That is what they are built for, so if it isn't, NVidia have deeply failed.
The CUDA software takes a little getting your head around - but not much. It is basically plain vanilla C (not C++) with vector processing. A single $400 8800 GPU board matches manufacturer-that-I-used-to-work-for's $50,000 hardware at evens. it isn't faster, but it isn't slower either. Two would be twice as fast.
It takes 4 cycles to do a floating point operation, and 4 cycles to do an integer add/subtract. It takes 16 cycles to do an integer multiply because it only has a 24-bit hardware multiplier (needed to achieve the 4-cycle flops, so it has to do long multiplies as four madds, This was for the first generation CUDA CPUs; the second generation, which should be out by now, was going to have double length floating and would be able to do 32 bit multiplies in the same four cycles.
While they can do integer, these machines are not very happy with it, and I found it much easier to do everything in floating point, even if you are talking about 8-bit colour data. It goes no slower, and everything is much better adapted to floating point. Then there are special instructions to get back to integer at the output.
While each operation takes 4 cycles, they are fully pipelined, so that it launched a new instruction per cycle, times 32 pipes per unit, times 8 units per GPU.
And madd is very useful for the sort of tasks for which supercomputers are traditionally used.
From my experience, you can buy support for any FOSS package worth mentioning, at a price that still beats commercial rivals. It is also my experience that the support thus purchased is outstandingly better than that for paid-for software. Problem responses within four hours, from somebody who really understands the system, instead of taking to weeks to dig through layers of ignoramuses to get to the expert. This, I conjecture, is because FOSS support teams live or die by the quality of support, whereas paid-for software put the best developers onto new features and regard support as very much a second-line function.
And, of course, many of the early car repair and servicing [places (and the early cars needed a lot of repairs) were the blacksmiths, because repairing a car was basically a matter of metalwork. Into the '70 s a lot of small village garages in the UK were recognisably descended from the village blacksmith. The farrier, however, did lose out.
Blacksmiths, buggy whip makers and all the other usual old time jobs that Slashdotters trot out each and every time they wish to denigrate a business case did not face competition from their own product being hawked with no requirement for any return on investment. Yes they did: the product was a means of transport. Few of the customers actually wanted horse, whips etc. they wanted transport. Cars are a different implementation of that product. By your definition, Ford are not competing with GM because Ford's product is different from GM's.
What is actually being supplied is the ability to listen to nice noises. The transport layer is irrelevant.
Not Boeing, composites. Rutan has made a significant number of aircraft using composites on large scale. However, none within sight is the size of a B787, few intended for large-scale production, and none intended for the 365 days a year utilisation of a commercial airliner.
I may be unselective, but while I don't find every woman beautiful, I have no trouble finding pictures of beautiful women without the assistance of a computer.
In the name of God, WHY? The whole point of being a man (or a gay woman etc.) is so you can make your own estimation of women's beauty. One of the things I definitely DO NOT need a computer to do is to look at women for me. Many other things I may try and get computers to do for me - some sensible, some not. But I'll look at my own women, thank you very much,
Parallelism was the New New Thing when the Inmos Transputer rolled out in 1984 - a CPU explicitly designed to allow multiple CPUs to co-operate, and with a hardware scheduler to provide on-chip parallelism. Then we had the GAPP (Generalised Arithmetic parallel processor), the Connection Machine, and lots of other weird architectures whose names I cannot recall. I designed one myself in the late eighties, and we took it to breadboard stage (essentially Hyperthreading write very large).
Forgive me for not getting too excited by this new dawn. I have seen too many false dawns. Yes, parallelism is getting more and more prevalent, but expect more incremental steps, not an explosion.
There have been som studies funded by the UK Department of Health which showed no convincing results from cellphones. Now, they migh, of course, be in the pocket of the cellphome manufacturers, like they might be in the pocket of the drug manufacturers. But with a socialised health service, they are the people who are going to end up payiong the first level, purely medical, costs of any ill effects that there may be. Whcih suggests to me that, if they were going to err, they would be likely to err in the direction of overcaution rather than recklessness.
All the accusations against cellphones have been generally anecdotal i.e. a number of people have been found who were both heavy cellphone users and got brain tumours. But when large scale statistical studues are done, these "clusters" disappear. If you ask averybody with a tumour whether they were a heavy cellphone user, some will say yes. Probably more than really are, becasue moderate users will tend to judge themselves heavier in order to have something to blame for their tragedy - randomness seems much more frightening that a technological accident.
I don't think it is company vs. individual, it is main focus vs peripheral interest. If looking after your data is either the person's or the company's priimary task, you are much more likely to get good, focussed service from that individual. If your data is only one of 47 differnt things the company, or the indiividual, does, and is not regarded as the main one, then inevitably you will get poor service, and may be dropped or otherwise inconvenienced when they regard other aspects of their business as more important.
Obviously, if you employ someone, it is easy to tell them that their primary job, on which their salary depends, is storing and searching your data. But you can say the same about a company whose sole focus is managing other people's data: if they screw you around, you can werck them by telling the truth about their service. That would not be true if your data were stores by some giant such as Google or Amazon whose main business is elsewhere.
A small, hightly specialised, company is likely to perform as well as, or better than, your single employee, but with a higher truck number. But keep an eye on their financials - if they go bust you could be truly screwed. It is in your interest for them to be profitable. And what you do when tney get takenover by Google, I don't know. But what you do when your one in-hous expert leaves is no easier.
The articla assimes that any monopolist or near monopolist is going to get hated just for that fact. And, with DoubleClick, Apple is going to be dominant in both inline and display advertising. Therefore, he concludes, Apple will get hated. But that dominance per se does't wory me, the sucker^Wconsumer. I don't care who isgenerating the ads, I only care if they are invasive - which has no relationship to who owns them. The adverisers may come to hate Google's dominance of web-based adverting. But why should I cry for the advertisers? Indeed, if you believe they have a fixed ad budget, then the more Googel gouges them the fewer ads they can afford - which is great.
Of course, I am not saying that Google cannot get hated - it already is, a little bit. But if they keep their current policy of keeping opt-out easy, there is always a quick answer "then walk, dummy". The things that have made IBM, Microsofs and (looking upwards) Apple hated - things that, intentionally or otherwise, make it difficult to pick-and-mix your hardware and software options. As long as GMail doesn't lock me into uusing Google for search or vice versa, or I find my GMail locked in so I cannot load it into something else, or I cannot use Google without giving all my security information, it is difficult to see a real head of hate building up. Yes, Google is already losing that fresh and innocent look - but it can (if it wants to) avoid becoming actively hated, IMO.
Only if voters allow it - even if only by default. Again, the problem is political. We should be holding the politicians much more to account. But if you have a two party system, wealthy companies have only two targets to focus lobbying on, and persuade both to do what the companys want. This leacves the voters no way to tell the politicans to clean up their act. A system which supports only two (or othr small number) of immovable parties is inherently corruptible: the rich can seriously contemplate buying the system. If the parties ar many, and less stable, it is much harder to buy (sorry - influence) all the relevant politians. The price you pay is long coalition negotiations and, occasionally, weaker government. (Though those who praise the strength of a two-party system underestimate, IMO, the weakness when Presisent and Congress are in opposition).
That, of course, depends upon the instructions given to the police by the politicians, and is a political decision not a police one. Oversight should check that the police are doing what the politicians ordered. We should also have public visibility of the orders the politicians are giving to the police - which arguably is not true at the moment.
I concede that your cross-border seeking of the weakest rule set has validity. The same logic justifies Guantanmo, and "Special Rendition". We need a solution to this problem across the board, not just in cyberspace.
Wasn't there a "virus" a few years back that cleaned up any infected pcs it found?
There was a demonstration a few years ago of such a system, but I don't think it was ever released because of the legal problem described by parents. This is an area in which I think the law could be improved. Just as the police are allowed to break down a door (a behaviour normally regarded as criminal) if the genuinely believe a crime is being committed behind it, it seems reasonable that they should be able to recapture and repair hi-jacked computers. But you do need a supervision system to ensure that powers thus granted are not abused. But you need a supervision system whenever you grant people special powers for any reason whatsoever.
Legislators have the attention span of a butterfly on cocaine. Bills come, lobbyista swarm, bills are passed, bills go. DMCA, PATRIOT.. just get a good acronym and they will pass it. They don't have the time to think things through - they respond to very short term pressures. Hopefully, a group of policemen full-time committed to cyber-crime will have the time and energy to master the subject. And the resources i.e. the funds to hire a few high level geeks (not new graduates, though you probably need a few of those) to tell them how it really goes. Which, given what the real security gurus can charge, could be quite a lot.
Just becasue they are cybercops does not mean that they are necessarily bad guys. Cybercops, like any kind of cops, are a necessary evil: laws must be enforced, so we need law enforcers. If people would do the right thing without enforcers, you wouldn't need a law to tell them to do so. (Or course, laws which cannot or should not be enforced should be repealed; see you local legislator about this, not the police.) The problem is not the existence of cops, not (within broad limits) their powers. It is the checks and balances that need to be in place to ensure that their powers are not exceeded and that they are used in pursit of the ends for which they were allocated.
I absolutely accept that there is a need for police to tap electronic communications at some times. But not at any time a single policeman, however senior, thinks that he would like to. It is not tapping that is bad, it is tapping without a warrant. The executive branch needs some oversight, which is usually provided by the judicial branch i.e. the policeman needs to get a warrant from a judge, whose appointment needs to be transparent enough to ensure that he is not in cahoots with the policeman.
So I think your knee-jerk response to the concept of cybercops is excessive, and damaging. I want them bugging Osama bin Laden's phone calls. I don't want them bugging my, or my neighbours (equally innocent of major crime, though probably mostly guilty of the odd misdemeanor) phones. And i want to know who, and how, it watching to see tha this is so. But I don't want the head in the sand attitud of "they are all evil". They won't go away, and you may make them evil becasue, since you assume they are evil, they have nothing to lose by being evil. Support your local cop *if* he can show he is squeaky clean.
But if it does go wrong, things could be bad. Superconductors are laready prone to explosive failure if a superconductor suddenly ceasews to superconduct. If that is inside a very high pressure vessel, the available energy from a destructive malfunction is frightening : Mega-amps of electicity and giga-pascals of pressure suddenly being unleashed in the wrong place.
That is super-pure single-crystal silicon suitable for fabricating nanometre-sized transistors. Common-or-garden bulk silicon is cheap. Presumably for power transfer, thism doesn't need to be that perfect. Think premium bottled water vs tap water: there are plenty of rivers out there for your drinking water needs (though maybe not those of your farm).
Well, I had 20 years of no-friends before I got on the Net, where I met most of my current friends, including two girlfriends. It is all very well to say Get A Life, but if you are too shy to talk to anybody who is not already a friend, and have too low self esteem to think they they could possibly like yo it is rather too difficult to get started. I go to pubs and sit in the corner with my pint, then go home again without talking more than to order my drinks. On line, I am willing to dive into a conversation like this, because nobody knows that I am a dog. Just occasionally it goes further. Not often, but it beats the altenatives by an infinite ratio.
OK, if you are like the Korean who literally killed himself gaming, that is too much. But how many emails a day is "too much"? How many hours gaming are you allowed? I admit that most of my friends are online, though I occasionally meat some IRL. If I don't communicate with them, I get feeligns of loss (withdrawal) But before the Net, I didn't have friends. How is it worse to have net-friends instead of no-friends?
No. I am assuming that they have laptops which they may not carry into the office every time, or which they do not want to keep long term, or which may be shared between several workers, or which they cannot guarantee to have the same one every day or they want to move the data easily between work desktop, home desktop and mobile laptop without re-docking or they find it a pain to boot up the laptop in the office just to transfer one file or... A USB stick is a lot lighter than even a Macbook Air. My windows laptop takes botu 1.5 minutes to start up (OK, it is not the world's latest).
This is a child care agency. They need to visit the child and/or parents in their home, and have access to the child's records, both to read them (e.g. to find if any allegations are repeat cases) and to update them to record new allegations. You cannot get parents and childern to come into a secure environment for interview. The case worker, who may have to do three or four emotionally draining interviews in one day, cannot be expected to remember all the facts accurately enough for (for example) legal proceedings to remove a child from parents. Tha alternative to USB keys is probably printout, pen and paper. And how secure is t that? At least USB keys can be encrypted.
There can be justifiable reasons for taking the data off site. Rather than banning it completely, you need to do a (security) cost/benefit analysis and justify the action. And if yu can reduce the security cost, you may well be able to access more benefits.
Slashdot Mirror
That is what they are built for, so if it isn't, NVidia have deeply failed.
The CUDA software takes a little getting your head around - but not much. It is basically plain vanilla C (not C++) with vector processing. A single $400 8800 GPU board matches manufacturer-that-I-used-to-work-for's $50,000 hardware at evens. it isn't faster, but it isn't slower either. Two would be twice as fast.
It takes 4 cycles to do a floating point operation, and 4 cycles to do an integer add/subtract. It takes 16 cycles to do an integer multiply because it only has a 24-bit hardware multiplier (needed to achieve the 4-cycle flops, so it has to do long multiplies as four madds, This was for the first generation CUDA CPUs; the second generation, which should be out by now, was going to have double length floating and would be able to do 32 bit multiplies in the same four cycles.
While they can do integer, these machines are not very happy with it, and I found it much easier to do everything in floating point, even if you are talking about 8-bit colour data. It goes no slower, and everything is much better adapted to floating point. Then there are special instructions to get back to integer at the output.
While each operation takes 4 cycles, they are fully pipelined, so that it launched a new instruction per cycle, times 32 pipes per unit, times 8 units per GPU.
And madd is very useful for the sort of tasks for which supercomputers are traditionally used.
From my experience, you can buy support for any FOSS package worth mentioning, at a price that still beats commercial rivals. It is also my experience that the support thus purchased is outstandingly better than that for paid-for software. Problem responses within four hours, from somebody who really understands the system, instead of taking to weeks to dig through layers of ignoramuses to get to the expert. This, I conjecture, is because FOSS support teams live or die by the quality of support, whereas paid-for software put the best developers onto new features and regard support as very much a second-line function.
And, of course, many of the early car repair and servicing [places (and the early cars needed a lot of repairs) were the blacksmiths, because repairing a car was basically a matter of metalwork. Into the '70 s a lot of small village garages in the UK were recognisably descended from the village blacksmith. The farrier, however, did lose out.
What is actually being supplied is the ability to listen to nice noises. The transport layer is irrelevant.
Not Boeing, composites. Rutan has made a significant number of aircraft using composites on large scale. However, none within sight is the size of a B787, few intended for large-scale production, and none intended for the 365 days a year utilisation of a commercial airliner.
I may be unselective, but while I don't find every woman beautiful, I have no trouble finding pictures of beautiful women without the assistance of a computer.
In the name of God, WHY? The whole point of being a man (or a gay woman etc.) is so you can make your own estimation of women's beauty. One of the things I definitely DO NOT need a computer to do is to look at women for me. Many other things I may try and get computers to do for me - some sensible, some not. But I'll look at my own women, thank you very much,
Parallelism was the New New Thing when the Inmos Transputer rolled out in 1984 - a CPU explicitly designed to allow multiple CPUs to co-operate, and with a hardware scheduler to provide on-chip parallelism. Then we had the GAPP (Generalised Arithmetic parallel processor), the Connection Machine, and lots of other weird architectures whose names I cannot recall. I designed one myself in the late eighties, and we took it to breadboard stage (essentially Hyperthreading write very large).
Forgive me for not getting too excited by this new dawn. I have seen too many false dawns. Yes, parallelism is getting more and more prevalent, but expect more incremental steps, not an explosion.
There have been som studies funded by the UK Department of Health which showed no convincing results from cellphones. Now, they migh, of course, be in the pocket of the cellphome manufacturers, like they might be in the pocket of the drug manufacturers. But with a socialised health service, they are the people who are going to end up payiong the first level, purely medical, costs of any ill effects that there may be. Whcih suggests to me that, if they were going to err, they would be likely to err in the direction of overcaution rather than recklessness.
All the accusations against cellphones have been generally anecdotal i.e. a number of people have been found who were both heavy cellphone users and got brain tumours. But when large scale statistical studues are done, these "clusters" disappear. If you ask averybody with a tumour whether they were a heavy cellphone user, some will say yes. Probably more than really are, becasue moderate users will tend to judge themselves heavier in order to have something to blame for their tragedy - randomness seems much more frightening that a technological accident.
I don't think it is company vs. individual, it is main focus vs peripheral interest. If looking after your data is either the person's or the company's priimary task, you are much more likely to get good, focussed service from that individual. If your data is only one of 47 differnt things the company, or the indiividual, does, and is not regarded as the main one, then inevitably you will get poor service, and may be dropped or otherwise inconvenienced when they regard other aspects of their business as more important.
Obviously, if you employ someone, it is easy to tell them that their primary job, on which their salary depends, is storing and searching your data. But you can say the same about a company whose sole focus is managing other people's data: if they screw you around, you can werck them by telling the truth about their service. That would not be true if your data were stores by some giant such as Google or Amazon whose main business is elsewhere.
A small, hightly specialised, company is likely to perform as well as, or better than, your single employee, but with a higher truck number. But keep an eye on their financials - if they go bust you could be truly screwed. It is in your interest for them to be profitable. And what you do when tney get takenover by Google, I don't know. But what you do when your one in-hous expert leaves is no easier.
The articla assimes that any monopolist or near monopolist is going to get hated just for that fact. And, with DoubleClick, Apple is going to be dominant in both inline and display advertising. Therefore, he concludes, Apple will get hated. But that dominance per se does't wory me, the sucker^Wconsumer. I don't care who isgenerating the ads, I only care if they are invasive - which has no relationship to who owns them. The adverisers may come to hate Google's dominance of web-based adverting. But why should I cry for the advertisers? Indeed, if you believe they have a fixed ad budget, then the more Googel gouges them the fewer ads they can afford - which is great.
Of course, I am not saying that Google cannot get hated - it already is, a little bit. But if they keep their current policy of keeping opt-out easy, there is always a quick answer "then walk, dummy". The things that have made IBM, Microsofs and (looking upwards) Apple hated - things that, intentionally or otherwise, make it difficult to pick-and-mix your hardware and software options. As long as GMail doesn't lock me into uusing Google for search or vice versa, or I find my GMail locked in so I cannot load it into something else, or I cannot use Google without giving all my security information, it is difficult to see a real head of hate building up. Yes, Google is already losing that fresh and innocent look - but it can (if it wants to) avoid becoming actively hated, IMO.
Only if voters allow it - even if only by default. Again, the problem is political. We should be holding the politicians much more to account. But if you have a two party system, wealthy companies have only two targets to focus lobbying on, and persuade both to do what the companys want. This leacves the voters no way to tell the politicans to clean up their act. A system which supports only two (or othr small number) of immovable parties is inherently corruptible: the rich can seriously contemplate buying the system. If the parties ar many, and less stable, it is much harder to buy (sorry - influence) all the relevant politians. The price you pay is long coalition negotiations and, occasionally, weaker government. (Though those who praise the strength of a two-party system underestimate, IMO, the weakness when Presisent and Congress are in opposition).
That, of course, depends upon the instructions given to the police by the politicians, and is a political decision not a police one. Oversight should check that the police are doing what the politicians ordered. We should also have public visibility of the orders the politicians are giving to the police - which arguably is not true at the moment.
I concede that your cross-border seeking of the weakest rule set has validity. The same logic justifies Guantanmo, and "Special Rendition". We need a solution to this problem across the board, not just in cyberspace.
Legislators have the attention span of a butterfly on cocaine. Bills come, lobbyista swarm, bills are passed, bills go. DMCA, PATRIOT.. just get a good acronym and they will pass it. They don't have the time to think things through - they respond to very short term pressures. Hopefully, a group of policemen full-time committed to cyber-crime will have the time and energy to master the subject. And the resources i.e. the funds to hire a few high level geeks (not new graduates, though you probably need a few of those) to tell them how it really goes. Which, given what the real security gurus can charge, could be quite a lot.
Police Against New Internet Crimes.
Nothing too knee-jerk.
Just becasue they are cybercops does not mean that they are necessarily bad guys. Cybercops, like any kind of cops, are a necessary evil: laws must be enforced, so we need law enforcers. If people would do the right thing without enforcers, you wouldn't need a law to tell them to do so. (Or course, laws which cannot or should not be enforced should be repealed; see you local legislator about this, not the police.) The problem is not the existence of cops, not (within broad limits) their powers. It is the checks and balances that need to be in place to ensure that their powers are not exceeded and that they are used in pursit of the ends for which they were allocated.
I absolutely accept that there is a need for police to tap electronic communications at some times. But not at any time a single policeman, however senior, thinks that he would like to. It is not tapping that is bad, it is tapping without a warrant. The executive branch needs some oversight, which is usually provided by the judicial branch i.e. the policeman needs to get a warrant from a judge, whose appointment needs to be transparent enough to ensure that he is not in cahoots with the policeman.
So I think your knee-jerk response to the concept of cybercops is excessive, and damaging. I want them bugging Osama bin Laden's phone calls. I don't want them bugging my, or my neighbours (equally innocent of major crime, though probably mostly guilty of the odd misdemeanor) phones. And i want to know who, and how, it watching to see tha this is so. But I don't want the head in the sand attitud of "they are all evil". They won't go away, and you may make them evil becasue, since you assume they are evil, they have nothing to lose by being evil. Support your local cop *if* he can show he is squeaky clean.
But if it does go wrong, things could be bad. Superconductors are laready prone to explosive failure if a superconductor suddenly ceasews to superconduct. If that is inside a very high pressure vessel, the available energy from a destructive malfunction is frightening : Mega-amps of electicity and giga-pascals of pressure suddenly being unleashed in the wrong place.
That is super-pure single-crystal silicon suitable for fabricating nanometre-sized transistors. Common-or-garden bulk silicon is cheap. Presumably for power transfer, thism doesn't need to be that perfect. Think premium bottled water vs tap water: there are plenty of rivers out there for your drinking water needs (though maybe not those of your farm).
Well, I had 20 years of no-friends before I got on the Net, where I met most of my current friends, including two girlfriends. It is all very well to say Get A Life, but if you are too shy to talk to anybody who is not already a friend, and have too low self esteem to think they they could possibly like yo it is rather too difficult to get started. I go to pubs and sit in the corner with my pint, then go home again without talking more than to order my drinks. On line, I am willing to dive into a conversation like this, because nobody knows that I am a dog. Just occasionally it goes further. Not often, but it beats the altenatives by an infinite ratio.
OK, if you are like the Korean who literally killed himself gaming, that is too much. But how many emails a day is "too much"? How many hours gaming are you allowed? I admit that most of my friends are online, though I occasionally meat some IRL. If I don't communicate with them, I get feeligns of loss (withdrawal) But before the Net, I didn't have friends. How is it worse to have net-friends instead of no-friends?
No. I am assuming that they have laptops which they may not carry into the office every time, or which they do not want to keep long term, or which may be shared between several workers, or which they cannot guarantee to have the same one every day or they want to move the data easily between work desktop, home desktop and mobile laptop without re-docking or they find it a pain to boot up the laptop in the office just to transfer one file or... A USB stick is a lot lighter than even a Macbook Air. My windows laptop takes botu 1.5 minutes to start up (OK, it is not the world's latest).
This is a child care agency. They need to visit the child and/or parents in their home, and have access to the child's records, both to read them (e.g. to find if any allegations are repeat cases) and to update them to record new allegations. You cannot get parents and childern to come into a secure environment for interview. The case worker, who may have to do three or four emotionally draining interviews in one day, cannot be expected to remember all the facts accurately enough for (for example) legal proceedings to remove a child from parents. Tha alternative to USB keys is probably printout, pen and paper. And how secure is t that? At least USB keys can be encrypted.
There can be justifiable reasons for taking the data off site. Rather than banning it completely, you need to do a (security) cost/benefit analysis and justify the action. And if yu can reduce the security cost, you may well be able to access more benefits.