JUDO CHOP!" bring toothpaste with them... or else you'll need a big boy to get you out of that situation.

What kind of laser beams? Will they terminate?

[LM741N]

Will these be in the visible or infrared range? Will the laser beams terminate or leak outside the unpackaged chip? I ask because engineers are constantly looking at decapped chips or doing various types of testing under the microscope of live circuitry. I'd hate to get hit by a laser beam through a microscope.

Chips with frikkin lasers!

[Mi1ez]

Sorry, hadda be said. :)

RE: Sun Turns to Lasers to Speed Up Computer Chips

[Anonymous Coward]

I didn't read TFA, but I did read the headline...
So you are telling me that Star at the center of our solar system (Sol or some people call it "Sun") is somehow changing its rate of rotation/turning to track lasers and the side effect of this turning is to increase the production speed of inedible chips made out of computers?
No wonder, I don't read TFA... the headline is just plain silly.

whoa, scared me there for a second.

[Jafafa Hots]

Thought it was saying the sun turned into a laser. That could be a bad thing, different kind of light and all that.

Hmm, missed this opportunity too

[Private.Tucker]

Back in High School, a friend named Tom had this crazy idea to use light to transmit information between components in a computer. Back in the mid-to-late 90's, I wonder if anyone else thought of this.

Re:Hmm, missed this opportunity too

[ZeroExistenZ]

Yes, they use called (silicon nano-) photonic chips for that purpose. The same technique will be used to communicated between different cores in chips. (check the press release of IBM)

I've been looking for an explinatory video from IBM I believe, explaining laser-computing and how they solved certain problems in their designs, but I've failed to locate the particular movie.

A really high bridge

[Pinky's Brain]

On chip they are pumping the signal over a traces with mm range lengths and um range widths, off chip it's over traces with dm range lengths and mm range widths. Timing and power consumption are hard enough problems on chip, off chip they become much harder ... not to mention that most of the power consumed either goes into EM or gets coupled into other signals.

Serial connections help with the timing, but do diddly for power and noise. That's where optical comes in.

Re: Sun Turns to Lasers to Speed Up Computer Chips

[Broken scope]

You twat. Stop trying to be a pedantic prick. It says "Sun", The shortened name of a company called Sun Microsystems thats typically used in conversation by a large number of people who don't have shit for brains. Lets not forget the logo displayed to the side of the article summary.

You might not be such a dumb fuck if the title said "The sun".

light bridges vs. tubes... hmmm...

[norminator]

Would a series of light bridges allow me to send my internets faster than a series of tubes?

Re: Sun Turns to Lasers to Speed Up Computer Chips

[shentino]

Pot. Kettle. Black.

I agree with the mods on GP (for once). It was an attempt at humor and was properly labelled as such.

Why light, why not wireless?

[ThreeGigs]

Remember the article not long ago about micro transmitters/receivers on a chip?

Considering no special connections are needed for wireless, unlike light which woud likely need fiber or line of sight, chips equipped with that mini wireless tech would, in theory, only need to be powered and placed in proximity to each other.

Not as sexy as SPARCs with friggin' lasers, but certainly a plus from a computer design perspective.

Re:Why light, why not wireless?

[imgod2u]

Even a directed wireless transmitter through a waveguide only manages to send a fraction of its signal power over to the receiver. There's also the problem that it's much more susceptible to interference, it drains a lot of power because RF signals are not easy to generate at high speeds, the extra logic required and the fact that the bandwidth is just nowhere near what traditional wired links are capable of might not make it all that attractive.

Re:Why light, why not wireless?

[morgan_greywolf]

Even a directed wireless transmitter through a waveguide only manages to send a fraction of its signal power over to the receiver. There's also the problem that it's much more susceptible to interference, it drains a lot of power because RF signals are not easy to generate at high speeds, the extra logic required and the fact that the bandwidth is just nowhere near what traditional wired links are capable of might not make it all that attractive. Exactly. Hence the reason 802.x wireless is much slower than its wired counterpart or why fiber optics are used for high-speed networking over great distances (like between North America and Europe) (as opposed to satellites).

Re:Why light, why not wireless?

[ThreeGigs]

Agreed on all your points, although I don't think getting 100% of the sigal power to the receiver is an issue. And maybe wired bandwidth is greater, but if you only need 1 gigabit, who cares if fiber can do terabit speeds?

Sun's research is aimed at supercomputers... getting 1024 processors to all talk to each other. Simultaneously. That's a lot of cross connections, and some heavy duty switching gear. But as long as any two processors can switch to the same frequency, they could communicate. Meaning 512 processors in a 1024 node cluster could be sending, at full speed, to the other 512, simultaneously. At 10 GBit per, that's 5 Terabit speed, in total. With the only limiting factor being the amount of channels available.

Siemens had 1 Gbit wireless in 2004. University of Essex had 10 Gbit wireless in 2005. And remember... it's not necessarily the speed that really matters, but the ability to easily mesh together hundreds or thousands of processors.

Me too

[davidwr]

You weren't the only one to confuse Sol with the computer company.

My first thought was "The sun is lasing? Cool!"

My second thought was "space sharks! Way Cool!!!"

Whenever anyone says 50%

[florescent_beige]

"This is a high-risk program," said Ron Ho, a researcher at Sun Laboratories who is one of the leaders of the effort. "We expect a 50 percent chance of failure, but if we win we can have as much as a thousand times increase in performance."

Whenever anyone says there is a 50% chance of something happening they really mean "I have no idea. No idea at all. I'm guessing."

In probability theory, "p" has a specific meaning which is roughly stated as "the ratio of the total number of positive outcomes to the total number of possible outcomes in a population". So for the number of 50% to be right, it must be known that if this research was repeated a million times, 500,000 times there would be success and 500,000 times there would be failure. But this makes no sense because the thing being measured is not a stochastic property. It is simply an unknown thing.

What is probably vaguely intended when a number like this is given is that if you took all the things in the history of the world that "felt" like this in the beginning, half of them will have worked out and half will have not.

How on earth could any mortal human know that?

But it gets even more complicated. One cannot state a probability like this without stating how confident one is in the estimate of the number. So really a person should say the probably of success of this endeavor is between 45% and 55% and this estimate will be correct 19 times out of 20.

With that as background here is what I humbly suggest 50% really means: it means "I have no idea how to quantify the error of this estimate. It doesn't matter what the estimate is because the error band could possibly stretch between 0% and 100%. So I'll split the difference and call it 50%". But that is wrong, the statement should be "I estimate the probability of success to be between 0% and 100%".

But nobody does that because it makes them look stupid.

So whenever anyone says there is a 50% chance, or a 50/50 probability of something happening, they might as well talk in made-up Klingon words, the information content of their statement will be equivalent.

Re:Whenever anyone says 50%

[Anonymous Coward]

as long as you set p=1.0, you can't be wrong.

(I do it with my girlfriend all the time.)

Re:Whenever anyone says 50%

[NeutronCowboy]

Absolutely. Personally, I do the same thing: if someone asks me about the likelihood of something happening about which I have no clue, I tell them flat out "50/50. Here, let me flip a coin." I expect the same thing to have happened here as well.

Now, someone please mod me redundant. Executive summaries should be discouraged wherever possible.

Re:Whenever anyone says 50%

[QuantumFTL]

In probability theory, "p" has a specific meaning which is roughly stated as "the ratio of the total number of positive outcomes to the total number of possible outcomes in a population". So for the number of 50% to be right, it must be known that if this research was repeated a million times, 500,000 times there would be success and 500,000 times there would be failure. But this makes no sense because the thing being measured is not a stochastic property. It is simply an unknown thing.

This is true, if by "probability theory" you mean "Frequentism". Frequentism is nice, for those cases where you are dealing with nice, neat ensembles. For a lot of real world situations which require probabilistic reasoning, there are no ensembles, only unique events which require prediction. For that, we often use Bayesian Probability.

Take the assertion "I'd say there's a 10% chance that there was once life on Mars." Well, from a Frequentist point of view, that's complete bullshit. Either we will find evidence of life, or we won't - either the probability is 100% or 0%. There's only one Mars out there.

In order to deal with this limitation, Bayesian Probability Theory was born. In it probabilities reflect degrees of belief, rather than frequencies of occurance. Despite meaning something quite different, Bayesian probabilities still obey the laws of probability (they sum/integrate to one, etc), thus making them mathematically compatible (and thus leading to confusion by those that don't study probability theory carefully.) Of course there are issues with paradoxes and the fact that prior distributions must be assumed rather than empirically gathered, but that does not prevent it from being very useful for spam filtering, machine vision and adaptive software.

As someone who professionally uses statistics to model the future performance of a very large number of high-budget projects at a major U.S. defense contractor, I can assure you that his statement was much more in line with the Bayesian interpretation of probability than the Frequentist view you implicitly assume.

Sorry for the rant, I just get very annoyed when people assume that Frequentism is all there is to statistics - Frequentism is just the beginning.

But it gets even more complicated. One cannot state a probability like this without stating how confident one is in the estimate of the number.

Of course! But where did the confidence interval come from, and how much confidence do we have in it? It's important to provide a meta-confidence score, so that we know how much to trust it! That too, however, should be suspect - indeed even moreso because it is a more complex quantity to measure! So a meta-2 confidence score is in order, for any serious statistician... But why stop there?!

With that as background here is what I humbly suggest 50% really means: it means "I have no idea how to quantify the error of this estimate. It doesn't matter what the estimate is because the error band could possibly stretch between 0% and 100%. So I'll split the difference and call it 50%".

So, if someone does not give an error bound on an estimate, we should assume that the error is maximal?

So whenever anyone says there is a 50% chance, or a 50/50 probability of something happening, they might as well talk in made-up Klingon words, the information content of their statement will be equivalent.

Or, it's entirely possible that that 50% number is somewhat accurate, because they know something about the subject that you do not.

Re:Whenever anyone says 50%

[Xmastrspy]

110% That is the other one that I think needs to be looked at. What does it mean to give 110%. Is that even possible? How do you get over 100%? If you say 110%, doesn't that mean the the scale has just gone from 0 - 100 to 0 - 1000... So 110 is on he very low end of the scale? Thats right boss, I am giving 110%!!!

Re:Whenever anyone says 50%

[florescent_beige]

Just my luck huh, here I go looking all smart then some uber Bayesian has to come along and spoil my party.

Anyway, with little expectation of anything good coming from this (for my ego I mean), here's why I don't usually think in Bayesian terms. Correct me if I'm wrong which I probably am.

While I have heard Bayesians talk about probability not meaning the same thing as as "normal", I've never seen any Bayes p which means anything other than a relative likelihood that I'm familiar with. If there is a bag with 3 red balls and 2 white balls in it, the probability of randomly drawing a red ball out is 3/5 even to a Bayesian, right?

I believe, and here is where I could be all wrong, that as Bayesians we should interpret the 50% number from the OP as an a-priori estimate which is to be refined if we ever get better information. But doesn't that have the same problem that I talked about, which is that the thing under consideration is not a stochastic variable?

And even if it is valid to do that, simply elevating the 50% to the status of an a-priori estimate doesn't suddenly make it a more accurate or even legitimate number. I mean, does it?

As for the error estimate which ended up being the crux of my previous argument, well, referring back to the wiki article on Bayes that you linked to, interestingly even they give the example here of a case where we have no prior knowledge of how many different colored balls are in the bag in which case we would use a uniform a-priori distribution which is exactly what I described originally, it could be anywhere between 0% and 100%, we don't know.

Interesting stuff.

Re:Whenever anyone says 50%

[grammar fascist]

My mod points expired recently, so could someone mod this up? I do machine learning and computer vision with Bayesian statistics, and the above poster is spot-on. The GP sounds like a frequentist trying to regain control over statistical vocabulary.

FWIW, the frequentists can keep "confidence interval". We don't want to sully our theoretically sound vocabulary with its filthy connotations. :p But "probability" is something we'll lay uncompromising claim to, however much detractors say that subjective probabilities don't count. If they don't count, how else would anyone model something like "belief" in a well-grounded way?

Re:Whenever anyone says 50%

[Anonymous Coward]

you're not worth taking the time to correct.

Re:Whenever anyone says 50%

[Anonymous Coward]

"With that as background here is what I humbly suggest 50% really means: it means "I have no idea how to quantify the error of this estimate. It doesn't matter what the estimate is because the error band could possibly stretch between 0% and 100%. So I'll split the difference and call it 50%".

I knew what he meant. I think we all did! ; -)

Re:Whenever anyone says 50%

[QuantumFTL]

Just my luck huh, here I go looking all smart then some uber Bayesian has to come along and spoil my party.

I'm hardly a Bayesian in spirit, but it's useful enough when treated properly. I'm actually much more likely to say "Bayesian statistics is absolute bollocks - which just so happens to work very reliably in many cases". This is due to the well known paradoxes with priors, and issues associated with the certainty of beliefs (which you referenced). I prefer Dempster Shafer evidence combination when I can use that, for that reason, but still it's a lot of inductive reasoning with a provably sketchy base case.

Anyway, with little expectation of anything good coming from this (for my ego I mean), here's why I don't usually think in Bayesian terms. Correct me if I'm wrong which I probably am.

While I have heard Bayesians talk about probability not meaning the same thing as as "normal", I've never seen any Bayes p which means anything other than a relative likelihood that I'm familiar with. If there is a bag with 3 red balls and 2 white balls in it, the probability of randomly drawing a red ball out is 3/5 even to a Bayesian, right?

Interestingly enough, I think that Bayesian probability more accurately reflects what nonmathemeticians often mean when they give statistics about predictions - indeed due to its connection with betting, you can see why it would be intuitively favored by the common folk. Of course I doubt they do explicit probability chaining, however I think bayesian inference is intuitively used for many statistical deductions made by the untrained... it just "feels" right for many real life situations - situations that do not involve ensembles.

As per your question, well, you left a lot there to be assumed. Assuming there's no other factors involved (the selection is purely random) then yes, a Bayesian would say "I have a 60% belief that the ball I pull out will be Red." So it is compatible in that sense (one of the many reasons it's useful at all, and why one can even call the Bayesian notion of belief a "probability".

I believe, and here is where I could be all wrong, that as Bayesians we should interpret the 50% number from the OP as an a-priori estimate which is to be refined if we ever get better information. But doesn't that have the same problem that I talked about, which is that the thing under consideration is not a stochastic variable?

I'd say that's actually inaccurate. If he was being a strict Bayesian (which I doubt), there would be an a-priori estimate about the difficulty of the challenge being faced. The actual belief (his stated 50%) would actually be a chaining of that a-priori estimate with all other information this individual happens to have about their efforts. Given that Ron Ho is one of the scientists leading the effort, I would tend to believe that he actually has quite a bit of information, indeed possibly enough to dwarf the influence of the prior. This is one of the features of Bayesian Probability that prevents it from being useless, the fact that in many real life situations, even a crappy prior will lead to estimates that agree well with experimental results.

And even if it is valid to do that, simply elevating the 50% to the status of an a-priori estimate doesn't suddenly make it a more accurate or even legitimate number. I mean, does it?

If someone makes an assertion, I generally try to estimate how likely they are to be accurate given what I know about what they've said. (Interestingly enough, I wrote a slashdot comment about this being a logical basis for the validity of ad hominem attacks - you may find this interesting). If this were a random slashdotter (such as myself) saying the 50% number, I'd say "bollocks" (which I inexplicably say despite being a quaint colonial), but this is an expert in th

Re:Whenever anyone says 50%

[florescent_beige]

This is so fascinating. I had no idea there could possibly be two-way advocacy about this. If I had of known I would have worn my asbestos underwear :)

Honest question, Bayesian-wise, how could/would one interpret the 50% number in the article? Is there an intuitive interpretation? Is it quantitative or qualitative?

Re:Whenever anyone says 50%

[QuantumFTL]

I think there are a lot of people who are not really taught Bayesian statistics, and so they are limited to think of probability solely in terms of frequentist terminology.

To be fair, many things about Bayesian statistics are odd, and possibly even unsound (yay prior distributions we just made up!) The confidence interval thing can get a bit ridiculous, but I prefer Dempster Shafer theory for the precise reason that I emphatically DO NOT want to treat all evidence with equal weight.

Interestingly enough, I did some research in school with that last bit, using Bayesian combination for ensemble classifiers, and varying the coefficient which determined how strong the exponential bayesian weighting was (how much more strongly slightly better classifiers were preferred to their peers in terms of voting weight). Turns out it affects the ensemble quite a bit, but not in an obviously predictable (read - useful) manner.

Not about single wafer design

[renoX]

If I understood correctly this is not about single wafer design but exactly the opposite: regaining the speed of 'single wafer design' with multiple chips by using optical communications between chips increasing the inter-chips bandwidth (normally intra-chip bandwith is much higher than inter-chip bandwith so this is a bottleneck).

Re:Not about single wafer design

[renoX]

A better link providing more information:
http://www.theinquirer.net/gb/inquirer/news/2008/03/24/sun-silicon-photonics-macro

+1, Superior Use of Meme

[Anonymous Coward]

n/t

link to the original story (!)

[spage]

Why, why, why do people submit second-hand links to Slashdot?

The byline of the Seattle Times story is "John Markoff New York Times". 5 seconds with Google's site:nytimes.com reveals the original story with better explanation and more quotes from Sun personnel.

Re:link to the original story (!)

[dwye]

> Why, why, why do people submit second-hand links to Slashdot?

Because the NY Times used to require registration to read their articles?

.

Of course, the article still makes some bonehead errors. They do not cut wafers of identical chips apart to be able to eliminate the few failures in a circuit, but because we want a hundred CPU chips more than we want a single four inch processor with about 100x4 or x8 cores. You do not need that many processors to do your own taxes (unless .Net is far more wasteful than I have heard) or run a word processor. You might get a kick-ass game of WoW out of it, I suppose :-) -- providing that your net connection and database update is not the bottleneck.

That estimate of 100 chips per wafer may be low, BTW, as it has been well over a decade since I last checked into semiconductor manufacture, wafer sizes, and yields.

Macrochip

[imgod2u]

Interesting, so what they want to do is to be able to create larger multi-chip packages where each the chips are connected to each other optically rather than the traditional wire-bonds on a SiP. I'm honestly not seeing the advantage here in terms of speed. A single LVDS pair across a chip pad and wire-bond can already carry "tens of billions of bits per second" of bandwidth. Many can be put in parallel. I can see this being an advantage if they've discovered some ultra-efficient electro-optical conversion device that's can be etched into silicon. LVDS drivers and receivers do suck up a lot of power....

Misleading title/resume

[Anonymous Coward]

I think it should read as (new text in italics):
"The company is announcing today a $44 million contract from the Pentagon to explore replacing the wires between computer chips with laser beams. The technology, part of a field of computer science known as silicon photonics, would [I]render useless[/I] the most daunting fear of the Pentagon: [I]EMP weapons[/I]."

Re:Misleading title/resume

[morgan_greywolf]

The technology...would [I]render useless[/I] the most daunting fear of the Pentagon: [I]EMP weapons[/I]." And that's exactly why the Pentagon would be investing in such technology. Any additional performance or other geeky coolness is just a side benefit. Ultra-high-performance computing is the DoE's gig, not the Pentagon's.

Re:Misleading title/resume

[Anonymous Coward]

My fear would be that they succeed, tell everyone that they failed, then launch their own EMP weapons.

Intel

[pak9rabid]

Intel's already been working on this for a few years. For Sun's sake, they better hope that Intel didn't file for a patent on this already, otherwise this could get messy.

Just make wafer-sized chips!

[robi5]

Instead of goofing around with connections, why not build a chip occupying the entire 300mm wafer? Any local manufacturing problem would disable just one specific core out of the hundreds of cores on the wafer-chip. Isn't it done already? Cell, AMD tri-core, old celerons... Even the memory could be on the wafer, or at worst, one wafer for the cores and one for the memory, vertically stacked with through-silicon vias.

Re:Just make wafer-sized chips!

[IdeaMan]

Cooling?
Oh and how long are those vias? Will you be trying to get heat to flow through the memory wafer?

Re:Just make wafer-sized chips!

[robi5]

This needs one large cooler instead of hundreds of smaller ones. You can do something useful with the concentrated heat, for example provide hot water. Better than letting it go useless. But I think a good tradeoff would be to lower the frequency an order of magnitude, and use the massive parallelism - hundreds or thousands of cores on a die. Better, make it fully three-dimensional for a massive explosion of processing units. The brain is large and is 3D and still does not get really hot.

Re:Just make wafer-sized chips!

[totoanihilation]

Because we don't have the fabrication technology to expose a whole wafer at once. Since we're essentially shining light on the surface, the wider we make the beam, the softer the features get, specially towards the edges (because the light hits the edge at a different angle).
There's a sweet spot of size-vs-yields. Trying to make bigger chips requires multiple exposures for the same die, and getting the exposures to line up properly is extremely tedious.

That's why it's easier to make lots of small chips on a wafer than several big ones, and why people are trying to find better ways to communicate between chips...

Re:Just make wafer-sized chips!

[robi5]

Maybe several hundred connection lines between any two "cores" or memory banks would do, in which casse the edge traces to align with can be orders of magnitude wider, making the alignment possible.

Re: Sun Turns to Lasers to Speed Up Computer Chips

[droopycom]

If it said, "The Sun" I would have been worried about the british tabloid... :)

power and heat

[Crazyswedishguy]

I'm not very well versed in chip design, as I only took one class a few years ago. Could someone please confirm or disprove the following hypothesis?

Assumption: The energy dissipated in a chip generates heat, which could be avoided by the use of lasers, resulting in lower heat generation and energy consumption.

I'm fully aware that my speculative hypothesis may be completely unfounded, especially given that not much heat should be dissipated when electricity flows through a superconductor. If someone who is more informed (i.e. physicist or chip designer) could answer my question, I would appreciate it.

And yes, the lowered energy consumption would be offset by the energy spent in feeding the sharks.

Re:power and heat

[KiwiCanuck]

Yes, that's true, but that's not the focus of the article. The article is aboot replacing electrical lines on the PCBs. The biggest bottleneck in a PC is the front side bus. This is the connection b/w the memory, the HDDs, and the CPU. If you could switch these types of connections from electrical to optical then you could increase the communication bottleneck b/w the chips. The next step would be faster RAM and then faster HDDs, next a faster CPU, then a faster bus, and the circle continues.

Re:And furthermore . . .

[seededfury]

I turned mine off the first day i saw it and haven't seen it since.


Preferences--->Discussions---->Viewing.

Not so hard.

Re:And furthermore . . .

[Xiph1980]

Haven't received enough hugs from your daddy??

Re:And furthermore . . .

[cbiltcliffe]

Weird. The new discussion display system works just fine on my computer, and in fact, on any computer that I've ever tested it on. Running either Firefox, or (shudder) IE.

I keep hearing people bitch about how it takes forever to load, and crashes their browser, and all sorts of other crap, but I've never seen it, and I've surfed /. on a very broad range of computer hardware.

Maybe your computer is infected with spyware, or something. Or maybe you've got a browser extension that screws something up.

It can't be the discussion system itself, as it works fine for a great number of people.

Computer science?

[digitally404]

The area of photonics is largely related to physics and electrical engineering, not so much with computer science, which deals with information processing and computations. Being someone who works in the area of silicon photonics, this is some pretty exciting news.

Haven't we been here before

[saccade.com]

It was quite the smoking crater last time around. Maybe technology has improved since then...

Re:Great idea! Cypress has been there, done that.

[aisnota]

Cypress Semiconductor has already figured this out with their tech. Check out Silicon Light Machines and you will see T. J. Rodgers acquired a former Cypress Semiconductor alumnus in that acquisition and all Sun needs to do is work with CY.

Is Sun running out of business ideas?

[TruthfulLiar]

I have to wonder, if Sun is pursuing Defense contracts, does Sun know where it's business is headed? Usually companies do the Defense contracts when they are small, need money, and don't really have a product yet. Since Sun made $740 million last year, you'd think they could afford to spend $40 million on this (probably over several years), and then they'd get to keep all the knowledge to themselves (including their R&D direction). So I can only assume that either Sun thinks this has too small a chance of success to invest in, or they can't think of any ideas for the future and are using government money to explore lots of ideas and hope that one of them keeps the company afloat.

Maybe it's just because I'm not in the server space, but it's unclear to me why exactly I would buy a Sun machine. I used to know--they were fast and had a nice version of Unix--but now Solaris is free and I'm not even sure if Sun makes their own chips any more.

Re:Is Sun running out of business ideas?

[dwye]

>blockquote>Usually companies do the Defense contracts when they are small, need money, and don't really have a product yet.

Like AT&T, the entire aircraft industry, IBM, etc.?

This is NOT just an SBIR grant.

Maybe it's just because I'm not in the server space, but it's unclear to me why exactly I would buy a Sun machine.

Yes, I would not recommend them for cheap laptops, or to give to your grandmother to handle her email needs. That would be a bit of overkill. OTOH, if you have a problem where a 48 processor machine and a few TB of disk running a few months at a time is a good start, you might start considering it.

Optical vs Magnetic

[dunc78]

I agree that dust will not be a problem, as the pathways through which the light signal would travel would probably be sealed in some way and I can't even begin to guess why the GP was concerned about a computer being knocked on its side. However, I would imagine that since the pickups for a hard drive are magnetic, dust would not make much of a difference. Now I don't know how big the gap is between a head and the platter, so I guess if this was close enough dust could scratch a platter? But our CD drives tend to work fairly reliably over an unsealed path.

Re:Optical vs Magnetic

[Nigel Stepp]

The head is close enough to the platter that it would hit a piece of dust. In fact, the head is *so* close to the platter, that it would hit a fingerprint on the surface. It floats on a cushion of air created by the high speed of the spinning platter.

Scratching the surface renders that part of the surface unusable, but also creates pieces of shrapnel which cause more problems.

I think it's absolutely incredible that hard drives work at all.

bandwidth, and I think the article missed it

[reiisi]

Not to make bigger chips, but to solve the interconnect problem when you use a lot of small chips in a big package.

Although, even on-chip, at 1 cm^2 and above, optical conversion might beat be able to beat the reactance+buffering on a channel that crosses the whole chip, especially when a single physical channel might be able to carry 64 logical channels.

It's not a new idea, it's just one that needs to be revisited from time to time, to see if the optical tech is up to the job yet.

price reduction?

[glitch23]

"The company is announcing today a $44 million contract from the Pentagon to explore replacing the wires between computer chips with laser beams.

I hope this means that servers with the new chips will not actually cost 2-4x as much as an equivalent Dell server. IMHO, Sun needs to do something about the cost of their servers. I try to only use them when required because of their cost and I'm told the inflated price is due to the low yields of the SPARCs.

Re:price reduction?

[Anonymous Coward]

Have you actually looked at their product line anytime in the last 2 years or so? They are pretty competitive with their servers. Take a look at the x4150 on their site as an example. 2 quadcore procs, 24G mem and about 1TB of disk for around 10 grand and in a 1u server.

Re:And furthermore . . .

[andersbergh]

I don't like the commenting system, it's quite slow and very annoying to use. Usually I don't log in, but I guess this is a good incentive :-) I do hope that they will enhance it, perhaps even provide an option to disable it without logging in.

Re:And furthermore . . .

[Anonymous Coward]

AC's don't get to change preferences, as explained here. And there is nothing else in that list that is enticing enough for me to get a UID.

Re:And furthermore . . .

[Anonymous Coward]

I don't like the commenting system, it's quite slow and very annoying to use. Usually I don't log in, but I guess this is a good incentive :-) I do hope that they will enhance it, perhaps even provide an option to disable it without logging in.

See Taco?? It seems that every time I post a rant like this, someone agrees with me. I wish more slashdotters to rant about it.

Yes, slow and annoying. There was nothing wrong with the old system. No one asked us if we wanted it changed. I can see what Taco is going for, and it looks like a good idea. It's just been poorly executed. It would have helped immensely if someone tested the fucking system before it was inflicted upon us. And I fail to see what is so difficult about providing an option to disable it. Maybe Taco doesn't realize that I can troll just as much through a UID as I can posting as AC.

Re:And furthermore . . .

[Anonymous Coward]

Then stop bitching...