Slashdot Mirror
New Optical Chip Claims 8 Trillion Operations/sec.
Richard Finney writes "Lenslet is announcing
the 'World's First Commercial Optical Processor.'. Reuters has the story here. The Inquirer has a cool graphic here on it. The processor is specified to run at a speed of 8 Tera (8,000 Giga) operations per second, one thousand times faster than any known DSP. When Lenslet releases its Enlight processor in a matter of weeks, a unit using the technology will be 1.7 centimetres high and measure 15 by 15 centimetres."
It can't handle 8192 Giga Operations per second?
There's a nice picture of the processor here
Interstingly, optical processors aren't faster because light is faster than electricity. They are faster because they have much faster rise and fall times between digital on and digital off.
Slashdot Syndrome: the sudden, extreme urge to correct someone in order to validate one's self.
This innovative new product will enable revolutionary, new applications in the fields of defense, homeland security, multimedia and communications. The exhibition being held at The World Trade Center, continues until October 15th, 2003.
The fact that...
1. its at the WTC
2. they mention defense and homeland security
3. its immensely powerful
...makes me question whether or not this is going to be available to end users.
besides the lack of a huge marketing campaign.
Anyone know anything different?
Get paid to code OSS
Optical processors have incredible potential. And if you think that's good, just wait. The combo of an optical processor with optical memory is a one-two punch.
But if you want to get the full speed out of your processor and memory, as I recall, all the buses must be optical as well.
Otherwise you're limited by silicon and PCB boards again...
"Processing at the speed of light, you can have safer airports"
Its really quite sick and disturbing that the aftermath of 9/11 has degraded to a marketing ploy.
How do I keep track of people who are fingering
Nope, you needn't worry. You'll still finish well before the computer does.
Prismatic
Optical
Refractive
Nacell
---- The Open Source Record Label : : LOCARECORDS.COM
I don't care if it fills my room, and I have to sleep on top of it in the 30cm space between it and the ceiling. I'd still take it.
My patent states:It's all there in black and white.
I appreciate that it's a great demonstration of new technology, but maybe it's a little premature to call this a new commercial chip. It sounds to me like a demonstration of a research project or an exposition of things to come.
It's quite possible that I'm completely ignorant about this, but to whom do they expect to sell the latest and greatest THREE ORDERS OF MAGNITUDE increase in memory bottleneck?
This is NOT a Harvard architecture part - this isn't fetching instructions from RAM and executing them, like a regular DSP would.
Think of this more like an FPGA - you have a device that is configured for a specific processing algorithm, and data is fed in at wire rate and processed at wire rate.
An example of how a device like this might be used may be in order:
I'm trying to find a radar pulse buried in the noise coming in from my receiver. I want to know the phase delay of the radar pulse - how long from when I sent it till I got it back.
Now, I know what my radar pulse looks like as it goes out. I know that any reflection is going to consist of versions of that pulse shape, delayed and of varying strengths. So what I do is called a correlation - the easiest way to think of this is to imagine having 2 transparencies, one of my outgoing pulse, and one of the incoming signal. Now, I hold them up to the light, and slide the incoming signal across the reference pulse until things match up - that's the point of maximum correlation, and that give me the delay of the signal.
A real correlation function is a bit more complicated as you have to allow for the signal level to be changed - if I am looking for a signal of N samples in a received data stream of M samples, I have to do M*N multiply and add operations to get my correlation. Now, for a radar signal I might be sampling at over a billion samples a second, and looking for a chirp of a 100 ns would give me over 100 billion MAC operations a second. There are ways to do that with conventional DSPs, but they are a galloping BITCH to do (you basically make a cluster of DSPs, and each DSP takes a part of the signal. Synchronising that is a bitch.)
This device would work by having the shape of the outbound pulse represented in the structure of the device itself, and the MACs are done by taking the incoming data stream and projecting it on the structure - thus you do all your processing in parallel, and at wire speed. You get a pulse out when the incoming signal matched the signal you ar looking for.
www.eFax.com are spammers
... at the Lenslet page, the unit actually has several components. The VMM (vector matrix multiplier) does 8000 MAC (matrix array calculations) but there is a VPU (vector processing unit) that comes in at 128 Giga-ops and which would be the bottleneck in the whole setup. No question this is a huge improvement BUT to put it in perspective, it is a DSP only, not a computer system (although some neural network weenies might see a way of turning this into something more than just a DSP). In any case, the bottlenecks will come from the equipment it has to operate with both onboard and off.
Still, note that it's developed with Matlab. Now surely that is the Holy Grail of research, a bitchin' language with an awesome tailored processor. Imagine the logo Matlab [Lenslet Inside].
"Consensus" in science is _always_ a political construct.
I'm not quite sure you understand what this processor is, and how it works. This is *NOT* like a Pentium with a faster clockspeed. This is a signal processing chip which, rather than really executing code, uses a series of optical filters to do massively expensive mathematical operations. This has the following set of properties:
The original poster was right, this IS the future, at least for now
Cheers,
Justin
Disclaimer: I'm one semester away from my bachelors in Physics
Seriously though, basically this chip can do very quickly what the SETI@Home software does on PCs. Fast fourier transforms and the like... Think about completing a calculation unit every 30 seconds instead of 8 hours and 40 minutes. That is the ball park. I wonder if the precision will be the same.
The article says "The Ablaze(TM) is the Spatial Light Modulator (SLM) in the optical core of the EnLight256(TM)". Going by the graphic in the Inquirer article, they shine a row of blinking lights through a LCD-like device (and some lenses and mirrors I assume) and collect the results in a column of light sensors on the other end.
Each pattern of on/off elements on the LCD-like device gives them a different transformation running at however fast you could emit and sense the light. I doubt they mechanically move the optical arrangment so that would seem to limit the number of transformations. Some of the LCD patterns might give useful transformations. A vector multiply, a Fast Fourier Transform (maybe) or a sort (I doubt it)?
If the numbers are an analog light intensity level the precision would depend on how precise the light emitters and sensors you have are. Packaging the mirrors and lenses small enough is a neat trick. Having a problem that fits the available transformations and can supply data in and out fast enougth is another. I wonder anything useful can be done by quickly switching LCD matrix pattern, or directly feeding outputs back as inputs?
HAH!!! I found the post I lost in my browser cache!!! Now you can read and enjoy it.
Sorry if I came off snide. I didn't mean too. I took your comment to be snide and responded a little harshly. I will be much more civil.
Here's a better and longer explanation of what I said before.
With the present theory of computing (electronic and optical) you have a clock that drives the processor. Actions that take place such as moves, adds, rotates, and multiplies all take place because of an enabling clock pulse. There are bits that will be set on or off that are read and written at the clock pulse. For any digital computer, there must be on and off thresholds - above a certain threshold is on, and below a certain threshold is off - in between is not used and possibly an error.
With your degree, I'm sure you know all this.
Once a bit is set, that is a voltage applied or a light turned on, there is a certain amount of wait time until that signal propogates and can be read. The slope of the voltage vs time plot is rather shallow compared to a light intensity vs time plot. So in order to be reasonably sure that all bits your enabled have reached the threshold will take longer for an electronic signal than for an optical signal. This is why overclockers often increase the voltage on their cpu's, to decrease the time it takes for the signal to reach the threshold value. However, since the intensity of light increases much faster (the packet is tigher) the clock can be set at a MUCH faster rate and still maintain good assurance that all signals have reached their threshold value.
Hopefully this better explains what I said before. You can only imagine if I had tried to type that all into my original post!
Slashdot Syndrome: the sudden, extreme urge to correct someone in order to validate one's self.
You can build both digital and analog computers from analog components. Indeed, aside from "ideal switches", I'd argue that most circuit components are analog. The poster to which you responded is correct -- a system built around discrete levels selected from a continuous domain is considered digital. A system built around continuous levels is considered analog.
So, if you build a computer around 4-level signaling, it'd still be digital. Each signal corresponds to a "digit," likely valued 0 thru 3. If you built a differential-equation analyser out of op-amps, resistors, capacitors and inductors to represent the derivatives/integrals in your system of equations, and operated it within the linear region of the op-amps, you'd have an analog computer. Those aren't very popular anymore, nor have they been for some time.
The options you suggested a couple posts up are all variants of digital signaling. The first one is binary signaling. The second one is discrete amplitude modulation. The third one is multi-frequency discrete amplitude modulation.
The real question is, can you build large numbers of sufficiently small gates for a given signaling method? Additionally, do those gates aggregate into computing structures we know how to use effectively?
Related side topic: 3-valued logic (ternary computing) likely won't take off until its proponents have good answers for both questions. Something tells me the circuits and the gates are the smaller of the two problems. The change in programming paradigm is too big.
--JoeProgram Intellivision!