Firefox NoScript is the answer to all this kind of stupidity; I think it's worth using Firefox for NoScript alone.
Also, remind your hosts file that intellitxt.com is a synonym for 127.0.0.1
Yes, this is depriving hexus of advertising revenue. If they want advertising revenue, they should produce adverts which do not deeply infuriate their readers. Intelligently-targetted intellitxt might be actually usable, but to have every occurence of 'computer' hyperlinked to Dell's store is of no use to anybody.
Basically, the mission hasn't yet succeeded, and it doesn't seem to be completely certain that it will.
The goal was to measure the frame-dragging effect of the Earth, which is of the order of 40 milli-arcseconds per year; the current calibration (http://einstein.stanford.edu/content/aps_posters/ ExperimentError.pdf) has a one-sigma error of 100 milli-arcseconds per year, significantly larger than the relativistic effect and significantly larger than the effect from the motion of the target star through space. The initial expectation was for an error budget of less than 0.5 mas per year, so there was a lot of work done on measuring the proper motion of the star to that precision.
The problems turn out to be really crazily subtle issues in solid-state physics -- the deposited metal films on the gyroscope and on its housing retain charge in patches large enough that they have to be modelled rather than averaged out -- plus an annoying issue from classical mechanics where the motion of a rigid body around three axes XYZ depends on the ratio of the differences of the moments of inertia X-Z and Y-Z. Whilst the gyroscopes are absurdly precisely made, so the moments of inertia are very close, the ratio of the differences of the moments of inertia is a macroscopic number; this is the same effect, and even a similar cause, to some of the oddities with low-precision floating-point arithmetic.
They'll probably be sorted out, sigma might be reduced by a factor ten after another year's modelling effort, but it seems unlikely that they'll get it down by the factor 200 they were hoping for.
The frame-dragging has already been measured indirectly by looking at the flickers of X-ray sources caused by frame-dragging-induced precession of the accretion discs around black holes, and most of the theories that suggest relativity is wrong would suggest that any oddities would be more pronounced in the huge gravitomagnetic fields near black holes than in the tiny fields near a mass as small, as non-dense and as slowly rotating as Earth.
This isn't 'installing Linux renders the warranty void from that point on', which would clearly be ridiculous and probably illegal, it's 'we can't troubleshoot hardware under Linux'.
Which I think makes some degree of sense; I caused quite a lot of fuss to a local computer supplier when I returned a machine as unable to run stably under Ubuntu 6.10, they changed parts and finally got it to a state they thought reasonable, I found that even then it was totally unstable when I ran large disc-intensive jobs on it, and then installed a Feisty beta and found it's worked perfectly ever since.
Driver bugs look enough like hardware faults that three Canonical employees who I happened to know reckoned the issue was hardware; that a modern laptop would use devices for which the drivers in standard Linux distributions are buggy is not unexpected.
OK, HP are large enough that they could easily use a service model where they put a known-good hard drive with test tools on into the machine, run the test tools on that, and test the machine's current hard drive as the second drive in a machine known-good everywhere else, which is clearly the right thing to do... but I could imagine that the average user would not be very happy with a report 'hardware worked on our test bed, $50 charge for returning a non-broken machine'.
http://www.dramx.org/#fspot gives a decent idea of pricing for the underlying chips; it's just over a dollar per gigabit at the moment, so before considering assembly charges, profit and the like you'd expect a 64G drive to cost about six hundred dollars. I'd expect them to be about a thousand dollars once all the overhead in assembling systems from chips is added in, and rather more than that initially when they've got novelty value.
Disc space hit eight dollars a gigabyte in about June 2000, and a dollar a gigabyte about last year, so give it five years...
I believe the current figure is something more like a hundred thousand writes per memory cell; with a load-distributing file system, this means the lifetime is about the amount of time taken to write the whole device 100k times, which (since flash writes are fairly slow; a few megabytes per second at best) is up there with hard-drive lifetimes.
America had 32,000 warheads in 1966, but has been dismantling them ever since, and now has just under ten thousand, of which about six thousand are operational.
For all practical purposes this is 5900 more than enough.
I am not sure what the ratio between uranium and plutonium in warheads is; I had the impression that, whenever the choice was available, you used plutonium because it's much cheaper to make plutonium in reactors than to make enriched uranium in centrifuges.
Iran and North Korea started off with plutonium programs and then moved to use uranium instead, because it's easier to hide centrifuges than reactors and reprocessing plants... you can ship material between distributed clusters of centrifuges in a truck, whilst reactors and reprocessing plants are unavoidably huge facilities that show up well on satellite imagery.
You convert the plutonium to the oxide and blend it down with natural uranium; the idea is then that, once the fuel has been in a reactor for a while, the fission of the uranium has caused enough of the non-239 plutonium isotopes to be produced that you don't get bomb-grade plutonium by separating out the plutonium fraction... you'd have to do isotope-separation, which is incredibly expensive.
(the problem with the other plutonium isotopes is that they have short half-lives and decay patterns that produce lots of neutrons; this imposes lots of engineering problems when handling the material, neutron radiation being even more inconvenient to deal with than other sorts, and means that a bomb is likely to detonate too early in the implosion process and give a much lower yield than expected)
The fission detonator for a thermonuclear weapon has to be of at least city-damaging proportions - several kilotons of yield - to get the X-ray flux up enough to ignite the fusion fuel. It's a real atom bomb, it requires several tens of kilograms of U235 or Pu239, neither of which is found in any fire detectors I've ever seen.
That's already being done, but there aren't all that many weapons; ten thousand warheads at 20kg each is about two thousand tonnes of fissile material (quite a lot is plutonium, to burn which you need a special mixed-oxide-burning reactor), whilst the known reserve of uranium is estimated at 3.6 million tons.
Err, that's an interesting thing to be taught; the core of the Earth is a sphere of liquid iron. Uranium isn't a siderophile (that is, it doesn't dissolve in liquid iron), so there won't be much uranium in the core (this also means there won't be much uranium in asteroids, in case space enthusiasts want to mention mining those for the uranium).
People have measured the uranium content of the inside of the Earth by looking for neutrinos of the right energy, which are produced during radioactive decay and fly straight through the Earth, and get that the quantity of uranium is enough to produce about 40 terawatts by radioactive decay. There is a crank theory that the core of the Earth has a fission reactor in it, but there's really very limited evidence for that.
Re:Which is why India's looking at thorium...
on
The Coming Uranium Crisis
·
· Score: 5, Interesting
Breeder reactors can also convert U238 to fissile plutonium, which is if anything more useful, since we already have reactors designed for Pu239 and I don't believe any reactors have yet been designed for Th233.
The problem is that people paranoid about nuclear proliferation have successfully made it very politically difficult (it's not technically completely straightforward, you're running rather fiddly chemistry by remote-control in a very high radiation environment) to reprocess spent fuel to get the plutonium out for reuse.
So the current nuclear fuel cycle is the equivalent of running a basic oil refinery, taking out the small jet-fuel fraction from crude oil, and then pumping the remainder back into the ground in places deliberately chosen to make it hard to take it out again. Breeder reactors are the equivalent of those catalytic-cracking columns in refineries which can make something useful out of the heavier crude-oil fractions.
The interesting feature of this announcement is how little computation and how much intelligence in software development was involved by the standards of other large computational projects. The calculation took three days on SAGE, which is an eight-socket dual-core Opteron system with 64GB of memory; it's perhaps three orders of magnitude less calculation than the factorisation of RSA200, or than IBM's work modelling hafnium silicates for developing 45nm processes. It is very much less work than is routinely done commercially for chip simulation or seismic inversion.
On the other hand, even if some of the tricks they used were fairly routine (you have a reasonable idea how large the coefficients are? The coefficients are obtained only by multiplication and addition? Why not calculate modulo lots of coprime one-byte integers and save a factor eight in storage space?) it's remarkably clever work.
are a description of the project aimed at reasonable mathematicians, with a lot more in them than the press release; but I think this item is mostly useful to tell mathematicians how to write a press release which gets picked up about what is fairly abstruse work by the standards of computational group theory.
The commercial market for large-scale computers exists, it just doesn't advertise itself on the top500 list. Banks have big clusters for monte-carlo work (why do you think the most-hyped benchmark application when Intel talks about eight-core Clovertown systems is an asset-pricing model); oil companies have enormous clusters for seismic work.
But both of those are intrinsically parallel jobs, so the clusters don't need the interconnect to run linpack, and there's no advantage to BP or Barclays in appearing on the top500 list -- they're not looking for the kind of researcher who picks his employer by acreage of computers, whilst Aldermaston and the Met Office may well be.
One of the embarrassing truths of computation is that an awful lot of jobs are either small enough to run on a single PC or too large to run on anything; there aren't that many things (climate modelling's the obvious one) constrained by amount of computation.
Cassini runs at 82kbit/second from Saturn, but it's a probe with a larger power budget.
The imager takes one-megapixel, 16bpp images, and compresses them to 100kbyte files for initial transmission, saving the originals in a few gigabytes of onboard flash; it can be instructed to send back uncompressed images if there's something interesting visible.
So an image takes about 20 seconds to transmit, plus about six minutes if you want the uncompressed version; and it takes 45 minutes to get to Earth from Jupiter. From Pluto, the images will take half an hour for the preview and twelve hours for the uncompressed image.
Did anyone else notice the very quick changes from white to red of the colour of the squid?
Humboldt squid ('diablo rojo') are red when annoyed, and have a fearsome reputation since people wanting to dive with them generally dive from ships in the Bay of California squid-fishing fleet, under which circumstances the squid are understandably usually annoyed... apparently they're a good deal more placid when not impaled on pointed hooks (who'd have thought it).
So, if architeuthis behaves like the Humboldt squid, the behaviour of this one at the surface is unlikely to be typical... getting a sufficiently quick and sufficiently quiet robot to observe them is a better approach to seeing them in the wild, but there's been a lot of work put into that, and it didn't work.
It is irritating that water is really only transparent in visible wavelengths, so you can't do the trick that works nicely for land mammals of observing in light that the squid can't itself sense; they may well be better at detecting our lights and sonar than our own cameras are.
That's why Intel's also working on better buses (arrays of silicon lasers), on 3D designs that put a die of level-3 cache directly under the die of CPUs.
It is entirely conceivable, made more so by the enormous Chinese reticence to publish the SMS4 encryption algorithm they're using and to open it to international review.
AES versus a Chinese government-approved algorithm which you can only get a specification for by agreeing to partner with one of eleven Chinese firms is not a difficult decision.
Well, it's probably a better use of $57 billion a year than the standard Japanese economy-boosting habit of building enormous public works; seven billion 1988-dollars for the Honshu to Hokkaido undersea tunnel (longer than the Chunnel), for example.
With all the current focus on China, people forget that Japan has (in dollar terms; the CIA World Factbook figures use slightly dubious purchasing-power-adjusted figures) the second largest economy in the world. It's an economy in a deep recession, but huge government spending is a traditional way out of those.
But presumably the "access denied" error message from an access point doesn't say "Access denied because your MAC isn't 00-20-E0-31-41-59", and there are enough MACs that picking ones at random to see if they happen to be allowed won't give you success at all quickly.
The integrated memory controller doesn't need to be too closely coupled to the CPU core; as far as I can see, it needs only a bus to tell it which location to handle next, and a connection to the L2 cache to put the new data in.
So it shouldn't be too difficult to graft in a new memory controller if a new memory technology becomes necessary; and if you have to replace the CPU to take account of new memory tech, so much the better for the CPU manufacturer.
No. As far as I can tell (I admit I only used long-longs, so interesting things might happen after 18 digits), there are four essentially-different series starting with numbers less than 10^4: the ones generated by 196, 879, 1997 and 7059.
The code to check this is trivial, and at http://www.chiark.greenend.org.uk/~twomack/lychrel.cpp
I've extended the search to starts up to 10^7, and found a little under 2000 routes; there are many more route-starts beginning with the digit 1 than with any other digit. The program is fast enough that it's not worth downloading the tables of route-starts.
http://www.xs4all.nl/~itsme/projects/math/196/ba se 2.html
has the observations required to show that 10110_bin *never* forms a palindrome in base 2.
Your comments about IA64 seem to make sense, but
unfortunately for your thesis about RDRAM, the IA64 systems that Intel is producing [which were supposed to come out before the P4] use two or four channels of normal PC133 SDRAM to provide their memory subsystem.
The IA64 cooling and power requirements make the P4 look like an embedded chip; you need a housebrick-sized "power pod" per processor, and the case designs I've seen use half a dozen large fans.
Why would I buy a pc that I CANT upgrade (or not easily) to play my games on, when I can buy a real PC, play my games, browse the web, write papers, code, and upgrade HOWEVER and WHENEVER I want
Because an X-Box costs $300 and a PC with monitor and ISP connections costs $800 plus monthly ISP charge.
What this article tells us is that the corporations are asking for hideous restrictions on the use of their works.
Go one better. Stop using the things. In the words of your grandfather, make your own entertainment - and license it the way you want. You won't make the sort of money the Spice Girls have made; you won't get the exposure, but at least you can think of it as adding to the resources of humanity.
The goods are tied to the license; if you take the goods and leave the license, you're committing theft. So leave the goods and leave the license also; stick to, and add to, the pool of liberally-licensed content out there.
Did RMS distribute someone else's editor or compiler against their license conditions? Hell no! He wrote emacs and gcc, then used the rights of a creator to license them pleasantly.
At the moment, Intel's price is $990 each and AMD's is $1299; it's pretty much unprecedented to see AMD charging a premium over Intel, especially for what is not as good a chip except for FP work.
Neither of those prices bear any relation to the cost of manufacturing the chips; I have a feeling that the yield at these speeds (at least for Intel) is not high, and that these might even be loss-leaders ("I've got a GHz chip. OK, we get three working chips out of each $5000 wafer, but think of the press release").
I expect dual 1GHz Athlons to be in the same position this Christmas that dual Celerons are now, and by this time next year I hope to be using a dual 1500MHz Willamette box.
Slashdot Mirror
Firefox NoScript is the answer to all this kind of stupidity; I think it's worth using Firefox for NoScript alone.
Also, remind your hosts file that intellitxt.com is a synonym for 127.0.0.1
Yes, this is depriving hexus of advertising revenue. If they want advertising revenue, they should produce adverts which do not deeply infuriate their readers. Intelligently-targetted intellitxt might be actually usable, but to have every occurence of 'computer' hyperlinked to Dell's store is of no use to anybody.
Basically, the mission hasn't yet succeeded, and it doesn't seem to be completely certain that it will.
/ ExperimentError.pdf) has a one-sigma error of 100 milli-arcseconds per year, significantly larger than the relativistic effect and significantly larger than the effect from the motion of the target star through space. The initial expectation was for an error budget of less than 0.5 mas per year, so there was a lot of work done on measuring the proper motion of the star to that precision.
The goal was to measure the frame-dragging effect of the Earth, which is of the order of 40 milli-arcseconds per year; the current calibration (http://einstein.stanford.edu/content/aps_posters
The problems turn out to be really crazily subtle issues in solid-state physics -- the deposited metal films on the gyroscope and on its housing retain charge in patches large enough that they have to be modelled rather than averaged out -- plus an annoying issue from classical mechanics where the motion of a rigid body around three axes XYZ depends on the ratio of the differences of the moments of inertia X-Z and Y-Z. Whilst the gyroscopes are absurdly precisely made, so the moments of inertia are very close, the ratio of the differences of the moments of inertia is a macroscopic number; this is the same effect, and even a similar cause, to some of the oddities with low-precision floating-point arithmetic.
They'll probably be sorted out, sigma might be reduced by a factor ten after another year's modelling effort, but it seems unlikely that they'll get it down by the factor 200 they were hoping for.
The frame-dragging has already been measured indirectly by looking at the flickers of X-ray sources caused by frame-dragging-induced precession of the accretion discs around black holes, and most of the theories that suggest relativity is wrong would suggest that any oddities would be more pronounced in the huge gravitomagnetic fields near black holes than in the tiny fields near a mass as small, as non-dense and as slowly rotating as Earth.
This isn't 'installing Linux renders the warranty void from that point on', which would clearly be ridiculous and probably illegal, it's 'we can't troubleshoot hardware under Linux'.
... but I could imagine that the average user would not be very happy with a report 'hardware worked on our test bed, $50 charge for returning a non-broken machine'.
Which I think makes some degree of sense; I caused quite a lot of fuss to a local computer supplier when I returned a machine as unable to run stably under Ubuntu 6.10, they changed parts and finally got it to a state they thought reasonable, I found that even then it was totally unstable when I ran large disc-intensive jobs on it, and then installed a Feisty beta and found it's worked perfectly ever since.
Driver bugs look enough like hardware faults that three Canonical employees who I happened to know reckoned the issue was hardware; that a modern laptop would use devices for which the drivers in standard Linux distributions are buggy is not unexpected.
OK, HP are large enough that they could easily use a service model where they put a known-good hard drive with test tools on into the machine, run the test tools on that, and test the machine's current hard drive as the second drive in a machine known-good everywhere else, which is clearly the right thing to do
http://www.dramx.org/#fspot gives a decent idea of pricing for the underlying chips; it's just over a dollar per gigabit at the moment, so before considering assembly charges, profit and the like you'd expect a 64G drive to cost about six hundred dollars. I'd expect them to be about a thousand dollars once all the overhead in assembling systems from chips is added in, and rather more than that initially when they've got novelty value.
...
Disc space hit eight dollars a gigabyte in about June 2000, and a dollar a gigabyte about last year, so give it five years
I believe the current figure is something more like a hundred thousand writes per memory cell; with a load-distributing file system, this means the lifetime is about the amount of time taken to write the whole device 100k times, which (since flash writes are fairly slow; a few megabytes per second at best) is up there with hard-drive lifetimes.
... if you look at the spec sheet http://www.samsung.com/Products/Semiconductor/NAND Flash/SLC_LargeBlock/16Gbit/K9WAG08U1A/ds_k9xxg08u xa_rev11.pdf (linked to from an HTML summary at http://www.samsung.com/products/semiconductor/NAND Flash/SLC_LargeBlock/16Gbit/K9WAG08U1A/K9WAG08U1A. htm ) for a Samsung flash chip which is probably the one in these drives, you'll observe that it has 256Mbits of spare capacity for that sort of purpose. Doing reads and writes in units of a whole sector means you can use quite efficient error-correcting codes.
Flash can easily be made to have ECC and bad-sector remapping
America had 32,000 warheads in 1966, but has been dismantling them ever since, and now has just under ten thousand, of which about six thousand are operational.
For all practical purposes this is 5900 more than enough.
I am not sure what the ratio between uranium and plutonium in warheads is; I had the impression that, whenever the choice was available, you used plutonium because it's much cheaper to make plutonium in reactors than to make enriched uranium in centrifuges.
... you can ship material between distributed clusters of centrifuges in a truck, whilst reactors and reprocessing plants are unavoidably huge facilities that show up well on satellite imagery.
... you'd have to do isotope-separation, which is incredibly expensive.
Iran and North Korea started off with plutonium programs and then moved to use uranium instead, because it's easier to hide centrifuges than reactors and reprocessing plants
You convert the plutonium to the oxide and blend it down with natural uranium; the idea is then that, once the fuel has been in a reactor for a while, the fission of the uranium has caused enough of the non-239 plutonium isotopes to be produced that you don't get bomb-grade plutonium by separating out the plutonium fraction
(the problem with the other plutonium isotopes is that they have short half-lives and decay patterns that produce lots of neutrons; this imposes lots of engineering problems when handling the material, neutron radiation being even more inconvenient to deal with than other sorts, and means that a bomb is likely to detonate too early in the implosion process and give a much lower yield than expected)
No part of that post is in any way true.
The fission detonator for a thermonuclear weapon has to be of at least city-damaging proportions - several kilotons of yield - to get the X-ray flux up enough to ignite the fusion fuel. It's a real atom bomb, it requires several tens of kilograms of U235 or Pu239, neither of which is found in any fire detectors I've ever seen.
That's already being done, but there aren't all that many weapons; ten thousand warheads at 20kg each is about two thousand tonnes of fissile material (quite a lot is plutonium, to burn which you need a special mixed-oxide-burning reactor), whilst the known reserve of uranium is estimated at 3.6 million tons.
Err, that's an interesting thing to be taught; the core of the Earth is a sphere of liquid iron. Uranium isn't a siderophile (that is, it doesn't dissolve in liquid iron), so there won't be much uranium in the core (this also means there won't be much uranium in asteroids, in case space enthusiasts want to mention mining those for the uranium).
People have measured the uranium content of the inside of the Earth by looking for neutrinos of the right energy, which are produced during radioactive decay and fly straight through the Earth, and get that the quantity of uranium is enough to produce about 40 terawatts by radioactive decay. There is a crank theory that the core of the Earth has a fission reactor in it, but there's really very limited evidence for that.
Breeder reactors can also convert U238 to fissile plutonium, which is if anything more useful, since we already have reactors designed for Pu239 and I don't believe any reactors have yet been designed for Th233.
The problem is that people paranoid about nuclear proliferation have successfully made it very politically difficult (it's not technically completely straightforward, you're running rather fiddly chemistry by remote-control in a very high radiation environment) to reprocess spent fuel to get the plutonium out for reuse.
So the current nuclear fuel cycle is the equivalent of running a basic oil refinery, taking out the small jet-fuel fraction from crude oil, and then pumping the remainder back into the ground in places deliberately chosen to make it hard to take it out again. Breeder reactors are the equivalent of those catalytic-cracking columns in refineries which can make something useful out of the heavier crude-oil fractions.
The interesting feature of this announcement is how little computation and how much intelligence in software development was involved by the standards of other large computational projects. The calculation took three days on SAGE, which is an eight-socket dual-core Opteron system with 64GB of memory; it's perhaps three orders of magnitude less calculation than the factorisation of RSA200, or than IBM's work modelling hafnium silicates for developing 45nm processes. It is very much less work than is routinely done commercially for chip simulation or seismic inversion.
On the other hand, even if some of the tricks they used were fairly routine (you have a reasonable idea how large the coefficients are? The coefficients are obtained only by multiplication and addition? Why not calculate modulo lots of coprime one-byte integers and save a factor eight in storage space?) it's remarkably clever work.
http://atlas.math.umd.edu/kle8.narrative.html
and
http://atlas.math.umd.edu/kle8.html
are a description of the project aimed at reasonable mathematicians, with a lot more in them than the press release; but I think this item is mostly useful to tell mathematicians how to write a press release which gets picked up about what is fairly abstruse work by the standards of computational group theory.
The commercial market for large-scale computers exists, it just doesn't advertise itself on the top500 list. Banks have big clusters for monte-carlo work (why do you think the most-hyped benchmark application when Intel talks about eight-core Clovertown systems is an asset-pricing model); oil companies have enormous clusters for seismic work.
But both of those are intrinsically parallel jobs, so the clusters don't need the interconnect to run linpack, and there's no advantage to BP or Barclays in appearing on the top500 list -- they're not looking for the kind of researcher who picks his employer by acreage of computers, whilst Aldermaston and the Met Office may well be.
One of the embarrassing truths of computation is that an awful lot of jobs are either small enough to run on a single PC or too large to run on anything; there aren't that many things (climate modelling's the obvious one) constrained by amount of computation.
http://pluto.jhuapl.edu/science/data_collection.ht ml says that the transmission is at 38kbit/second from Jupiter, and will be at around 450bit/second from Pluto.
Cassini runs at 82kbit/second from Saturn, but it's a probe with a larger power budget.
The imager takes one-megapixel, 16bpp images, and compresses them to 100kbyte files for initial transmission, saving the originals in a few gigabytes of onboard flash; it can be instructed to send back uncompressed images if there's something interesting visible.
So an image takes about 20 seconds to transmit, plus about six minutes if you want the uncompressed version; and it takes 45 minutes to get to Earth from Jupiter. From Pluto, the images will take half an hour for the preview and twelve hours for the uncompressed image.
Did anyone else notice the very quick changes from white to red of the colour of the squid?
... apparently they're a good deal more placid when not impaled on pointed hooks (who'd have thought it).
0 3-10-2005/cover/Article.cover_story
... getting a sufficiently quick and sufficiently quiet robot to observe them is a better approach to seeing them in the wild, but there's been a lot of work put into that, and it didn't work.
Humboldt squid ('diablo rojo') are red when annoyed, and have a fearsome reputation since people wanting to dive with them generally dive from ships in the Bay of California squid-fishing fleet, under which circumstances the squid are understandably usually annoyed
http://www.montereycountyweekly.com/issues/Issue.
So, if architeuthis behaves like the Humboldt squid, the behaviour of this one at the surface is unlikely to be typical
It is irritating that water is really only transparent in visible wavelengths, so you can't do the trick that works nicely for land mammals of observing in light that the squid can't itself sense; they may well be better at detecting our lights and sonar than our own cameras are.
That's why Intel's also working on better buses (arrays of silicon lasers), on 3D designs that put a die of level-3 cache directly under the die of CPUs.
? ilc=125-6&ilg=english&isort=1&is=%3CISEARCH%3E&ip= no&itrack=+&itech_topic=+&itarget_audience=+&idate =+&isession_id=&iabstract=
http://www28.cplan.com/cv125/sessions_catalog.jsp
and look at the TCRS* series of talks (user 'idf', password 'fall2006')
It is entirely conceivable, made more so by the enormous Chinese reticence to publish the SMS4 encryption algorithm they're using and to open it to international review.
AES versus a Chinese government-approved algorithm which you can only get a specification for by agreeing to partner with one of eleven Chinese firms is not a difficult decision.
Well, it's probably a better use of $57 billion a year than the standard Japanese economy-boosting habit of building enormous public works; seven billion 1988-dollars for the Honshu to Hokkaido undersea tunnel (longer than the Chunnel), for example.
With all the current focus on China, people forget that Japan has (in dollar terms; the CIA World Factbook figures use slightly dubious purchasing-power-adjusted figures) the second largest economy in the world. It's an economy in a deep recession, but huge government spending is a traditional way out of those.
But presumably the "access denied" error message from an access point doesn't say "Access denied because your MAC isn't 00-20-E0-31-41-59", and there are enough MACs that picking ones at random to see if they happen to be allowed won't give you success at all quickly.
The integrated memory controller doesn't need to be too closely coupled to the CPU core; as far as I can see, it needs only a bus to tell it which location to handle next, and a connection to the L2 cache to put the new data in.
So it shouldn't be too difficult to graft in a new memory controller if a new memory technology becomes necessary; and if you have to replace the CPU to take account of new memory tech, so much the better for the CPU manufacturer.
No. As far as I can tell (I admit I only used long-longs, so interesting things might happen after 18 digits), there are four essentially-different series starting with numbers less than 10^4: the ones generated by 196, 879, 1997 and 7059.
l .cpp
a se 2.html
The code to check this is trivial, and at http://www.chiark.greenend.org.uk/~twomack/lychre
I've extended the search to starts up to 10^7, and found a little under 2000 routes; there are many more route-starts beginning with the digit 1 than with any other digit. The program is fast enough that it's not worth downloading the tables of route-starts.
http://www.xs4all.nl/~itsme/projects/math/196/b
has the observations required to show that 10110_bin *never* forms a palindrome in base 2.
Your comments about IA64 seem to make sense, but
unfortunately for your thesis about RDRAM, the IA64 systems that Intel is producing [which were supposed to come out before the P4] use two or four channels of normal PC133 SDRAM to provide their memory subsystem.
The IA64 cooling and power requirements make the P4 look like an embedded chip; you need a housebrick-sized "power pod" per processor, and the case designs I've seen use half a dozen large fans.
Why would I buy a pc that I CANT upgrade (or not easily) to play my games on, when I can buy a real PC, play my games, browse the web, write papers, code, and upgrade HOWEVER and WHENEVER I want
Because an X-Box costs $300 and a PC with monitor and ISP connections costs $800 plus monthly ISP charge.
Tom
No.
What this article tells us is that the corporations are asking for hideous restrictions on the use of their works.
Go one better. Stop using the things. In the words of your grandfather, make your own entertainment - and license it the way you want. You won't make the sort of money the Spice Girls have made; you won't get the exposure, but at least you can think of it as adding to the resources of humanity.
The goods are tied to the license; if you take the goods and leave the license, you're committing theft. So leave the goods and leave the license also; stick to, and add to, the pool of liberally-licensed content out there.
Did RMS distribute someone else's editor or compiler against their license conditions? Hell no! He wrote emacs and gcc, then used the rights of a creator to license them pleasantly.
Tom
At the moment, Intel's price is $990 each and AMD's is $1299; it's pretty much unprecedented to see AMD charging a premium over Intel, especially for what is not as good a chip except for FP work.
Neither of those prices bear any relation to the cost of manufacturing the chips; I have a feeling that the yield at these speeds (at least for Intel) is not high, and that these might even be loss-leaders ("I've got a GHz chip. OK, we get three working chips out of each $5000 wafer, but think of the press release").
I expect dual 1GHz Athlons to be in the same position this Christmas that dual Celerons are now, and by this time next year I hope to be using a dual 1500MHz Willamette box.