Posts: 16,102
Threads: 733
Joined: Sep 2012
I don't think there's really an issue with reversible computing producing a small amount of heat in and of itself. Although, the process probably isn't going to be perfect in reality, regardless of what may be possible in principle.
The biggest issue I'm seeing here is two-fold:
a) How much energy is going to be involved in intercepting and launching these 'rocks'. Unless you are willing to wait an extremely long time for your computer to get up and running, to take in new input, and to deliver results to the outside world, there is going to be some energy release involved in launching and catching them. Probably also in course corrections over the sort of distances you are proposing, although that would presumably be less of a factor over much smaller distances that could also potentially be used.
b) This system is going to be a lot slower than a conventional irreversible system using conventional cooling methods or even an reversible system using conventional data transmission methods and cooling to deal with the heat generated (I'm assuming that you haven't mentioned using optical or other EM based input/output systems that these are not reversible). Under most circumstances, it seems unlikely that users would be willing to wait for the months to years (or decades) it would take for information to be input/output from this kind of system. Also, it sounds like it would be limited in the types of computation it could do if the we're just talking 'a little new date mixed with a lot of old results or purely old results'.
Whatever performance pluses it would get from producing minimal heat via reversible computing would seem to be overwhelmed by the long wait times to input data or get useful results from it. That doesn't mean that no one would find a use for it (a group like the Backgrounders, who seek to minimize waste heat production might build something like this for some major computing effort, perhaps - they aren't generally bothered by long wait times anyway).
But generally speaking it sounds like the long wait times negate whatever benefit is to be gained by playing ones bit erasing processes at a distance and limiting oneself to portable data storage for input/output rather than using EM based methods. At least under most circumstances. And assuming this all works as advertised of course. I keep having this nasty concern that something is being missed here and that thermodynamics will somehow prevent this from working (the I/O bit, not the reversible computing bit).
So, generally speaking, if I'm understanding this correctly, it might have niche application within the setting, but I don't see an I/O system like this being widely used, because the costs outweigh the benefits in most cases. Reversible computing might have much wider application in the setting in places where waste heat needs to be minimized.
Todd
Posts: 16,102
Threads: 733
Joined: Sep 2012
Meant to loop back to this earlier...
As mentioned, I can see more applications for reversible computing where heat needs to be minimized. A big one could be where there are large collections of nano computers, such as a major construction project in a fairly tight space. Another might be internal nano medical systems operating inside a biont body, which is sensitive to even fairly minor temp increases.
In terms of groups using RC, the Backgrounders were already mentioned. The Negentropists might make extensive use of it given their philosophy. Perhaps using it everywhere max computing speed isn't needed.
As far as the large decentralized system you describe, as mentioned earlier, the Backgrounders might go in for such a thing, both because it would be easier to hide and because they generally operate at a slower speed with much colder metabolisms and a slowed time sense. So the longer wait might not bother them. Such a construct might be made from a Kuiper body or interstellar planet orbiting thru space far from any star.
If you'd like to do a writeup for he EG about reversible computing in general and/or along the lines of any of the above or similar, please write it up and post it to the EG sub forum for consideration.
Thanks!
Todd
Posts: 15
Threads: 1
Joined: Jun 2015
Sorry, new to the discussion groups, I didn't notice for awhile that the 11th post and onwards are on separate pages.
For virtual life where the internal computation is the entire universe, time to interact with the outside universe is a minor consideration. Small data could be beamed to the core and back by light, but yes large data would have to wait for rocks to be moved. I don't know if there are limits on how efficiently energy can be transferred from incoming rocks to outgoing rocks. I doubt that cold temperatures imply slow computations ... cold only implies low entropy. Defense against the outside universe would have to be addressed.
At the moment I'm stuck on the problem of packing a lot of matter into a small space, all at zero gravity. A tube shape has a major radius (central circle to the center axis) and minor radius (central circle to outside surface of the tube). A rotating tube could have centrifugal force mostly cancel gravity of the tube as a whole, but there's still gravity pointed down the minor radius, and variable centrifugal force for the inner and outer portions of the tube. Rotating the surface around the circle as well can cancel the gravity along the minor axis, but ... too many moving parts, I need to do a simulation.
Posts: 16,102
Threads: 733
Joined: Sep 2012
(06-18-2015, 01:47 PM)Bob Jenkins Wrote: Sorry, new to the discussion groups, I didn't notice for awhile that the 11th post and onwards are on separate pages.
No worries
(06-18-2015, 01:47 PM)Bob Jenkins Wrote: For virtual life where the internal computation is the entire universe, time to interact with the outside universe is a minor consideration. Small data could be beamed to the core and back by light, but yes large data would have to wait for rocks to be moved. I don't know if there are limits on how efficiently energy can be transferred from incoming rocks to outgoing rocks. I doubt that cold temperatures imply slow computations ... cold only implies low entropy. Defense against the outside universe would have to be addressed.
True - if your computroniun is running a totally self-contained and solipist virtuality, I suppose that it wouldn't care about I/O latency. Although, I will point out that you hadn't indicated that this is what you were going for until this point
I/O latency would be an issue re 'defense', including protecting against natural phenomena. Perhaps it would be better all around to use optical data links, which can transmit a lot of data for a relatively small amount of energy and heat? Although, I suppose if low temps are the main goal whoever's doing this wouldn't go for that.
Reversible computation is always going to be slower than irreversible, all else being equal, because in order to maintain those low temps it has to run all the computations in reverse, thereby taking twice as long as a system that isn't doing that. This probably isn't a major issue in many cases, simply because the computation rate is so high that there really isn't any inconvenience from an end-user perspective or where other factors (such as wanting to pack a LOT of processors in a small volume) outweigh the raw speed issue.
(06-18-2015, 01:47 PM)Bob Jenkins Wrote: At the moment I'm stuck on the problem of packing a lot of matter into a small space, all at zero gravity. A tube shape has a major radius (central circle to the center axis) and minor radius (central circle to outside surface of the tube). A rotating tube could have centrifugal force mostly cancel gravity of the tube as a whole, but there's still gravity pointed down the minor radius, and variable centrifugal force for the inner and outer portions of the tube. Rotating the surface around the circle as well can cancel the gravity along the minor axis, but ... too many moving parts, I need to do a simulation.
Hm. What are you trying to do this for? How much matter are you going for here and why are you trying to pack it into such a small space (how small of a space).
Bottom lining this: It would probably be much more effective and helpful to provide us with a complete explanation of what it is you're thinking of here from the get go. Based on your posts so far, it sounds like you have an idea or goal at least partially worked out, but it's not clear what it is at this point or where it is to fit into the setting. If you could give us the full picture, we might be able to help you more effectively with it.
Todd
Posts: 15
Threads: 1
Joined: Jun 2015
07-04-2015, 05:06 PM
(This post was last modified: 07-04-2015, 05:08 PM by Bob Jenkins.)
Generally, I'm trying to maximize computation while minimizing entropy increase. This particular design is http://burtleburtle.net/bob/future/bigdata.html , the rotating torus is http://burtleburtle.net/bob/future/dense.html , and the shell is roughly http://burtleburtle.net/bob/future/balloon.html . It's definitely a work in progress, might not be workable at all.
I have many web pages using a Java nbody simulator, like http://burtleburtle.net/bob/scifi/dyson.html . Java as a web display language has committed hara-kiri. All my applets are effectively dead. It's very sad. I'm translating the Java simulator to C++ for exploration now, to see if I can find some orbits for http://burtleburtle.net/bob/future/dense.html that actually work. Later I'll have to translate it to JavaScript too to display the results, and to get all my old applets working again for the modern web.
Posts: 611
Threads: 41
Joined: Dec 2014
You can do that, but I think there's no advantage in it.
The heat produced in reversible computing can be produced at a designated location by doing bit erasures at a distance removed from the computing. That said, it would introduce a speed-of-light roundtrip between the computation and the heat disposal, and that would slow the computation down at least twice as much as just using a superconductor conduit to _directly_ remove heat from the reversible computation site after (local) bit erasure. Ultimately, it would produce the same amount of heat if we get the same amount of reversible efficiency.
When conducting energy (electricity or heat - aka entropy or bits in this case) through a superconductor, the energy propagates at the speed of light. And the superconductor loses its superconductive property if heated above its critical point. That means there is too much energy contained within a given volume of the material. The volume centered at the point of energy input and bounded by the speed-of-light constraint and the superconductor's physical shape, must be large enough that its energy density doesn't exceed the critical point for its material.
So, as with any kind of energy transport, there's a lower bound on the space required to transmit it via superconductor. Probably not a relevant one on the scale of conventional engineering, but extreme engineering seems to be the order of the day here.
Posts: 15
Threads: 1
Joined: Jun 2015
(07-05-2015, 01:50 AM)Bear Wrote: ... it would introduce a speed-of-light roundtrip between the computation and the heat disposal, and that would slow the computation down at least twice as much as just using a superconductor conduit to _directly_ remove heat from the reversible computation site after (local) bit erasure.
It wouldn't slow down computation at all. You can send one batch of bits off to be erased, and immediately start using zeros from another batch that have already been erased.
Posts: 11,610
Threads: 450
Joined: Apr 2013
Welcome to OA, by the way, Bob!
I wonder if you have seen the 'Jenkins' swarm images I've made, variants of which used on several pages, and based on your articles elsewhere.
Posts: 15
Threads: 1
Joined: Jun 2015
(07-05-2015, 08:10 AM)stevebowers Wrote: Welcome to OA, by the way, Bob!
I wonder if you have seen the 'Jenkins' swarm images I've made, variants of which used on several pages, and based on your articles elsewhere.
Yes, thank you for doing that! I was fishing for a place on the internet where productive design discussions about megastructures still happen, and Orion's Arm is my current best guess. There was also this paper that seemed very close to what I'm looking at: http://www.aleph.se/Nada/Jupiter/Brains2.pdf , but that was a long time ago.
Posts: 11,610
Threads: 450
Joined: Apr 2013
Well, I've found your pages very inspiring. Hopefully we can come up with some more images or speculative articles about this fascinating subject.
|