Slashdot Mirror
10GHz Processors and Ultraviolet Lithography
hoyosa writes "This article on zd-net reports that Extreme Ultraviolet LLC has built the first ultraviolet lithography stand for manufacturing processors. Will this make silicone obsolete? "
Some interesting bits in there. Also "Soon" means we won't see actual
chips until oh, say 2005, so don't hold your breath or anything.
Egads, I hope not! It's my favorite synthetic rubber!
Anonymous Luddite: "What do you think of the dehumanizing effects of the Internet?"
Andy Grove: "Not Much."
No, this does not have anything to do with breast implants, but it WILL make silicon obsolete.
Remember, there were no nuclear weapons before women were allowed to vote.
Would not UV lithography work on silicon?
Bruce
Bruce Perens.
UV lithography has nothing to do with silicon (or silicone, for that matter...)
It just means using light with a shorter wavelength to etch the silicon wafer, allowing you to use a smaller micron process than you could with longer wavelengths.
You'd still use silicon for the wafer. To say otherwise is like saying that deisel fuel makes cars obsolete. They're entirely different problems.
Kevin Fox
--Mike--
That article is over a year old
Actually I believe right now the fastest available CPU is the 1.2Ghz EV7. That is if you're talking about speed of computation. If your talking about clock speed (Which is still measuring the speed of something, though not necissaritly performance) the fastest chips run much faster then 1Ghz. It seems to me like the article is talking about clock speed.
I noticed that in the article, the author mentions Moore's Law as stating that transistor densities double every 18-24 months. Wasn't it originally 12 months, then changed to every 12-18 months?
"Will this make silicone obsolete?"
Silicone, used in breast implants, will NEVER become obsolete, as long there breathes horney men everywhere!
Silcon, on the other hand, may be superseded by some other material as a semiconductor.
Heh!
Note the date of the article: January 11, 2001 2:41 PM PT
now - 2ghz
June 2003 - 4ghz
January 2005 - 8ghz
Spring 2005 - 10ghz
It's hard to be religious when certain people are never incinerated by bolts of lightning.
the fastest processors on terms of clock speed may be the p4 and its multigigahertz, but think of all the other processor lines that are MUCH faster at lower clock speeds computationaly wise(powerpc-power4, alpha, itanium...)
i'm wondering what will happen with 10 Ghz processor because every cycle, lights can only travel about 3 cm...?
Extreme Ultraviolet LLC has built the first ultraviolet lithography stand for manufacturing processors.
Um, we've been using UV for a while now. This company has built the first _Extreme_ UV rig. This is especially obvious as a press release when you realize that they can define EUV as beginning more or less wherever they feel like. The term "EUV" was coined when "X-Rays" got a bad name in lithography circles (it used to be "deep UV", "Soft X-Rays", "Hard X-Rays").
Will this make silicone obsolete?
a) "Silicon".
b) No.
The article says:
"EUV technology is very extendable...and we have demonstrated that it would work down to the 30-nanometer level," Gwyn said.
Barring a new invention, which is always possible, "It should take us to the end of silicon...as we know it today," he said.
In english: The limits of silicon technology will run out before the limits of EUV technology.
They're not ending silicon - they're saying that as long as silicon will be around, photolithography will be around.
While EUV technology is very likely to dominate the mass markets like x86s CPU, northbridges, etc. E-beam technology could bring much more competition to the market. As the article and serveral other source told us, e-beam tech. "draws" the transitors one by one to the silicone. This drawing process is much slower than the normal mask-based lithography. But you do not need a mask, you can make changes to the chip layout much faster because you don't need to make new masks and must just change the programming of your e-beam chip printer.
This could enable cost-effective low-volume chip series made with a cutting edge manufacturing process. It could also make expensive and "slow" fpga based chip emulators obsolete. It could also be the break-through for open hardware because open chip design could be manufatured without big finacial problems.
Jan
Heh, something called "freudian slip" comes to mind..
;-)
The age old question of "Whats on a man's mind" can now be awnsered: Silicone and not silicon
Did you notice the date on this article?? Jan. 11th 2001!! Last year! hmmm.. old news?
Great... so can we expect Intel to keep extending the pipeline in the processor so they can up the clock rate? Or are they going to actually going to improve the processor for something other than Quake?
Nowadays all I associate clockrate with is Intel's marketing machine. AMD has slowly increased the clock rate and kept the price/performance gap decent AFAIK.
At any rate, sounds like good engineering innovation.
While e-beam technology is, as stated, extremely flexible and useful as a 'one-off' prototyping method, it is also SLOW AS HECK. With millions of transistors, drawing one by one is not a solution for mass-production. With a wafer holding perhaps thousands of dies, I have heard from those in the industry that it can take up to 10 hours (hours!) for one wafer to be 'drawn'.
This Will Not Do.
-- "They say that time changes things. The truth is, you have to change them yourself." (Andy Warhol, adapted)
And remember, you have to make a pass through the E-beam machine for each layer.
In which case, the top-of-the-line are about 2 GHz, at room temp. The best overclocking I've seen almost trippled performance, putting the theoretical top-of-the-line at 6 GHz.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
RISC's are above 1GHz :)
UltraSparc III @ 1.05GHz
Alpha 21264 @ 1.001GHz
Hmm... what we need then is an e-beam with some sort of prism-like splitter in front of it... sort of like when you you glue ten pens together to write 10 lines of text at once. If you could split a single e-beam into 1,000 parallel e-beams, separated by the right distances, you could then write 1,000 chips in parallel while still avoiding the need to draw up a mask.
I don't care if it's 90,000 hectares. That lake was not my doing.
There are two big unsolved problems with "extreme ultraviolet" lithography, which is really X-ray lithography. First, you need a coherent X-ray source. The proposed options are a synchrotron, which is big (house-sized) and expensive, or an X-ray laser, which nobody has yet made work. Sandia has claimed a laser-pumped "plasma" source, but it doesn't yet have enough power to do the job.
The other problem is that the masks have to be almost perfect down to the atomic level. Surprisingly, there are ways to do this. It looks like that problem will be solved.
However, the whole technology is nowhere near working. The major web pages on the subject haven't been updated for a year or so, which is a bad sign. Much of the work is being done at the old A-bomb labs (LLNL and Sandia), which today are sort of senior activity centers for old physicists. All the articles seem to come from there. We're not seeing much in the way of EUV articles from semiconductor-fab equipment manufacturers yet.
There's considerable speculation in the industry that there might be a hiatus of a few years around 2004-2006, during which there won't be much progress in line width. This happened once before in the semiconductor industry, in the 1970s. But it's not the end; EUV should eventually work.
Sometime around 2014 or so, we reach the End of Silicon, or at least the end of improvements to lithography on flat silicon, because atoms are too big. Further progress will require a new technology.
Out of curiousity, I compared mpeg play back on both windows2K and BeOS on a dual P3 450 system with 32mb video and 512mb of ram. BeOS was able to play back full screen without skipping. Win2K skipped probably a few frames every minute and was very noticeable. Now I wasn't being very scientific about the test obviously, I just wanted to see if the hardware was capable of full screen (1280 X 1024 res) play back. The answer from a viewing perspective is yes.
Although the older P3 architecture running on 100mhz bus with a single CPU isn't good enough for professional quality video editing (non-linear editing), the newer systems would perform much better. In comparison, a lot of professionals use Mac and Final Cut Pro 3. Having a 10ghz CPU will do very little for non-linear video editing.
The hardware needs better bus architecture and the OS needs to be designed for streaming large amounts of data rapidly, which windows NT kernel currently does not do well. Microsoft has tried to get their systems into film school for editing with poor results. Linux isn't any better in that respect, so the only viable solution (BeOS) for PC video editing is gone.
As more consumers get comfortable with video and music editing, the OS will have to change to meet the demand. If microsoft and intel doesn't, some one else will. This whole mhz battle won't go on forever. At some point, it will cease being the primary factor for consumer PC's.
I don't know that I'd want to use e-beam for prototyping. Your electrical performance might be so different that you could get fooled into thinking you have something which works. Prototyping isn't only for functional verification, it's also needed to see if you're meeting setup/hold times, jitter specs, etc, and that stuff is process sensitive.
A better use for e-beaming is fixing/moding of prototype parts when a bug is found. Mask sets are so expensive now a days, if you suspect you've found the cause of a problem (and its small), you're better off trying to fix a few parts first.
With a wafer holding perhaps thousands of dies, I have heard from those in the industry that it can take up to 10 hours (hours!) for one wafer to be 'drawn'.
I think you mean 1 step in the wafer building process. In 0.13um you currenly get anywhere from 1-3 steps per day, and there are roughly 200 steps to making a wafer. That works out to about 3 month to make one batch of wafers.
Well the 10GHz is the speed of the processor they think they are able to produce with EUV. It has nothing to do with the wavelength used in the etching process. The 10GHz just means that there is a clock thingy inside the processor which says 'tick' 10*10^9 times per second. Although this _does_ create a problem, as you hinted at. As the speed of light in vacuum is constant, it means that the clock signal will only propagate a certain length before the next 'tick'. With a 10GHz clock, the signal will propagate a maximum of about 3cm before the next 'tick' (high school physics, remember; c=lf, where c=speed of wave propagation, l=wavelength, f=frequency). Of course in reality it will propagate even less than 3cm (which is the wavelength you get if c=speed of light in vacuum=3*10^8m/s), as the speed of light inside the chip is somewhat slower than in vacuum. This will mean that the parts of the chip that are further away from the clock will be somewhat out of sync with the parts that are close to the clock. This is something chip designers certainly have to take into account. I'm not sure, maybe it is already an issue today?
And the mips processors too, this one from PMC-Sierra is a 1GHz, 64bit, dual core processor.
Silicon: Chemical Formula Si, Atomic Number 14 in period table of elements, 2nd most common element in Earth's crust behind oxygen. Semiconductor. If silicon were to become obsolete we would need a replacement for stuff like rocks and materials as well as glass and concrete.
Silica: SiO2, as pure a white crystaline material abundant in nature. Fused quartz is pure amorphus silica.
Silicate: chemical compound containing silicon, oxygen, and one or more metals, e.g., aluminum, barium, beryllium, calcium, iron, magnesium, manganese, potassium, sodium, or zirconium. Found in quartz minerals such as agate, amethyst, chalcedony, flint, jasper, onyx, and rock crystal, opal, sand, sandstone, clay, granite, and many other rocks; in skeletal parts of various protists and animals, such as certain sarcodines, diatoms, and sponges, and in the stems and other tissue of higher plants.
Silicone: inorganic polymer in which atoms of silicon and oxygen alternate in a chain; various organic radicals, such as the methyl group, CH3, are bound to the silicon atoms. As linear polymers silicones form a large class of useful fluids and greases. When crosslinked they form a useful class of synthetic rubbers.
If you look more carefully, they meant April 1, 2001. Not that this makes the date selection any more shrewd, but it does help to show the irrelevance of this particular article.
This is a process for lithography into silicon, which is what we already use, except it's slightly more precise.
autopr0n is like, down and stuff.
And ironically you misquoted as well, Moore said that transistor density would double every 18 months, not chip speed (which is closely related).
autopr0n is like, down and stuff.
There are two big unsolved problems with "extreme ultraviolet" lithography, which is really X-ray lithography. First, you need a coherent X-ray source. The proposed options are a synchrotron, which is big (house-sized) and expensive, or an X-ray laser, which nobody has yet made work. Sandia has claimed a laser-pumped "plasma" source, but it doesn't yet have enough power to do the job.
Or, you can use a frequency-doubled UV laser (frequency-doubled Ar:F lasers are the current favourite, if memory serves).
Shining a laser beam through certain types of material produces an output beam that contains frequencies that are harmonics of the input beam's frequency, due to nonlinear interactions between the incident beam and the electrons in the material.
This has been used as a tool in the lab for years, and has been under intense investigation for lithography for quite a while now. My understanding is that frequency-doubled EUV sources are already shipping.
In the mean time, the folks at places like CAMD have had coherent xrays for a while. There are supposed to be about five other labs like this around. I supose you could try to miniturize this technology. If someone comes up with something better, great, but the techniques that can take advantage of it ARE being worked out today.
Friends don't help friends install M$ junk.