9) Develop a suicidal immigration policy that keeps out educated, hardworking men and women from friendly nations and, instead, takes in vast numbers of angry, uneducated immigrants from nations that hate us. This, too, leads to the shrinking of our knowledge base and the eventual disappearance of social cohesion.
This person has no clue what an H-1B is.
If you flood the simple jobs, then it will encourage people to get educated. But, if a programmer has to work at McD's because an H-1B took his/her job, then there is little incentive to be a programmer anyhow. Why is flooding skilled labor better than flooding unskilled labor?
See also my earlier comments (yes, I'm a British H-1B holder). I think he knows very well what an H-1B is, and just how difficult getting a green card is. I most certainly know.
I'll reply to your question by repeating my earlier comments. It isn't just about internal competition. You don't stop competing with foreign engineers for jobs just because they're in another country. The most pressing problem for the US engineering industries (and by extension, US engineering jobs) isn't how much competition there is in the US. Exact figures are difficult to nail down, but Chinese universities are currently turning out (ballpark) 600k EEs a year, compared to US universities (ballpark) 60k. The biggest issue is whether team US loses out to team China. If team China wins, many more US engineers are going to be flipping burgers.
Decide which team you're on, and start playing for the team. China is coming to get us, and they mean to win. Worry about that first.
I really have to respond to this. I am, I must say, pretty sick of seeing this same argument trotted out every time somebody talks about the future of American engineering industries. And yes, before I get started I'll let you know that I'm one of the many engineers (British, in my case) working in the US, currently on an H-1B visa (green card application in progress, albeit paaiiiinfullllyyy slow progress).
First in the "Hall of Shame" is the H1-B Visa bill that is importing mass quantities of foreign workers to compete for US jobs
ok, I'll start with this. Never mind that this point is contradicted later on (I'll get to that in a minute), but what is it that makes you think that foreign engineers won't be competing with American engineers just because they're in another country? People, this is a global market. Get used to it. And if you want to see good engineering jobs in the US, make sure first of all that America wins. The H-1B program is abused by some companies, for sure (but you should know that there are measures in place to catch and take action against those who do) - but the rules set a pretty high bar for entry, and present the only way that it is possible to immigrate into the US. Coming from the UK, I've seen exactly who gets hit worst by having the best engineers getting sucked out of one country and deposited in another (hint - it isn't the US - many UK high-tech industries were heavily damaged in the '70s by what was known as the "brain drain". Some would argue that the UK has never fully recovered).
These workers must keep their jobs, or they are deported. "You don't mind working 80 hour weeks do you? I thought not...". Plus 40K a year is a fortune to many of these folk!
Ugh. This kind of FUD is particularly offensive (not least to my intelligence). As you might guess, I know a large number of other aliens currently working on H-1B visas, and what you're saying here doesn't match my experience.
Firstly, transferring H-1Bs isn't hard. It isn't at all hard. I know people who move jobs without any difficulty. Processing the paperwork is routine and takes about a month if you have a good lawyer. Compared to the time required to get a visa in the first place, or to renew a visa, or to get a green card, it's a triviality. I honestly don't know anybody who worries about this.
Secondly, the requirements for H-1B approval place lower limits on the salary you can set. I don't know of anybody on an H-1B who is getting paid anywhere near as low as $40k. You may know different - but remember that if any employer is paying below the going rate, or using the fear of deportation, they are abusing the rules. Complain about them if you want to - in fact, I'd urge you to do so - but please don't assume that this is the general case.
This whole story about being treated as slave labor is repeated so often many people seem to believe it's true. I have seen many on Slashdot swear it's true. But speaking from experience, it just doesn't square with the facts. It's also an abuse of the rules, if you think anybody cares (admittedly, unfortunately, debatable). But I have to say this really pisses me off more than anything. It's just plain insulting to keep on trotting out this BS which seems to imply that all foreigners are stupid and happy to be treated as slaves.
Second is the abysmal rate at which we're turning out math, science and engineering graduates in the US. The US (if memory serves correctly) will turn out about 50,000 EE grads this year, of which 2/3 are foreign citizens likely to leave with their new expertise at some point. Mainland China alone will graduate over 600,000 EE grads this year, of which the vast majority are Chinese citizens. Also, in 2001 Mainland China opened ten shiny brand new software engineering universities. India is another large counry with a vast potential tech workforce.
This is the contradiction I was talking about, and speaks to the point I made earlier. The big picture is that if China (for instance) is able to accrue an engineering workforce of scale and caliber that greatly surpasses the US, the American engineering industry is in deep shit. I'll repeat the comment I made at the start:
If you want to see good engineering jobs in the US, make sure first of all that America wins
If you want to make sure this happens, stop griping about how much competition there is for jobs internal to the US jobs market and worry about how strong the US industry is relative to the rest of the world. I have a Stars and Stripes flying outside my house. Whose side are you on?
No, don't think so. I'm rather skeptical of the claims for an "isochronous network" having better latency than a CSMA protocol. This is actually very well researched and I haven't yet seen anything that shows this kind of lower latency.
Of course, I'm guessing as to what protocol they're using, because there are no details here, but I'd guess that by "isochronous" they mean a TDMA scheme. If this is true, then I would have to disagree with the assertion of isochronicity - and I can't see how else this could have been done. Others may disagree, but the only definition I know of "isochronous" has little to do with the protocol by itself, but is about the system. A phone (that uses a TDMA channel access protocol) is not isochronous because it uses TDMA, but because the source voice codec is clock-locked to the underlying protocol. That is, the voice codec produces bursts of data at precisely the time that the allocated slot comes around. This is what gives the system the low latency.
Translate this into something where the data source is asynchronous (such as buttons being pressed on a controller) and the relation goes away. Now, when the data turns up at an uncontrolled time, it has to wait for its TDMA slot to come around before it can send. Worse, if the first try is corrupted (and this happens a lot in WLANs, btw) it has to wait for the next slot to come around to have a go. Compare this with CSMA, where you can send the data as soon as it arrives, and if it fails, have another go right away, and you actually get lower latency than a typical TDMA scheme, all things being equal.
There's a good deal of data to support this. See, for instance:
The thing about collisions is pretty much a red herring. For sure, contention based schemes lose packets to collisions, but for most wireless LAN physical layers, this rate of packet loss is dwarfed by what you lose to vagaries of the medium. Either way, you end up with the need to retransmit failed packets, and it is an important measure of the performance of the protcol as to how well it can deal with this.
The 802.11 MAC is CSMA/CA, rather than CSMA/CD - the difference being that the wireless LAN can't detect collisions, so it tries a little harder to avoid them.
Be careful. Everything is not as simple as it seems....
First, actual throughput doesn't reflect the "nominal" speed on the box. Anybody who has used 802.11b will know that the most you actually get is about 6Mb/s (and that's under very ideal conditions). This matters in your choice of 802.11b,.11g or.11a., because the difference between "nominal" and "actual" are very different in each case.
The reasons for the deviation are twofold:
1. MAC & PHY layer overheads. 802.11 has much higher intrinsic overheads like inter-packet spaces and preambles than with traditional wired Ethernet. This means that for every packet there is a time on-air that isn't used for data. As the "nominal" data rate is pushed up, this "dead time" stays numerically the same, so becoming a higher proportion of the total. Bottom line: "22Mb/s" 802.11b isn't actually going to be anything like twice as fast. On the other hand, 802.11a is quite different - because all of the overheads have been shrunk; it isn't just a higher nominal rate. For instance, the preamble changes from 96us (802.11b) to 16us (802.11a). That stuff matters, big time.
2. Errors. Packets in a wireless LAN frequently get corrupted in transmission. Much of the complexity of the 802.11 protocol is detecting and retransmitting the corrupted packets. But of course, every time you do that, you're spending more bandwidth. Bottom line - as the channel deteriorates, meaning more packet errors, your actual throughput goes down. This matters here because 802.11b has only three channels, in a very polluted band (Bluetooth, microwaves, cordless phones, X10 cameras, and other 802.11b systems). This pollution is only going to get worse. On the other hand, 802.11a has eight channels (and there's potentially more to be allocated over the next few years) and the spectrum rules are built to deliberately disadvantage "narrow band" systems, making it effectively uneconomic to build cordless phones or garage door openers that use the band (and no microwaves, btw).
What many people here seem to be missing is, the question is not "wouldn't everybody switch to ogg vorbis" but "why wouldn't people stick with the version of mp3 they already have?"
Many people already have mp3 files, tools and players that do exactly what they want. Why would they "upgrade" to a deliberately crippled version that limits what they can do? To persuade people to upgrade, you have to provide them with something new of value that they didn't have before, not less.
I would have to disagree. There are huge differences between each of the processors you mention. Just because the instruction sets don't appear to change, doesn't mean the architectures are the same.
The big differences between the 8086/88, 286 and 386 are in the memory models. The 8086/88 can only address 1MB of memory based on its 20-bit address space. The 286 added a segmented memory model, which was largely judged to be a dead-end, but the 386 added a full paged memory model with the necessary MMU supporting a 32-bit address space. This is far from a trivial improvement and the whole basis of any useful platform for a multitasking OS.
The big change from the Pentium to the PPro architecture (which became P-II and P-III) is out of order execution. You can harly consider the Pentium and PPro to be "basically the same" when considering this. Even though the instruction set didn't change at this point, that a massive upgrade to the underlying architecture. The cores don't look even remotely similar.
There's a huge leap in the argument that says that if the US doesn't export cryptographic technology, nobody outside the US will be able to get hold of it.
Much encryption technology was invented outside of the US. That includes AES, NIST's replacement for DES, which is a cipher with 128, 192 or 256 bit keys. The algorithm was selected after a competition, and the winning algorithm was designed in Europe.
So how exactly does anybody believe that export restrictions from the US are going to work?
been there, done that. I'm a silcon design engineer and I've done silicon validation testing where you take a sample on a test rig and check its function at various extremes of temperature. The "cold" end of the test spec can actually be very cold (and I've designed some stuff that was designed to be used in an outdoor enclosure in Canada - chlly).
It's very easy when operating in a test environment that's significantly above the target temperature to cool the device to the point where you get significant condensation on and around the device under test and the whole test rig fries.
Not really - Sklyarov was an employee of a company that sold such software. I don't remember Sklyarov selling ripped-off copies of his employer's software in private transaction for his own profit.
Since when did employees (not company officers) become criminally liable for the actions of their employer?
agreed that you could run a low-performance CPU on this. By the way, it would be very low performance - we're talking 10um process on amorphous silicon. If your're thinking of getting something in the league of, say, a 6502, forget it. This isn't going to get close.
That might be good enough (and certainly fun) for sure. However, you still have the problem of low runs - I don't see anything in the description of this process that says it can do that. Just because it's "printing" doesn't mean that there aren't large tooling costs, and I suspect (but can't verify) that there might be.
This actually isn't going to be any use for processors, and it won't necessarily work for small runs either. But it is a very cool technology for different reasons, and the article does explain this.
Look at it this way - CPUs have become much cheaper over time because manufacturing processes have been developed to reduce the size of each transistor, allowing you to fit more on a wafer of a given size, whilst increasing the size of the wafers - and the cost per wafer has not increased at the same exponential rate as the transistor count per wafer. Since it is ok that the transistors are smaller, this is a good thing. However, this model doesn't work for things where you actually want the device to be big - like a flat panel display or (as the article says) an X-Ray panel (think detector in an airport that you walk through).
What this technology does is provides a compelling economic model for large items that isn't served by conventional manufacturing processes.
One more thing - this uses amorphous silicon, also a good thing for large items in that the absence of the need for crystallinity is a big help for yield - but at the same time don't expect it to break any speed records.
Bottom line - this is a really cool technology, but it isn't going to be used for one-off roll-your-own CPUs.
I know this is going to be unpopular, and I fully expect to be modded down for speaking up for Microsoft, but I'd like to make a plea for a little objectivity here.
A couple of questions: can you really put your hand on your heart and say that the choice of the name "Lindows" isn't a deliberate attempt to ride off Microsoft's name? Let me put that a different way: if Microsoft Windows wasn't called Windows, but something else, say "Microsoft Doors" do you really think that Lindows.com would still have chosen that name? Somehow, I don't think so. But you might disagree.
Secondly, examine the issue of handing over email addresses and so on for a minute. This issue is raised in the context of trying to start a campaign in defense of Lindows.com. Do you think this isn't an attempt to drum up sympathy? Why do you think Microsoft's attorneys (note - this isn't actually Microsoft, a subtle but important distinction) would have specifically asked for this data? Isn't it possible, or even probable, as part of the normal discovery process, that they would just ask for everything on their servers? That is, after all, the normal procedure. The reason it's called "discovery" is because they're asking for everything in an attempt to find what might be there - in other words, they don't know a priori what's on their servers, or in their filing cabinets. Why focus on this one piece of information? It could just as well be that Lindows.com have chosen to focus on the fact that personal details get scooped up in the normal discovery process, and publicise that as part of a campaign to get people on-side, and whip up some anti-Microsoft frenzy. Do you think that Microsoft's attorneys walked in and demanded the contact database, and nothing else? If not, why only mention the one set of data?
Just trying to apply some healthy skepticism here...
In the chipset maybe, not the chip. So you have to use a chipset that has the clock throttling feature. I'm reading the datasheets for the Athlon and Pentium 4 (you can get them from the relevant websites).
Go back to my original point. I wasn't making any point about the Athlon, rather asking how this is somehow a negative for the Pentium 4. So how is it?
What "malfunctions" are we talking about here?
And what's the deal with "overheating"? Neither of these has been reported, except for some ill-informed commentary on the thermal protection feature. Just to be clear:
1. If you run an Athlon and a Pentium 4 side by side, running the same application, the numbers I have seen say that the Pentium 4 dissipates less power, not more.
2. In the event that the temperature limits get exceeded, due to an inadequate thermal solution, the Pentium 4 thermal protection diode causes it to clock throttle to prevent it from blowing up. In the same situation, the Athlon fries itself.
I have seen many contorted arguments presented by the rabid Intel-haters who post here from time to time attempting to cast this as somehow a negative for the Pentium 4. Can somebody please explain to me how this can possibly be so? If not, can we please stop posting this kind of misinformed or plain wrong FUD
Was that max price or min price? The point here is that the contract above includes a minimum price. This was common at one time, but (and, of course, IANAL) I believe it is no longer legal in the UK. What this is basically saying is that discounting is not permitted by retailers. Until very recently there was an exception made for selling books in the UK, known as the "Net book agreement" (can somebody who still lives in the UK confirm that this has now gone?) which enforced the same controls. There was a great deal of debate about abolishing it, with the argument made very strongly that it would push small booksellers out of business, because they would be unable to compete with large bookstores who would now be free to discount their prices.
I accept what you are saying here, but it wasn't what I was getting at in my original (anonymous) post. What I was specifically answering was this:
That 400mhz figure is a fairytale... its a 100mhz buss that xmits 4 whatevers per clock cycle... They're flat out lying...
This seems to me to be a little strong, to say the least. Accepted that the transfers can be active or inactive, but when they're running, the word transfer rate is 400MHz. Sure you could put some additional explanation around that (and I think you will find it in Intel's statements, although I haven't looked).
Put it this way: for the purposes of a simple explanation, you need to say how many Hz it is. That's as much detail as most people want or need to know. What is that number?
Slashdot Mirror
This person has no clue what an H-1B is.
If you flood the simple jobs, then it will encourage people to get educated. But, if a programmer has to work at McD's because an H-1B took his/her job, then there is little incentive to be a programmer anyhow. Why is flooding skilled labor better than flooding unskilled labor?
See also my earlier comments (yes, I'm a British H-1B holder). I think he knows very well what an H-1B is, and just how difficult getting a green card is. I most certainly know.
I'll reply to your question by repeating my earlier comments. It isn't just about internal competition. You don't stop competing with foreign engineers for jobs just because they're in another country. The most pressing problem for the US engineering industries (and by extension, US engineering jobs) isn't how much competition there is in the US. Exact figures are difficult to nail down, but Chinese universities are currently turning out (ballpark) 600k EEs a year, compared to US universities (ballpark) 60k. The biggest issue is whether team US loses out to team China. If team China wins, many more US engineers are going to be flipping burgers.
Decide which team you're on, and start playing for the team. China is coming to get us, and they mean to win. Worry about that first.
minard
First in the "Hall of Shame" is the H1-B Visa bill that is importing mass quantities of foreign workers to compete for US jobs
ok, I'll start with this. Never mind that this point is contradicted later on (I'll get to that in a minute), but what is it that makes you think that foreign engineers won't be competing with American engineers just because they're in another country? People, this is a global market. Get used to it. And if you want to see good engineering jobs in the US, make sure first of all that America wins. The H-1B program is abused by some companies, for sure (but you should know that there are measures in place to catch and take action against those who do) - but the rules set a pretty high bar for entry, and present the only way that it is possible to immigrate into the US. Coming from the UK, I've seen exactly who gets hit worst by having the best engineers getting sucked out of one country and deposited in another (hint - it isn't the US - many UK high-tech industries were heavily damaged in the '70s by what was known as the "brain drain". Some would argue that the UK has never fully recovered).
These workers must keep their jobs, or they are deported. "You don't mind working 80 hour weeks do you? I thought not...". Plus 40K a year is a fortune to many of these folk!
Ugh. This kind of FUD is particularly offensive (not least to my intelligence). As you might guess, I know a large number of other aliens currently working on H-1B visas, and what you're saying here doesn't match my experience.
Firstly, transferring H-1Bs isn't hard. It isn't at all hard. I know people who move jobs without any difficulty. Processing the paperwork is routine and takes about a month if you have a good lawyer. Compared to the time required to get a visa in the first place, or to renew a visa, or to get a green card, it's a triviality. I honestly don't know anybody who worries about this.
Secondly, the requirements for H-1B approval place lower limits on the salary you can set. I don't know of anybody on an H-1B who is getting paid anywhere near as low as $40k. You may know different - but remember that if any employer is paying below the going rate, or using the fear of deportation, they are abusing the rules. Complain about them if you want to - in fact, I'd urge you to do so - but please don't assume that this is the general case.
This whole story about being treated as slave labor is repeated so often many people seem to believe it's true. I have seen many on Slashdot swear it's true. But speaking from experience, it just doesn't square with the facts. It's also an abuse of the rules, if you think anybody cares (admittedly, unfortunately, debatable). But I have to say this really pisses me off more than anything. It's just plain insulting to keep on trotting out this BS which seems to imply that all foreigners are stupid and happy to be treated as slaves.
Second is the abysmal rate at which we're turning out math, science and engineering graduates in the US. The US (if memory serves correctly) will turn out about 50,000 EE grads this year, of which 2/3 are foreign citizens likely to leave with their new expertise at some point. Mainland China alone will graduate over 600,000 EE grads this year, of which the vast majority are Chinese citizens. Also, in 2001 Mainland China opened ten shiny brand new software engineering universities. India is another large counry with a vast potential tech workforce.
This is the contradiction I was talking about, and speaks to the point I made earlier. The big picture is that if China (for instance) is able to accrue an engineering workforce of scale and caliber that greatly surpasses the US, the American engineering industry is in deep shit. I'll repeat the comment I made at the start:
If you want to see good engineering jobs in the US, make sure first of all that America wins
If you want to make sure this happens, stop griping about how much competition there is for jobs internal to the US jobs market and worry about how strong the US industry is relative to the rest of the world. I have a Stars and Stripes flying outside my house. Whose side are you on?
minard
Of course, I'm guessing as to what protocol they're using, because there are no details here, but I'd guess that by "isochronous" they mean a TDMA scheme. If this is true, then I would have to disagree with the assertion of isochronicity - and I can't see how else this could have been done. Others may disagree, but the only definition I know of "isochronous" has little to do with the protocol by itself, but is about the system. A phone (that uses a TDMA channel access protocol) is not isochronous because it uses TDMA, but because the source voice codec is clock-locked to the underlying protocol. That is, the voice codec produces bursts of data at precisely the time that the allocated slot comes around. This is what gives the system the low latency.
Translate this into something where the data source is asynchronous (such as buttons being pressed on a controller) and the relation goes away. Now, when the data turns up at an uncontrolled time, it has to wait for its TDMA slot to come around before it can send. Worse, if the first try is corrupted (and this happens a lot in WLANs, btw) it has to wait for the next slot to come around to have a go. Compare this with CSMA, where you can send the data as soon as it arrives, and if it fails, have another go right away, and you actually get lower latency than a typical TDMA scheme, all things being equal.
There's a good deal of data to support this. See, for instance:
http://grouper.ieee.org/groups/802/11/Documents/Do cumentHolder/1-525.zip
The thing about collisions is pretty much a red herring. For sure, contention based schemes lose packets to collisions, but for most wireless LAN physical layers, this rate of packet loss is dwarfed by what you lose to vagaries of the medium. Either way, you end up with the need to retransmit failed packets, and it is an important measure of the performance of the protcol as to how well it can deal with this.
The 802.11 MAC is CSMA/CA, rather than CSMA/CD - the difference being that the wireless LAN can't detect collisions, so it tries a little harder to avoid them.
There's no way any "open source hardware" that relies on FPGAs is ever going to touch custom silicon on price or performance.
Face facts. Hardware isn't the same as software, and an open source model isn't going to work.
First, actual throughput doesn't reflect the "nominal" speed on the box. Anybody who has used 802.11b will know that the most you actually get is about 6Mb/s (and that's under very ideal conditions). This matters in your choice of 802.11b, .11g or .11a., because the difference between "nominal" and "actual" are very different in each case.
The reasons for the deviation are twofold:
1. MAC & PHY layer overheads. 802.11 has much higher intrinsic overheads like inter-packet spaces and preambles than with traditional wired Ethernet. This means that for every packet there is a time on-air that isn't used for data. As the "nominal" data rate is pushed up, this "dead time" stays numerically the same, so becoming a higher proportion of the total. Bottom line: "22Mb/s" 802.11b isn't actually going to be anything like twice as fast. On the other hand, 802.11a is quite different - because all of the overheads have been shrunk; it isn't just a higher nominal rate. For instance, the preamble changes from 96us (802.11b) to 16us (802.11a). That stuff matters, big time.
2. Errors. Packets in a wireless LAN frequently get corrupted in transmission. Much of the complexity of the 802.11 protocol is detecting and retransmitting the corrupted packets. But of course, every time you do that, you're spending more bandwidth. Bottom line - as the channel deteriorates, meaning more packet errors, your actual throughput goes down. This matters here because 802.11b has only three channels, in a very polluted band (Bluetooth, microwaves, cordless phones, X10 cameras, and other 802.11b systems). This pollution is only going to get worse. On the other hand, 802.11a has eight channels (and there's potentially more to be allocated over the next few years) and the spectrum rules are built to deliberately disadvantage "narrow band" systems, making it effectively uneconomic to build cordless phones or garage door openers that use the band (and no microwaves, btw).
What many people here seem to be missing is, the question is not "wouldn't everybody switch to ogg vorbis" but "why wouldn't people stick with the version of mp3 they already have?"
Many people already have mp3 files, tools and players that do exactly what they want. Why would they "upgrade" to a deliberately crippled version that limits what they can do? To persuade people to upgrade, you have to provide them with something new of value that they didn't have before, not less.
I mis-typed. The 286 adds protected mode, not segmented mode
The big differences between the 8086/88, 286 and 386 are in the memory models. The 8086/88 can only address 1MB of memory based on its 20-bit address space. The 286 added a segmented memory model, which was largely judged to be a dead-end, but the 386 added a full paged memory model with the necessary MMU supporting a 32-bit address space. This is far from a trivial improvement and the whole basis of any useful platform for a multitasking OS.
The big change from the Pentium to the PPro architecture (which became P-II and P-III) is out of order execution. You can harly consider the Pentium and PPro to be "basically the same" when considering this. Even though the instruction set didn't change at this point, that a massive upgrade to the underlying architecture. The cores don't look even remotely similar.
Much encryption technology was invented outside of the US. That includes AES, NIST's replacement for DES, which is a cipher with 128, 192 or 256 bit keys. The algorithm was selected after a competition, and the winning algorithm was designed in Europe.
So how exactly does anybody believe that export restrictions from the US are going to work?
Has anybody tried hooking up a dehumidifier to the air inlet of an otherwise sealed box to make sure this doesn't happen?
It's very easy when operating in a test environment that's significantly above the target temperature to cool the device to the point where you get significant condensation on and around the device under test and the whole test rig fries.
Since when did employees (not company officers) become criminally liable for the actions of their employer?
That might be good enough (and certainly fun) for sure. However, you still have the problem of low runs - I don't see anything in the description of this process that says it can do that. Just because it's "printing" doesn't mean that there aren't large tooling costs, and I suspect (but can't verify) that there might be.
Look at it this way - CPUs have become much cheaper over time because manufacturing processes have been developed to reduce the size of each transistor, allowing you to fit more on a wafer of a given size, whilst increasing the size of the wafers - and the cost per wafer has not increased at the same exponential rate as the transistor count per wafer. Since it is ok that the transistors are smaller, this is a good thing. However, this model doesn't work for things where you actually want the device to be big - like a flat panel display or (as the article says) an X-Ray panel (think detector in an airport that you walk through).
What this technology does is provides a compelling economic model for large items that isn't served by conventional manufacturing processes.
One more thing - this uses amorphous silicon, also a good thing for large items in that the absence of the need for crystallinity is a big help for yield - but at the same time don't expect it to break any speed records.
Bottom line - this is a really cool technology, but it isn't going to be used for one-off roll-your-own CPUs.
A couple of questions: can you really put your hand on your heart and say that the choice of the name "Lindows" isn't a deliberate attempt to ride off Microsoft's name? Let me put that a different way: if Microsoft Windows wasn't called Windows, but something else, say "Microsoft Doors" do you really think that Lindows.com would still have chosen that name? Somehow, I don't think so. But you might disagree.
Secondly, examine the issue of handing over email addresses and so on for a minute. This issue is raised in the context of trying to start a campaign in defense of Lindows.com. Do you think this isn't an attempt to drum up sympathy? Why do you think Microsoft's attorneys (note - this isn't actually Microsoft, a subtle but important distinction) would have specifically asked for this data? Isn't it possible, or even probable, as part of the normal discovery process, that they would just ask for everything on their servers? That is, after all, the normal procedure. The reason it's called "discovery" is because they're asking for everything in an attempt to find what might be there - in other words, they don't know a priori what's on their servers, or in their filing cabinets. Why focus on this one piece of information? It could just as well be that Lindows.com have chosen to focus on the fact that personal details get scooped up in the normal discovery process, and publicise that as part of a campaign to get people on-side, and whip up some anti-Microsoft frenzy. Do you think that Microsoft's attorneys walked in and demanded the contact database, and nothing else? If not, why only mention the one set of data?
Just trying to apply some healthy skepticism here...
In the chipset maybe, not the chip. So you have to use a chipset that has the clock throttling feature. I'm reading the datasheets for the Athlon and Pentium 4 (you can get them from the relevant websites).
Go back to my original point. I wasn't making any point about the Athlon, rather asking how this is somehow a negative for the Pentium 4. So how is it?
What "malfunctions" are we talking about here?
And what's the deal with "overheating"? Neither of these has been reported, except for some ill-informed commentary on the thermal protection feature. Just to be clear:
1. If you run an Athlon and a Pentium 4 side by side, running the same application, the numbers I have seen say that the Pentium 4 dissipates less power, not more.
2. In the event that the temperature limits get exceeded, due to an inadequate thermal solution, the Pentium 4 thermal protection diode causes it to clock throttle to prevent it from blowing up. In the same situation, the Athlon fries itself.
I have seen many contorted arguments presented by the rabid Intel-haters who post here from time to time attempting to cast this as somehow a negative for the Pentium 4. Can somebody please explain to me how this can possibly be so? If not, can we please stop posting this kind of misinformed or plain wrong FUD
Was that max price or min price? The point here is that the contract above includes a minimum price. This was common at one time, but (and, of course, IANAL) I believe it is no longer legal in the UK. What this is basically saying is that discounting is not permitted by retailers. Until very recently there was an exception made for selling books in the UK, known as the "Net book agreement" (can somebody who still lives in the UK confirm that this has now gone?) which enforced the same controls. There was a great deal of debate about abolishing it, with the argument made very strongly that it would push small booksellers out of business, because they would be unable to compete with large bookstores who would now be free to discount their prices.
Thanks for posting the link to the inflation calculator... I didn't have such a thing at hand when I posted the article.
That 400mhz figure is a fairytale ... its a 100mhz buss that xmits 4 whatevers per clock cycle ... They're flat out lying ...
This seems to me to be a little strong, to say the least. Accepted that the transfers can be active or inactive, but when they're running, the word transfer rate is 400MHz. Sure you could put some additional explanation around that (and I think you will find it in Intel's statements, although I haven't looked).
Put it this way: for the purposes of a simple explanation, you need to say how many Hz it is. That's as much detail as most people want or need to know. What is that number?