Of course, if you're just referring to electricity flowing both ways, that's easily solved by a simple semiconductor junction.
Apparently you didn't read my message.
When the thermal energy of carriers within the diode is much greater than that imbued by the voltage drop across it, the junction conducts in both directions. Diodes stop working when they get too hot.
Just a thought. I wonder if it's reasonable to pump water to elevated storage and use this as overnight power. Overnight power needs are MUCH less than peak day time needs.
You could, but energy density is extremely low (a few tens of joules per kilogram for something you could install in your backyard, vs. tens of megajoules per kilogram for fuel cells or hundreds of kilojoules per kilogram for batteries). The plumbing and storage itself is cheap, but the pump/generator will probably cost more than batteries and a power converter would.
Two black holes combining into one huge black hole isn't going to do anything that they wouldn't do otherwise....Except releasing gravity waves of strong enough magnitude to be detected from great distances.
Even *one* processor often ends up being memory-bound - 25 on one die will cause most to be idly stalled on memory loads.
Did I mention that each one has an on-chip block of memory?
It doesn't matter. Working set size for most problems is far larger than you can reasonably cram into 1/25th of a die (or arguably even a whole die, though that claim's harder to make now that HP's embedded DRAM caches are maturing). Or to put it another way, only a small subset of problems will have a small enough footprint for this processor to be better than a less aggressively muticore design at solving them.
Give it a few more linewidth shrinks, and sure, you'll have enough cache per core, but by then everyone and their kid brother will also be rolling out CMP systems.
I'm afraid that in the absence of hard data, I remain skeptical.
and the very impressive 25X chip design (25 asynchronous processors in one tiny, low power chip, interfacing directly to an SDRAM or whatever else you want -- each of the output pins is software controlled by one processor)
I'd be careful calling this impressive. Even *one* processor often ends up being memory-bound - 25 on one die will cause most to be idly stalled on memory loads.
Also, the previous article on this chip said that the pinout was chosen so that it could be put back-to-back with a specific SRAM chip, not SDRAM.
Another poster called into question the claim that you could have all of those processors active at once without overheating, but without actually checking a chip or reading a detailed electrical specs sheet, I can't confirm or refute that allegation.
What about the rest of the EM spectrum. If electricity could be gotten from that, it would be even better.
Not by much. Most of the energy emitted by a hot object is near the peak of the black-body curve. The sun's surface is hot enough to put this well into the visible range (and enough of it beyond that range to give beachgoers a nasty sunburn). If you can process everything from near-IR to near-UV (or farther), you've got almost all of it.
Here's a question for you, and for everyone else: Would a solar cell continue to operate in an ambient temperature sufficient to generate that frequency in black-body radiation?
I *think* the answer is "no", as thermal energy would cause current to flow both ways across the junction you're trying to use to generate power, but as this is not my area of expertise, I could easily be wrong.
As I understand it, UV light hits the earth at all hours.
Does anyone know how much UV hits the earth during the night?
Almost none. Virtually all of the light that strikes Earth comes from the sun.
As another poster pointed out, you may be confusing this with the mid-IR glow that warm objects (including the ground and the air) give off. The amounts of energy involved are very low, and room-temperature thermal IR is difficult to convert to electricity efficiently.
Any solar power scheme (and so any photovoltaic scheme) has to have enough storage capacity to power the load overnight. Ideally, it should be able to provide power for several days, in case of cloud cover/rain/whatever. This is why most home-powering schemes involve large battery arrays. A city-powering solar plant would probably use fuel cells (energy density is much higher, and there are off-the-shelf models of power-plant scale already available and in use).
Why do you think Java and, to a lesser extent, C# are so popular right now? ESPECIALLY for teaching? Because with Java and C#, it's very, very hard to write code that can break the system it's running on.
It's also very hard with C/C++. The most you break on any system without very broken protection-handling is the faulty program itself.
The reason Java is taught as an introductory language is that it was stylish about 5 years ago. The reason C# is taught as an introductory language is that Microsoft threw a lot of money at universities to teach it, and at marketing to attempt to make it stylish.
It boggles my mind that people in second-year programming courses at my university don't know what a pointer is, because it wasn't covered in their first-year programming course (which used Java).
Languages with built-in safeguards are great, if that's your primary concern, but programming courses in university are supposed to teach you about all aspects of programming that you might reasonably encounter. If someone graduates without knowing how to debug memory errors and then has to maintain a C++ program, God help us all. This is also why we're forced to learn Lisp/Scheme and exposed to Fortran at some point - exposure to the concepts is what's important.
As far as what's used in industry is concerned, first likelihood is whatever the shop has used for the past several years (anything from VC++/VB down to Cobol, depending on where you're working), and second likelihood is whatever the industry fad was when upper management was setting up specifications.
This chip is more interesting than just the normal megahertz hike. It's the first of the desktop hyperhreaded chips - previously only available in the Xeon range (well, from Intel anyway. Other manufacturers had them).
Which other manufacturers?
To the best of my knowledge, nobody else has built a SMT chip. The Power4 was a CMP chip (multiple cores on one die, not multiple instruction streams sharing the same core). Everything else that I've heard of outside of paper-land has had one and only one instruction stream.
SMT was a great idea, but with transistor count being less of a limit nowadays, CMP seems to have the advantage (as you don't have functional-unit contention between threads).
Each gate requires a certain amount of power to maintain it's state, and a certain amount to change its state. This is where the dissipated Watts number comes from. The faster you want each one to switch (higher MHz), the more current will be consumed in the switch. Multiply this by the number of gates and you get values like 130W. This is however a number that often refers to the power requirement when most of the chip is in operation. Different computations exercise different parts of a modern microprocessor and therefore will require various levels of power.
ObNitPick - it's much simpler to think of this in terms of the total capacitance of the chip, as opposed to counting gates (which have capacitance that radically changes as devices scale).
Layout rules have been more or less the same for a while, so regardless of device size, you'll have roughly the same proportion of the chip being gate, reverse-biased diffusion region (for non-SOI chips), metal that's near the substrate, and so forth. Multiply the area of the chip by this fraction and by the capacitance per unit area for the region type in question, and you get the total capacitance. Assume some fraction of this is switched on each clock, and use CV^2/2 to get the energy lost per clock (it's dissipated resistively in the charging/discharging transistors).
Summary: power loss is (mostly) proportional to the square of the core voltage, times the core's area, times a capacitance per area value, times a scaling factor.
The computers of yesteryear, as you confirm, had much higher levels of power consumption. This, I believe, is mostly due to larger, less efficient gates and more discrete logic (less functional consolidation). Also, the equipment of yesterday had to spin larger hard drives (more energy required) and big tape motors, etc.
The computers of yesteryear (for the last decade or so, at least) had high power dissipation due to much higher supply voltage (the "V^2" in CV^2/2).
We've kept cores at the same size or larger (due to fancier implementation designed to improve performance per clock), and we've driven up the clock speed. Something has to give to keep power sane, and so far it's been voltage (though SOI helps by decreasing some of the capacitance).
There are other factors that change too (leakage is a problem in large, fast SRAM arrays [cache], and the capacitance per area shifts for several reasons), but as an approximation the analysis above holds quite well.
We're actually in for a bit of stickiness soon, as we're approaching the useful limits to the supply voltage for silicon (though we still have quite a ways to go, and there are biasing tricks you can play to make the swing lower for a given supply voltage).
Many of the papers on the subject are online, and make quite interesting reading.
every irreversible computation creates a net increase entropy (the 2nd law of thermodynamics in action) and unless something really weird is going on (eg supernova producing neutrinos just before she blows)you will see it as heat.
The amount of power dissipated by current microprocessors is many orders of magnitude higher than the minimum required due to entropy arguments.
Thus, entropy arguments aren't terribly useful when trying to figure out how much power a chip will dissipate.
There are a few interesting papers on the subject floating around; the ones that discuss the limits of transistor technology usually touch on this.
The *only* disadvantage I found so far is battery life. With a wireless CF card, you can use it for about 1 - 1.5 hours. Now that's bad. Of course, new 802.11b CF cards (type 2) are out and use less power, but I don't feel like shelling another 80$.
There's always the solution that I considered for my TI-81 calculator: Duct-tape a D-cell pack to the back...
Acoording to: Internation copper study group [nrcan.gc.ca], the world copper production is about 15000000 Kg/yr.
See my previous post on this topic (better yet, go to my original post, set your config to threshold -1 nested, and read the whole tree).
Upshot: Aluminum is probably the best (resistive) material to build this out of, and if we assume we're power-limited (i.e. put a smelter and railway beside each power plant in the project), we can produce enough within a reasonable length of time. There's no shortage of ore (we'd just end up with strip-mines dotting the landscape).
Same type of argument applies for LHe-cooled metallic superconductors. We'd use less material, but it would be something less common-as-dirt than aluminum (though nothing exotic). For cooling, we'd probably use LH2 instead of LHe as it would be easier to acquire the quantities needed (about the same operating temperature range).
Please provide numbers and formulas backing up your argumentation. I am very doubtfull.
Consider this construct to be similar in characteristics to a solenoid of radius 6.5e6 metres and length, oh, 4e6 metres. Because length is not substantially larger than radius, we can't just use the solenoid field equation, as field strength outside the solenoid would not be zero. But, if we assume the effective path for integration at the average field strength is a circle, we get about 4e7 metres. Let's be pessimistic and say 1e8 metres.
This gives:
B = 1.26e-6 * I / 1e8
Substituting in B = 1e-4 T (stronger than Earth's current magnetic field), we solve to get:
I = 8e9
So we need a sheet current on the order of 10 billion amperes in the solenoid (divide by the number of windings or mesh cables to get the current in one mesh cable).
The copper cable supplying power to your house has a rated current density of about 100 amperes per square centimetre of cross-sectional area. This is both using air cooling and adding a substantial safety factor. Using the same numbers, we get about 1e6 A/m^2 carrying capacity, giving us a cable of 1e4 m^2 cross-section (100m x 100m) required to carry *all* of the solenoid's sheet current. Parcel this out in smaller cables as you see fit.
Let's sanity check the cable numbers. Copper has a resistivity of 1.7e-8 ohm*m at 20 degrees C. A cable 4e7 metres long with cross-section 1e4 m^2 has a resistance of 1.7e-8 * 4e7 / 1e4 = about 7e-5 ohms. At 1e10 amperes, this gives a power dissipation of about 7e15 W....So we in fact need a thicker cable, or a smaller current (weaker field), either one by a factor of about 30, if we only have 10 TW of available power. If we decide to use a superconducting cable, of course, we have no resistive power losses; we just have to make sure the field strength near any of the cables is less than about 1 T (for LN2-cooled superconductors) or 10 T (for LHe-cooled superconductors). A superconducting cable is much more expensive than a copper cable, but much less cabling is required. Alternatively, we could build more power plants, but building more cabling is almost certainly cheaper.
To calculate energy stored, we need the inductance of the solenoid:
L = mu0 * A / length (1 winding)
A loop with radius 6.5e6 metres has an area of about 1.27e14 square metres. Therefore:
L = 1.26e-6 * 1.27e14 / 1e8 L = about 1.6 H
Stored energy is therefore:
E = 0.5 * 1.6 H * (1e10 A)^2 E = 8e19 J
Assuming 10 TW as our rate of energy transfer to the magnet, we need 8e6 seconds to charge the solenoid, or about 14 weeks. If our power supply is more modest (or if resistive losses are substantial), charging time is longer. However, it would still be accomplished within the span of a few years even in the worst case.
In summary, building and powering up the device is possible, though superconducting cables would make the construction much easier. As per my previous messages, you can certainly smelt enough aluminum for a resistive cable within a reasonable length of time. Producing the required amounts of high-temperature superconductor for a liquid nitrogen cooled cable is open to question. An ordinary metal cable could be made to superconduct at liquid helium temperatures, but maintaining the liquid helium envelope would consume a substantial amount of power (liquid nitrogen is much easier to make). Both methods use off-the-shelf technology that's already widely used in industry (LN2 for power transfer cables, LHe for magnets in MRI machines and particle accelerators).
Also, you would need to take into account the fact that the melted iron current in the earth's core would react to this field. Or, seen otherwise, if the pole were to flip, they'd rip appart you little wire around the equator in no time.
The solenoid described above actually stores more energy than the Earth's magnetic field does in *total*. If the entire core decided to do a back flip, the solenoid could take it. In practice, core disturbances are almost certainly much smaller - the core's field represents the balance point where power input matches resistive and other losses within the dynamo. The core's heat source doesn't have the ability to change the dynamo's state very quickly (field flips take thousands of years).
Satisfied now? It's easy enough to check my numbers.
Wouldnt something that dense have a tendancy to try to burst into flames?
At 17 W/processor, not really. According to one of the many press releases, this is using a 0.13 micron version of one of their older processors clocked at something like 600 MHz. I'd worry about the bus chipset heating up more than the processors.
It's interesting to look at the implications of a design like this. Highly parallel systems tend to be communications-limited, and systems that deal with large workloads tend to be memory-bandwidth-limited in general. All of this points to the processor not being the bottleneck. SGI appears to have designed with this in mind, using processors optimized for power instead of performance to improve density.
One reason why the Earth doesn't have an atmosphere like that of Venus is because of photosynthesis. On Earth, organisms using photosynthesis to produce energy consumed great volumes of carbon dioxide and produced oxygen as a waste product.
Photosynthesis indeed changed the chemical composition of the atmosphere, but was not responsible for the fact that Venus has a whole lot _more_ atmosphere....On further reading, I find several references that say that most of Earth's original CO2 atmosphere was locked in carbonate rocks, made possible by the presence of liquid water as a solvent. So, the lunar perturbation argument I'd heard was incorrect. My mistake.
There isn't a strong concenus that the "dynamo" theory is correct. It is a bit of a mystery how the Earth's magnetic field is generated. There isn't a better theory currently, but the mechanism isn't fully understood.
My point was that we are far from knowing *nothing* about how planetary magnetic fields are generated, and can in fact be reasonably confident that some extension or variant of a dynamo model will in fact accurately reflect reality.
Ok wiseguy. So you've got a forward model of MHD in spherical shells at high Reynolds number with perfect predictive power? Yeh Sure.
Ok, I'll bite.
How does absence of an exquisitely detailed model with perfect predictive power qualify as "not knowing [at all] how the planet's magnetic field is generated"?
Especially when you'll have a lot of fun finding a model with perfect predictive power for *anything*? Should we go back to the "well, maybe it _is_ a giant bar magnet" stage alluded to in the post my reply was attached to?
Nice try.
PA's dream come true.
on
Gaming Goodness
·
· Score: 5, Funny
I've got another currently feasible experiment that could provide enough power for this, that could likely be implemented over the course of the next 20 years if anyone wanted to do it.
But it's long so if you want to talk about it email me because this thread will probably drop off the map soon. =)
That would be easier to do with your email address.
Or perhaps make a post about the topic in your journal? That's probably the simplest way for us to get into a long discussion without losing context.
So if there's that much energy in the magnetic field, couldn't we build a Great Loop to suck the energy out of the Earth's natural field?
We could only actually tap much energy either by having a convenient large conducting object moving quickly with respect to the Earth with kinetic energy larger than the energy we want to tap, or by adding our load within the electric current loops in the dynamo (in the Earth's outer core). Neither is likely to be practical.
If the loop were built, there would be coupling between it and the dynamo, so we might get some power out, but it would probably be much less than the total amount stored.
The only way we could damage the field would be to affect the subsurface eddies that generate it. And if we could do that, then the reverse is true -- the Great Loop could be used to move the eddies back into position, averting this 'disaster'.
This would work, though on a much coarser level (we'd have a hard time affecting individual eddies, but turbulence domains would tend to align with our field when they settled down)....Actually, they'd probably be aligned _against_ our field, making a weak pocket in the middle (our field drops off more slowly than the core's field due to larger radius).
So if the Earth's magnetic field is generated by eddies, how much power would it take to push them around?
To completely rearrange them would probably take energy comparable to the energy stored in the field. What power this translates to depends on how long you want it to take. The minimum amount of power required - and the amount required to stabilize the dynamo, if we want to do that - would likely be much lower (just enough to offset parasitic resistance losses within the dynamo).
just out ouf curiosity, did you bother to calculate just how much raw copper you'd need to make ten thousand one-meter cables each long enough to circle the earth's equator?
*pulls out calculator* About four trillion metric tonnes, give or take a factor of four or so. If you're using aluminum, divide by about a factor of two (it's a third the density, but slightly more resistive).
and if you did, how does that compare to the amount of copper mined in the world every year? or even the entire amount of available copper in the world?
The amount available is obscenely large - you'd just have to strip-mine the faces of all continents to get at it.
Annual production of copper is around 16 Mt. Annual production of aluminum, which is probably more abundant given that the crust is aluminosilicates, is 26 Mt.
If we needed to build the cable badly enough to invest the effort, we'd vastly increase production (probably of aluminum, again because it's common). Assuming that we have enough bauxite strip mines and smelters to make power the limiting production factor, we'd produce about 0.1 Mt per second using the amount of power we'd devote to powering up the loop. It would take us about 5-10 years to produce the required quantity, not counting the time required to build smelters next to all of the power plants, railways for transport, and so forth (though some of the transport work has already been done, at coal-fuelled plants, at least).
It's not something that's _likely_ to be done, but it's _possible_ to do with the world's current industrial capability. Which is what makes it a fun thought-experiment.
oh, and by the way, given the amount of force that woukd be required to cut or break a hundred meter copper cable in the first place, i dont think the 10-gigaton or so discharge that results would be all that much more destructive.
It would, by about a factor of at least a hundred million.
How much dynamite does it take to turn a city block full of office buildings into a hundred-metre crater? That's about the level of force involved (maybe add a factor of 100 for the added weight and strength).
Slashdot Mirror
Of course, if you're just referring to electricity flowing both ways, that's easily solved by a simple semiconductor junction.
Apparently you didn't read my message.
When the thermal energy of carriers within the diode is much greater than that imbued by the voltage drop across it, the junction conducts in both directions. Diodes stop working when they get too hot.
Just a thought. I wonder if it's reasonable to pump water to elevated storage and use this as overnight power. Overnight power needs are MUCH less than peak day time needs.
You could, but energy density is extremely low (a few tens of joules per kilogram for something you could install in your backyard, vs. tens of megajoules per kilogram for fuel cells or hundreds of kilojoules per kilogram for batteries). The plumbing and storage itself is cheap, but the pump/generator will probably cost more than batteries and a power converter would.
Two black holes combining into one huge black hole isn't going to do anything that they wouldn't do otherwise. ...Except releasing gravity waves of strong enough magnitude to be detected from great distances.
Magnitude matters.
Even *one* processor often ends up being memory-bound - 25 on one die will cause most to be idly stalled on memory loads.
Did I mention that each one has an on-chip block of memory?
It doesn't matter. Working set size for most problems is far larger than you can reasonably cram into 1/25th of a die (or arguably even a whole die, though that claim's harder to make now that HP's embedded DRAM caches are maturing). Or to put it another way, only a small subset of problems will have a small enough footprint for this processor to be better than a less aggressively muticore design at solving them.
Give it a few more linewidth shrinks, and sure, you'll have enough cache per core, but by then everyone and their kid brother will also be rolling out CMP systems.
I'm afraid that in the absence of hard data, I remain skeptical.
and the very impressive 25X chip design (25 asynchronous processors in one tiny, low power chip, interfacing directly to an SDRAM or whatever else you want -- each of the output pins is software controlled by one processor)
I'd be careful calling this impressive. Even *one* processor often ends up being memory-bound - 25 on one die will cause most to be idly stalled on memory loads.
Also, the previous article on this chip said that the pinout was chosen so that it could be put back-to-back with a specific SRAM chip, not SDRAM.
Another poster called into question the claim that you could have all of those processors active at once without overheating, but without actually checking a chip or reading a detailed electrical specs sheet, I can't confirm or refute that allegation.
What about the rest of the EM spectrum. If electricity could be gotten from that, it would be even better.
Not by much. Most of the energy emitted by a hot object is near the peak of the black-body curve. The sun's surface is hot enough to put this well into the visible range (and enough of it beyond that range to give beachgoers a nasty sunburn). If you can process everything from near-IR to near-UV (or farther), you've got almost all of it.
Here's a question for you, and for everyone else: Would a solar cell continue to operate in an ambient temperature sufficient to generate that frequency in black-body radiation?
I *think* the answer is "no", as thermal energy would cause current to flow both ways across the junction you're trying to use to generate power, but as this is not my area of expertise, I could easily be wrong.
As I understand it, UV light hits the earth at all hours.
Does anyone know how much UV hits the earth during the night?
Almost none. Virtually all of the light that strikes Earth comes from the sun.
As another poster pointed out, you may be confusing this with the mid-IR glow that warm objects (including the ground and the air) give off. The amounts of energy involved are very low, and room-temperature thermal IR is difficult to convert to electricity efficiently.
Any solar power scheme (and so any photovoltaic scheme) has to have enough storage capacity to power the load overnight. Ideally, it should be able to provide power for several days, in case of cloud cover/rain/whatever. This is why most home-powering schemes involve large battery arrays. A city-powering solar plant would probably use fuel cells (energy density is much higher, and there are off-the-shelf models of power-plant scale already available and in use).
With all respect to universities, people can and should learn to use pointers, debug, and other stuff on their own spare time.
University time should be spent teaching important basics.
Reading these two lines in succession is either ironic or saddening, and I haven't figured out which.
Explain to me how the concept of pointers, or the modes of thought needed for debugging, are not both fundamental and extremely important?
Why do you think Java and, to a lesser extent, C# are so popular right now? ESPECIALLY for teaching? Because with Java and C#, it's very, very hard to write code that can break the system it's running on.
It's also very hard with C/C++. The most you break on any system without very broken protection-handling is the faulty program itself.
The reason Java is taught as an introductory language is that it was stylish about 5 years ago. The reason C# is taught as an introductory language is that Microsoft threw a lot of money at universities to teach it, and at marketing to attempt to make it stylish.
It boggles my mind that people in second-year programming courses at my university don't know what a pointer is, because it wasn't covered in their first-year programming course (which used Java).
Languages with built-in safeguards are great, if that's your primary concern, but programming courses in university are supposed to teach you about all aspects of programming that you might reasonably encounter. If someone graduates without knowing how to debug memory errors and then has to maintain a C++ program, God help us all. This is also why we're forced to learn Lisp/Scheme and exposed to Fortran at some point - exposure to the concepts is what's important.
As far as what's used in industry is concerned, first likelihood is whatever the shop has used for the past several years (anything from VC++/VB down to Cobol, depending on where you're working), and second likelihood is whatever the industry fad was when upper management was setting up specifications.
This chip is more interesting than just the normal megahertz hike. It's the first of the desktop hyperhreaded chips - previously only available in the Xeon range (well, from Intel anyway. Other manufacturers had them).
Which other manufacturers?
To the best of my knowledge, nobody else has built a SMT chip. The Power4 was a CMP chip (multiple cores on one die, not multiple instruction streams sharing the same core). Everything else that I've heard of outside of paper-land has had one and only one instruction stream.
SMT was a great idea, but with transistor count being less of a limit nowadays, CMP seems to have the advantage (as you don't have functional-unit contention between threads).
Each gate requires a certain amount of power to maintain it's state, and a certain amount to change its state. This is where the dissipated Watts number comes from. The faster you want each one to switch (higher MHz), the more current will be consumed in the switch. Multiply this by the number of gates and you get values like 130W. This is however a number that often refers to the power requirement when most of the chip is in operation. Different computations exercise different parts of a modern microprocessor and therefore will require various levels of power.
ObNitPick - it's much simpler to think of this in terms of the total capacitance of the chip, as opposed to counting gates (which have capacitance that radically changes as devices scale).
Layout rules have been more or less the same for a while, so regardless of device size, you'll have roughly the same proportion of the chip being gate, reverse-biased diffusion region (for non-SOI chips), metal that's near the substrate, and so forth. Multiply the area of the chip by this fraction and by the capacitance per unit area for the region type in question, and you get the total capacitance. Assume some fraction of this is switched on each clock, and use CV^2/2 to get the energy lost per clock (it's dissipated resistively in the charging/discharging transistors).
Summary: power loss is (mostly) proportional to the square of the core voltage, times the core's area, times a capacitance per area value, times a scaling factor.
The computers of yesteryear, as you confirm, had much higher levels of power consumption. This, I believe, is mostly due to larger, less efficient gates and more discrete logic (less functional consolidation). Also, the equipment of yesterday had to spin larger hard drives (more energy required) and big tape motors, etc.
The computers of yesteryear (for the last decade or so, at least) had high power dissipation due to much higher supply voltage (the "V^2" in CV^2/2).
We've kept cores at the same size or larger (due to fancier implementation designed to improve performance per clock), and we've driven up the clock speed. Something has to give to keep power sane, and so far it's been voltage (though SOI helps by decreasing some of the capacitance).
There are other factors that change too (leakage is a problem in large, fast SRAM arrays [cache], and the capacitance per area shifts for several reasons), but as an approximation the analysis above holds quite well.
We're actually in for a bit of stickiness soon, as we're approaching the useful limits to the supply voltage for silicon (though we still have quite a ways to go, and there are biasing tricks you can play to make the swing lower for a given supply voltage).
Many of the papers on the subject are online, and make quite interesting reading.
every irreversible computation creates a net increase entropy (the 2nd law of thermodynamics in action) and unless something really weird is going on (eg supernova producing neutrinos just before she blows)you will see it as heat.
The amount of power dissipated by current microprocessors is many orders of magnitude higher than the minimum required due to entropy arguments.
Thus, entropy arguments aren't terribly useful when trying to figure out how much power a chip will dissipate.
There are a few interesting papers on the subject floating around; the ones that discuss the limits of transistor technology usually touch on this.
OK... but I don't remember what iPAQ runs on. Perhaps a good review of the two devices is in order.
Check google, or browse the manufacturers' sites for the spec pages.
Palm, last I heard, has a Dragonball processor (slow, but extremely low power).
iPaq has a Super-H processor (200 MHz, decent FP, not-as-low power [in the 0.5-1w range]).
Zarus has an ARM or Xscale processor (depending on model) (200+ MHz, low-power, good integer/poor FP).
This is off the top of my head. Check the respective manufacturers for more info.
The *only* disadvantage I found so far is battery life. With a wireless CF card, you can use it for about 1 - 1.5 hours. Now that's bad. Of course, new 802.11b CF cards (type 2) are out and use less power, but I don't feel like shelling another 80$.
There's always the solution that I considered for my TI-81 calculator: Duct-tape a D-cell pack to the back...
That gives you 3.584x10^15 Kg of Copper.
:
Acoording to
Internation copper study group
[nrcan.gc.ca], the world copper production is about 15000000 Kg/yr.
See my previous post on this topic (better yet, go to my original post, set your config to threshold -1 nested, and read the whole tree).
Upshot: Aluminum is probably the best (resistive) material to build this out of, and if we assume we're power-limited (i.e. put a smelter and railway beside each power plant in the project), we can produce enough within a reasonable length of time. There's no shortage of ore (we'd just end up with strip-mines dotting the landscape).
Same type of argument applies for LHe-cooled metallic superconductors. We'd use less material, but it would be something less common-as-dirt than aluminum (though nothing exotic). For cooling, we'd probably use LH2 instead of LHe as it would be easier to acquire the quantities needed (about the same operating temperature range).
Please provide numbers and formulas backing up your argumentation. I am very doubtfull.
...So we in fact need a thicker cable, or a smaller current (weaker field), either one by a factor of about 30, if we only have 10 TW of available power. If we decide to use a superconducting cable, of course, we have no resistive power losses; we just have to make sure the field strength near any of the cables is less than about 1 T (for LN2-cooled superconductors) or 10 T (for LHe-cooled superconductors). A superconducting cable is much more expensive than a copper cable, but much less cabling is required. Alternatively, we could build more power plants, but building more cabling is almost certainly cheaper.
Consider this construct to be similar in characteristics to a solenoid of radius 6.5e6 metres and length, oh, 4e6 metres. Because length is not substantially larger than radius, we can't just use the solenoid field equation, as field strength outside the solenoid would not be zero. But, if we assume the effective path for integration at the average field strength is a circle, we get about 4e7 metres. Let's be pessimistic and say 1e8 metres.
This gives:
B = 1.26e-6 * I / 1e8
Substituting in B = 1e-4 T (stronger than Earth's current magnetic field), we solve to get:
I = 8e9
So we need a sheet current on the order of 10 billion amperes in the solenoid (divide by the number of windings or mesh cables to get the current in one mesh cable).
The copper cable supplying power to your house has a rated current density of about 100 amperes per square centimetre of cross-sectional area. This is both using air cooling and adding a substantial safety factor. Using the same numbers, we get about 1e6 A/m^2 carrying capacity, giving us a cable of 1e4 m^2 cross-section (100m x 100m) required to carry *all* of the solenoid's sheet current. Parcel this out in smaller cables as you see fit.
Let's sanity check the cable numbers. Copper has a resistivity of 1.7e-8 ohm*m at 20 degrees C. A cable 4e7 metres long with cross-section 1e4 m^2 has a resistance of 1.7e-8 * 4e7 / 1e4 = about 7e-5 ohms. At 1e10 amperes, this gives a power dissipation of about 7e15 W.
To calculate energy stored, we need the inductance of the solenoid:
L = mu0 * A / length (1 winding)
A loop with radius 6.5e6 metres has an area of about 1.27e14 square metres. Therefore:
L = 1.26e-6 * 1.27e14 / 1e8
L = about 1.6 H
Stored energy is therefore:
E = 0.5 * 1.6 H * (1e10 A)^2
E = 8e19 J
Assuming 10 TW as our rate of energy transfer to the magnet, we need 8e6 seconds to charge the solenoid, or about 14 weeks. If our power supply is more modest (or if resistive losses are substantial), charging time is longer. However, it would still be accomplished within the span of a few years even in the worst case.
In summary, building and powering up the device is possible, though superconducting cables would make the construction much easier. As per my previous messages, you can certainly smelt enough aluminum for a resistive cable within a reasonable length of time. Producing the required amounts of high-temperature superconductor for a liquid nitrogen cooled cable is open to question. An ordinary metal cable could be made to superconduct at liquid helium temperatures, but maintaining the liquid helium envelope would consume a substantial amount of power (liquid nitrogen is much easier to make). Both methods use off-the-shelf technology that's already widely used in industry (LN2 for power transfer cables, LHe for magnets in MRI machines and particle accelerators).
Also, you would need to take into account the fact that the melted iron current in the earth's core would react to this field. Or, seen otherwise, if the pole were to flip, they'd rip appart you little wire around the equator in no time.
The solenoid described above actually stores more energy than the Earth's magnetic field does in *total*. If the entire core decided to do a back flip, the solenoid could take it. In practice, core disturbances are almost certainly much smaller - the core's field represents the balance point where power input matches resistive and other losses within the dynamo. The core's heat source doesn't have the ability to change the dynamo's state very quickly (field flips take thousands of years).
Satisfied now? It's easy enough to check my numbers.
Wouldnt something that dense have a tendancy to try to burst into flames?
At 17 W/processor, not really. According to one of the many press releases, this is using a 0.13 micron version of one of their older processors clocked at something like 600 MHz. I'd worry about the bus chipset heating up more than the processors.
It's interesting to look at the implications of a design like this. Highly parallel systems tend to be communications-limited, and systems that deal with large workloads tend to be memory-bandwidth-limited in general. All of this points to the processor not being the bottleneck. SGI appears to have designed with this in mind, using processors optimized for power instead of performance to improve density.
One reason why the Earth doesn't have an atmosphere like that of Venus is because of photosynthesis. On Earth, organisms using photosynthesis to produce energy consumed great volumes of carbon dioxide and produced oxygen as a waste product.
...On further reading, I find several references that say that most of Earth's original CO2 atmosphere was locked in carbonate rocks, made possible by the presence of liquid water as a solvent. So, the lunar perturbation argument I'd heard was incorrect. My mistake.
Photosynthesis indeed changed the chemical composition of the atmosphere, but was not responsible for the fact that Venus has a whole lot _more_ atmosphere.
There isn't a strong concenus that the "dynamo" theory is correct. It is a bit of a mystery how the Earth's magnetic field is generated. There isn't a better theory currently, but the mechanism isn't fully understood.
My point was that we are far from knowing *nothing* about how planetary magnetic fields are generated, and can in fact be reasonably confident that some extension or variant of a dynamo model will in fact accurately reflect reality.
Ok wiseguy. So you've got a forward model of MHD in spherical shells at high Reynolds number with perfect predictive power? Yeh Sure.
Ok, I'll bite.
How does absence of an exquisitely detailed model with perfect predictive power qualify as "not knowing [at all] how the planet's magnetic field is generated"?
Especially when you'll have a lot of fun finding a model with perfect predictive power for *anything*?
Should we go back to the "well, maybe it _is_ a giant bar magnet" stage alluded to in the post my reply was attached to?
Nice try.
A first-person Starcraft?
Someone tell Penny Arcade.
I've got another currently feasible experiment that could provide enough power for this, that could likely be implemented over the course of the next 20 years if anyone wanted to do it.
But it's long so if you want to talk about it email me because this thread will probably drop off the map soon. =)
That would be easier to do with your email address.
Or perhaps make a post about the topic in your journal? That's probably the simplest way for us to get into a long discussion without losing context.
So if there's that much energy in the magnetic field, couldn't we build a Great Loop to suck the energy out of the Earth's natural field?
...Actually, they'd probably be aligned _against_ our field, making a weak pocket in the middle (our field drops off more slowly than the core's field due to larger radius).
We could only actually tap much energy either by having a convenient large conducting object moving quickly with respect to the Earth with kinetic energy larger than the energy we want to tap, or by adding our load within the electric current loops in the dynamo (in the Earth's outer core). Neither is likely to be practical.
If the loop were built, there would be coupling between it and the dynamo, so we might get some power out, but it would probably be much less than the total amount stored.
The only way we could damage the field would be to affect the subsurface eddies that generate it. And if we could do that, then the reverse is true -- the Great Loop could be used to move the eddies back into position, averting this 'disaster'.
This would work, though on a much coarser level (we'd have a hard time affecting individual eddies, but turbulence domains would tend to align with our field when they settled down).
So if the Earth's magnetic field is generated by eddies, how much power would it take to push them around?
To completely rearrange them would probably take energy comparable to the energy stored in the field. What power this translates to depends on how long you want it to take. The minimum amount of power required - and the amount required to stabilize the dynamo, if we want to do that - would likely be much lower (just enough to offset parasitic resistance losses within the dynamo).
just out ouf curiosity, did you bother to calculate just how much raw copper you'd need to make ten thousand one-meter cables each long enough to circle the earth's equator?
*pulls out calculator* About four trillion metric tonnes, give or take a factor of four or so. If you're using aluminum, divide by about a factor of two (it's a third the density, but slightly more resistive).
and if you did, how does that compare to the amount of copper mined in the world every year? or even the entire amount of available copper in the world?
The amount available is obscenely large - you'd just have to strip-mine the faces of all continents to get at it.
Annual production of copper is around 16 Mt. Annual production of aluminum, which is probably more abundant given that the crust is aluminosilicates, is 26 Mt.
If we needed to build the cable badly enough to invest the effort, we'd vastly increase production (probably of aluminum, again because it's common). Assuming that we have enough bauxite strip mines and smelters to make power the limiting production factor, we'd produce about 0.1 Mt per second using the amount of power we'd devote to powering up the loop. It would take us about 5-10 years to produce the required quantity, not counting the time required to build smelters next to all of the power plants, railways for transport, and so forth (though some of the transport work has already been done, at coal-fuelled plants, at least).
It's not something that's _likely_ to be done, but it's _possible_ to do with the world's current industrial capability. Which is what makes it a fun thought-experiment.
oh, and by the way, given the amount of force that woukd be required to cut or break a hundred meter copper cable in the first place, i dont think the 10-gigaton or so discharge that results would be all that much more destructive.
It would, by about a factor of at least a hundred million.
How much dynamite does it take to turn a city block full of office buildings into a hundred-metre crater? That's about the level of force involved (maybe add a factor of 100 for the added weight and strength).