howard@cpocd2.UUCP (03/12/87)
This discussion should move into comp.arch or comp.lsi or ? as it has little to do with space. In article <7718@utzoo.UUCP> henry@utzoo.UUCP (Henry Spencer) writes: >> Then there are Josephson junctions... >> This wasn't possible before because the transistors would dissipate too >> much heat to cool to liquid helium temperatures. Room temperature >> Josephson junctions would blow computing wide open. > >Well, maybe not. The cooling problem wasn't the only reason why JJs haven't >swept the field. They are also fundamentally low-gain devices, which makes >it very hard to build working JJ LSI -- the precise control of characteristics >needed to make low-gain devices work well is nearly impossible in LSI. IBM >concluded that the problem wasn't fixable. There are still better reasons. JJs give you very fast gates, but are hard to integrate (as noted). Now, where does delay come from in large-scale high- performance computers (the kind JJs would be used in)? Gene Amdahl believed (~5 years ago, answering a question at a Berkeley talk) that it was: 1/3 gate delay 1/3 on-chip interconnect 1/3 system interconnect Assume this is approximately right. Now suppose you have a magical technology with *ZERO* gate delay. How much does your system speed up if all other things are held constant? (If you get an answer other than 50% faster, check your arithmetic.) This limits the gains to be gotten from JJs. And if they have less dense integration levels, the system might actually run *slower*. Communication is more important than gate speed, and integration is the only way to get improved communication speeds because we already know how to send signals at more than half the speed of light. This limits the gains to be gotten from superconducting wires. The wires must get shorter; the system must get smaller. That means putting more circuitry on a chip. Then look at the reliability/maintainability issue. To do repair work on a JJ computer, you need to warm it from 4K to room temp. What does that do to all your delicate wires and transistors? How many temp cycles like that will it take to utterly destroy the chip? You could try to change the temperature slowly, but then what happens to the mean repair time? The conclusion is that superconductivity in general, and JJs in particular, just won't be of much use in general-purpose computers. We're far better off looking for good architectures to make use of the fabrication technology we have. Try reading "The Connection Machine" by W. Danny Hillis for a taste of what computers might look like in 10 or 20 years, when the Von Neumann architecture (and bottleneck) is fading from the scene. Here we have a potential gain of several orders of magnitude, not just a piddling 50%. The only factor in recent discussions which might argue in the other direction is that the brittleness of the reported liquid-N2-temp superconductors is not really a problem for ICs, since the wires would not be drawn but rather etched from a sheet. This suggests that it will be *possible* to make ICs with high-temp superconducting elements, but not that it is worth doing so. -- Howard A. Landman ...!intelca!mipos3!cpocd2!howard
keithl@vice.UUCP (03/14/87)
I worked on Josephson junction logic gates 10 years ago. I work on silicon now. Some comments on JJs in computers: Advantages: JJ's prime advantage is their very low power, allowing them to be placed closely together. With proper design, you can design JJ gates with switch energies of just a few kT. JJ gates can be made to match transmission lines - hard to do at low power with semiconductor circuits requiring >100mV logic swings. At liquid helium temperatures, it's possible to use superfluid effects in the LHe to cool them, allowing 3D stacking of die. Imagine a CRAY, 100x speed, a centimeter on a side (You've heard THAT one before, I'll guess!). Just don't ask how to build it! Given a computer that consumes milliwatts, the cooling problem is simplified, even with Carnot efficiencies included. What is the biggest box in YOUR computer center? Ours says "Liebert". In all the LSI random logic systems I've seen, the active area of the chip is a fraction of a percent, the rest being interconnect. Imagine what superconducting transmission lines can do for you here (Note: superconducting wires are still dispersive - at a Terahertz or so). There is a lot of room for improvement in circuit design. With proper design and close attention to device matching, a JJ circuit shouldn't be much harder to design than an NMOS circuit. While some things are looser, such as junction critical currents, other things like 2e/h seem pretty well under control. Proper circuit design (which IBM didn't appear to understand) relies on those things that are easy to control, and matches out or designs around hard-to-control parameters, and avoids dependence on too may of them (such as the gap voltage). Disadvantages: JJs are W*E*I*R*D. You need new test equipment, design tools, manufacturing equipment and techniques. GaAs has problems with its special equipment requirements; imagine the problems with JJ. JJs don't benefit much from the billions spent in silicon R&D that eventually ends up in consumer and military equipment. However, this isn't stopping Fujitsu, who just built a 16 bit ALU slice with JJ technology. I hope American manufacturers don't find themselves on the wrong side of a huge technological barrier. Even on the right side, we can't always keep up with Asia. Interconnect with the "real world" is a problem. The signals are small and fast. Testing is a similar problem, although Hypress sampling technology can be used, completely in the dewar at low temp. Not cheap or easy, though, and not off-the-shelf technology. Design is odd! Ever tried using Kirchoff's law for quantum mechanical phase? With speed-of-light and unpaired electron currents and flux pinning and wave interference thrown in? Not harder, but very different. When the circuits get fast enough, quantization of the signals themselves (Your logic signals become packets of milli-electron-volt energy) may become a problem. The inconvenience of temperature cycling changes test philosophies. Built-In-Self-Test is a MUST. Redundancy would be very useful. Thermally activated failures such as electromigration and junction leakage are absent, but vibration and ESD may still reach inside the dewar, so "board-swapping" may still be necessary. IBM had a lead junction process that could stand temp cycling, and the Fujitsu Niobium/Aluminum Oxide process should be even more robust. I would worry about connections unmating, though, which is where I'd put the redundancy. Conclusions: It's a lot of work. The engineering probably costs more than other alternatives such as massive parallelism, which should be developed first. But when those schemes hit the wall (and on some problems they will), JJs may be the way out. Don't sell your Niobium mining stock yet :-) -- Keith Lofstrom MS 59-316, Tektronix, PO 500, Beaverton OR 97077 (503)-627-4052
jewett@hpl-opus.HP.COM (Bob Jewett) (03/16/87)
> JJ gates can be made to match transmission lines Because the currents are ~0.2mA, and voltages about 2mV for impedances of about 10 ohms, and power dissipation of roughly 1uW. > At liquid helium temperatures, it's possible to use superfluid > effects in the LHe to cool them, allowing 3D stacking of die. The present excitement is due to the higher temperature superconductors (a thin film has been fabricated at Stanford, Tc=35K) where superfluidity doesn't occur. > Given a computer that consumes milliwatts IBM projected 8 watts for a 370-equivalent with ~3ns cycle time. > simplified, even with Carnot efficiencies included. The typical power multiplier (power_room_temp/power_He) is 2000, or 1000 for very efficient refrigerators to 4.2K, giving a projected room temperature AC requirement of the hypothetical LHe 370 of 16kW. > (Note: superconducting wires are still dispersive - at a Terahertz or so). IBM was able to maintain ~100ps risetimes from chip to circuit carrier. The problem was inductance and crosstalk in the pins to the circuit carrier. This implies a useful inter-chip bandwidth of 3GHz. > With proper design and close attention to device matching, a JJ circuit > shouldn't be much harder to design than an NMOS circuit. IBM had (and Japan has) a lot of bright poeple working on circuit design. None of them has come up with a design that can tolerate a +-50% error in junction current levels. > However, this isn't stopping Fujitsu, who just built a 16 bit ALU slice > with JJ technology. The circuit that Fujitsu reported on at the ISSCC had only 900 gates. It was not clear from either the presentation or the Digest paper whether they ever got all 900 gates to work on one chip. Their reported timing measurement was (using their more conservative number) 1.7ns for 16 bit carry propagation, but this was not measured in the ALU itself. It was measured in a special 36-gate test circuit that duplicated the critical delay path. > But when those schemes hit the wall (and on some problems they will), > JJs may be the way out. IBM dropped JJs when it appeared that there would be further delays and the speed advantage over semiconductor circuits would only be a factor of 3. When the project started, that factor was larger than 10. IBM did demonstrate a cross-section model (CSM) of a signal processor. It ran with a 3ns cycle time, including memory fetch time. The I/O, cooling, shielding and interconnect problems all seemed to have been solved. > Keith Lofstrom Bob Jewett hplabs!jewett
jbuck@epimass.UUCP (Joe Buck) (03/16/87)
In article <1486@vice.TEK.COM> keithl@vice.TEK.COM (Keith Lofstrom) writes: >[ Lots of neat stuff about Josephson junctions ] Seems like the newly discovered liquid-nitrogen-temperature superconductors could change the economics and feasibility for JJ computers radically. Anyone know if some of the people who dropped JJ research (like IBM) are reconsidering? >Conclusions: > It's a lot of work. The engineering probably costs more than other >alternatives such as massive parallelism, which should be developed first. >But when those schemes hit the wall (and on some problems they will), >JJs may be the way out. Don't sell your Niobium mining stock yet :-) Unfortunately, the new superconductors seem to be rare-earth oxides. Sorry, niobium fans! Apparently no one had thought of testing ceramics for superconductivity before, and now that they have, there's an explosion going on. The Japanese have just announced their intention to pour lots of money into high-temperature superconductor research. -- - Joe Buck {hplabs,ihnp4,sun,ames}!oliveb!epimass!jbuck seismo!epiwrl!epimass!jbuck {pesnta,tymix,apple}!epimass!jbuck Entropic Processing, Inc., Cupertino, California
howard@cpocd2.UUCP (Howard A. Landman) (03/16/87)
In article <1486@vice.TEK.COM> keithl@vice.TEK.COM (Keith Lofstrom) writes: >Advantages: > JJ's prime advantage is their very low power, allowing them to be placed >closely together. With proper design, you can design JJ gates with switch >energies of just a few kT. JJ gates can be made to match transmission lines > - hard to do at low power with semiconductor circuits requiring >100mV >logic swings. One problem I failed to mention in my previous posting is data storage. While JJ gates are very low power, the standard way to store a bit in a JJ circuit is to put a current into a superconducting loop. There's a really interesting feature to this current, which is that it is equivalent to some "trapped" magnetic flux through the loop. Now, this flux is governed by the simple Bohr-Sommerfeld quantization that most of you first heard about in the Bohr model of the hydrogen atom: there are discrete energy levels that can be trapped. Thus the minimum you can switch is a single flux quantum; and there exist JJ circuits that do just that! However, the quantization levels depend on the size of the loop, and increase as the loop gets smaller (remember "particle in a box"?). This is all freshman physics, but the implication is that the amount of energy required to store one bit of information goes *UP* as JJ circuits get smaller, and hence the power consumption goes up, and the power density (power per area) goes *WAY* up. Hence JJ cicuits will not scale gracefully to smaller sizes. They are already "quantum limited". This same effect, taken the other way to large loops (say 1 meter across), is what makes JJ flux detectors so extremely sensitive. >Imagine what superconducting transmission lines can do for you here (Note: >superconducting wires are still dispersive - at a Terahertz or so). As I mentioned before, not much. It's easy to get transmission speeds of .5 to .8 the speed of light using non-superconducting wires. Try teflon twin-ax driven by ECL, for example, as used in large Amdahl and Fujitsu machines. Not even superconductors can get > 1.0. -- Howard A. Landman ...!intelca!mipos3!cpocd2!howard
keithl@vice.TEK.COM (Keith Lofstrom) (03/18/87)
In article <504@cpocd2.UUCP>, howard@cpocd2.UUCP (Howard A. Landman) writes: > > However, the quantization levels depend on the size of the loop, and increase > as the loop gets smaller (remember "particle in a box"?). This is all > freshman physics, but the implication is that the amount of energy required > to store one bit of information goes *UP* as JJ circuits get smaller, and > hence the power consumption goes up, and the power density (power per area) > goes *WAY* up. Hence JJ cicuits will not scale gracefully to smaller sizes. > They are already "quantum limited". The flux quantization in a superconducting loop storage cell IS an issue ( but keep in mind there are many ways to store data). Flux is quantized in packets of 2e-15 Weber, which is 2e-15 tesla-meter2, or .002 tesla-micron2. This says that the voltage in the cell times the switching time in the cell is constant - 2mV will switch the cell in one picosecond, or 200 microvolts for 10 ps switching. Since the flux density goes up as the cell gets smaller, the loop current does go up as the cell gets smaller (I=2e-15Webers/Inductance); a 1 micron loop might have an inductance of 1e-12 henries for a circulating current of 2mA. Thus the power level for a flux-storage cell changing state is perhaps 200 nanowatts when switching in 10ps, 0 the rest of the time. How many memory cells get switched in a clock cycle? In a Von Neumann machine, a pretty small multiple of the word size, if readout is non-destructive. Lets say 100 cells; thats 20 microwatts for the main store. Spacing won't be limited by heat here. It sure is on ECL or GaAs gates. CMOS isn't yet in the speed ballpark, but it will be thermally limited when it does get there. One can store data in latch cells, which can be resistively terminated and aren't constrained by flux quantization (though quantization still shows itself in circuit design). There are some interesting superconducting shift register devices that work sort-of like bubble memories; they will probably be used like disk memory is now. For logic: Plank's constant is 6.6e-34 Joule-sec. That's the energy limit. Thus a 50 Gigahertz clock cycle machine would be able to use 3.3e-23 Joules to switch a gate, and if all this energy were wasted, that would be 1.6 picowatts per gate. A more important energy limit is thermal - the gate energy must be more than a few kT or the circuit will be tripped by occasional thermal quanta (phonons). A good limit would be, say, ~20kT or ~1e-21 Joules at 4K. This makes the gate power a whopping 50 picowatts. IBM switched their gates out to something called the gap voltage, 2mV or so, with on-chip impedances of 10 ohms, requiring 10 microwatts per gate. There are a number of reasons why this is stupid, but what else would you expect from the people who (put your favorite IBM design screw-up here :-) ). The otherwise-wonderful Keyes paper on limits of logic technology used the same silly switch-to-gap assumption, so the idea of 10 microwatt JJ gates has become entrenched. > >Imagine what superconducting transmission lines can do for you here (Note: > >superconducting wires are still dispersive - at a Terahertz or so). > > As I mentioned before, not much. It's easy to get transmission speeds of > .5 to .8 the speed of light using non-superconducting wires. Try teflon > twin-ax driven by ECL, for example, as used in large Amdahl and Fujitsu > machines. Not even superconductors can get > 1.0. You miss the point; speed of light delay is of course a factor of the dielectric, not the conductor. However, those wonderful twin-ax transmission lines don't have the same signal coming out as going in, because of dielectric loss and skin effect on the conductors. The signal is dispersed, and gets slow and ringy. Not an insurmountable problem at 200MHz, but quite a problem at 50 GHz. Superconductors help a lot. The other point is the transmission lines are ON CHIP. Not because superconductors make faster transmission lines, but because JJ circuit impedances are a better match to real-world transmission line impedances. If the permiability of the universe was 10e3 higher, then this would be true of ECL instead. With transmission lines on chip, a designer can do rude and wonderful things like pipelining data in interconnect wires. Imagine a wire running across a die with three successive logic signals propagating down it (?!). Most fast mainframes computers are SSI or MSI, because LSI circuits with 20 milliwatt 40 ps ECL gates are very difficult to cool. Device density for bipolar isn't an insurmountable problem, but busting up your circuit into separate 5 watt chunks is. If the die could be butt-fit and stacked, you could do a heck of a mainframe, often referred to as a "hairy smoking billiard ball". Hairy because of interconnect wires, smoking because of power, and spherical and billiard-ball sized for speed of light delay. With present device power levels replace "smoking" with "incandescent". With JJ replace "smoking" with "gurgling". --- Again, I don't want to claim designing with JJ is cheap, or that the advent of easier-to-cool superconductors will cause silicon IC designers to be sleeping on grates any time soon. Even a room temperature superconducting technology would take vast amounts of design work. However, I suspect that the work will get done, and I would hate to see most of my colleagues too far behind to catch up. JJs are as much of an advance over silicon as silicon is over RELAYS as far as speed/power product goes. An understanding of what the ultimate limits of the technology really are could save us from some rude surprises when we are getting old but not-quite-ready to retire... which is about the time frame I expect JJ to be important in. There are some sampling scope folk around Tek here that were surprised by Hypress, and Sadig Faris gave us a talk two years ago. The next group to be embarassed will be the Spectrum Analyser folk, when the Terahertz mixers start appearing (ever spectrum analysed the infrared from a glass of water?). Then the standards folk. Then the current probe folk. Then the logic analyzer folk. Then... Meanwhile, there's still a buck to be made pushing silicon. Back to work... -- Keith Lofstrom MS 59-316, Tektronix, PO 500, Beaverton OR 97077 (503)-627-4052
reid@sask.UUCP (03/20/87)
In article <978@epimass.UUCP>, jbuck@epimass.UUCP (Joe Buck) writes: > Seems like the newly discovered liquid-nitrogen-temperature > superconductors could change the economics and feasibility for JJ > computers radically. Anyone know if some of the people who dropped > JJ research (like IBM) are reconsidering? There's been some good stuff on superconductors in Electronics lately, including notes about a company that is using superconducting IC's to make very fast oscilloscopes. Cute stuff. Unfortunately, the newest superconductors are _not_ up to liquid nitrogen temperatures. The best I've heard yet was 53 degrees kelvin, which is well below liquid nitrogen (~70 K, I can't remember exactly). - irving - -- reid@sask.uucp {alberta, ihnp4, utcsri}!sask!reid What the world REALLY needs is a good Automatic Bicycle Sharpener.
sewilco@meccts.UUCP (03/22/87)
Someone wrote: >In article <978@epimass.UUCP>, jbuck@epimass.UUCP (Joe Buck) writes: >> Seems like the newly discovered liquid-nitrogen-temperature >> superconductors could change the economics and feasibility for JJ >> computers radically. Anyone know if some of the people who dropped >> JJ research (like IBM) are reconsidering? >... >Unfortunately, the newest superconductors are _not_ up to liquid nitrogen >temperatures. The best I've heard yet was 53 degrees kelvin, which is well >below liquid nitrogen (~70 K, I can't remember exactly). At least eight materials which function at ~90 K have been created. I assume these are what the original poster was referring to. The materials are ceramics, created and shaped as powders and then baked. We're not the only ones who are pleasantly surprised. My information is from a page 1 story in Friday's Minneapolis Star and Tribune. The story is from the N.Y. Times, so I assume it was in their Thursday or Friday paper. The story describes the "Woodstock of physics" which took place Wednesday night during a special overnight conference. Participants were quite excited. Telegraphic presentations until 3 AM, conversation at least until 6 AM. Quotes like "Our lives have changed." AT&T Bell Labs showing flat shapes made just last weekend from the materials. Renewed hopes for room-temp (the new materials could be used on dark side of a satellite in space). -- Scot E. Wilcoxon (guest account) {ihnp4,amdahl,dayton}!meccts!sewilco (612)825-2607 sewilco@MECC.COM ihnp4!meccts!sewilco It may be the event of the century, but "Supernova 1987A" isn't a good catchword.
byron@gitpyr.UUCP (03/22/87)
In article <644@sask.UUCP> reid@sask.UUCP (I am NOT your Sweet Baboo) writes: >In article <978@epimass.UUCP>, jbuck@epimass.UUCP (Joe Buck) writes: > >Unfortunately, the newest superconductors are _not_ up to liquid nitrogen >temperatures. The best I've heard yet was 53 degrees kelvin, which is well >below liquid nitrogen (~70 K, I can't remember exactly). > > - irving - I've got an Atlanta Constitution article (3/20) that refutes that claim. I quote: " In February the key tempreture barrier fell when scientists in Houston and Huntsville reported the appearence of superconductivity in a new yttrium-based compound at tempretures only 283 degrees below zero (since they didn't give a temp scale I assume it's F) nearly 40 degrees above the tempreture of liquid nitrogen." They go on to say how liquid N can be cheaply manufactured (5 cents/quart) and how the elements used in these compounds (yttrium, luteium, scandium, barium, and calcium) can be cheaply had. Another quote " A Los Alamos physicists say that anyone with the proper knowledge, $50 of the right materials and a microwave oven could make their own." A damn interesting concept. -- Byron Jeff E-mail address: ...!{akgua,allegra,amd,hplabs,ihnp4,seismo,ut-ngp}!gatech!gitpyr!byron
jbuck@epimass.UUCP (03/23/87)
In article <644@sask.UUCP> reid@sask.UUCP (I am NOT your Sweet Baboo) writes: >Unfortunately, the newest superconductors are _not_ up to liquid nitrogen >temperatures. The best I've heard yet was 53 degrees kelvin, which is well >below liquid nitrogen (~70 K, I can't remember exactly). Please don't post misinformation to the net. The whole reason for the current excitement is that superconductors with critical temperatures above 90K, and with larger critical magnetic fields than ever seen before, have been discovered. Your local paper must not be too good; this has been front-page news in the NYTimes and San Jose Mercury. -- - Joe Buck {hplabs,ihnp4,sun,ames}!oliveb!epimass!jbuck seismo!epiwrl!epimass!jbuck {pesnta,tymix,apple}!epimass!jbuck Entropic Processing, Inc., Cupertino, California
reid@sask.UUCP (03/23/87)
In article <644@sask.UUCP>, reid@sask.UUCP (stupid me) writes: > Unfortunately, the newest superconductors are _not_ up to liquid nitrogen > temperatures. The best I've heard yet was 53 degrees kelvin, which is well > below liquid nitrogen (~70 K, I can't remember exactly). > - irving - It has been pointed out to me that I'm entirely mistaken. Due to the long delay between our site and the "central" net this may not arrive in time to quell the growing flames, but I have to try... - irving - -- reid@sask.uucp {alberta, ihnp4, utcsri}!sask!reid What the world REALLY needs is a good Automatic Bicycle Sharpener.