Ok, here's a question. Why should they design a low power desktop when:
1. Your lightbulbs, AC, refrigerator consumer about 5x as much electricity.
2. The first thing the consumer does is shut it off....hmmmm....
Maybe you should pick a different cause..
Actually, that's not the point. The problem is that your algorithm is recursive and scales very poorly. The point of the article was to find a way to solve NP problems in a a way that's similar to P problems. In order to solve the problem the way you are doing it, the computer has to brute force it, by looking at ALL the data. The challenge is to find a way to do it that doesn't involve traversing every path, which hasn't been solved yet, but will need to be solved yet. Many algorithms, are brute forcing things, and will fail to scale. Some examples are designing the fastest route through the internet so that traffic can be routed in the quickest way possible. Using the brute force approach of testing all possible paths, will not scale, and will eventually slow things such as the internet to a crawl as the number of paths increases.
As someone who programs. I disagree completely. One thing that you have to keep in mind, is that modern software is enormously complex. To say that a modern programmer is lazy, is to misunderstand what it is that they do. Think about this, the software that I program on for a living, takes up close to a million lines of code, has it's own windowing API, as well as cross platform stdio functions that replace C's functions, and many, many tools. So, how do you go about programming something like that in assembly? Remember, you need portability, so assembly would be stupid choice for something so large and complex. It's nice to think that we could all go back and write our stuff in assembly, but I think that when you start thinking like that, you're overlooking that simple programs today are so large and complex, and do so much, that they would easily overflow the memory banks of 'advanced' machines of two decades ago. That's why programmers used to be able to get away with assembly, the stuff they were doing wasn't nearly as complex, and they didn't have to worry about portability as much.
How long do you think it will be before marathon runners start using this stuff instead of blood doping? It sounds like it would be much safer than blood doping since the user wouldn't have extra blood driving up blood pressure, but instead would be replacing the blood with a more efficient oxygen carrier.
Re:That's not the point, tho...
on
Should You Vote?
·
· Score: 1
Hmmmm.., I guess it depends on what number n is. If you pick n to be 10, and there are 10 candidates, then the 10th choice would only get 1. However, if you pick n to be 100, then the tenth choice would get 90, so I guess it depends on how you pick n. You could acheive a similar effect to what I was proposing by picking n to be 5 for 10 candidates, and not allowing the number 0, then the 10th choice would get -5. The reason that I thought my system was a good idea, is that a candidate that has 25% of the votes, but is desperately despised by the other 75%, shouldn't be made president just because he happens to get more votes than the next candidate in line, who may have only 20% of the votes. This way, a candidate who is disliked by more people than he is liked, will not be able to become president, therefore, not only must you be chosen by more people than any other candidate, but you must also make sure that your views aren't so extreme(or conservative for that matter), that you alienate the other voters. I only think that my way of doing things would fail for the main parties if alot of people don't want them in office, and in that case, if there are so many people that don't want them, that they can't get in office, then they shouldn't have a way of getting there.
How about 2 votes, one negative, the otherpositive
on
Should You Vote?
·
· Score: 1
How about instead, a system where each voter gets one positive vote, and another negative vote. This would have a similar effect to what you are proposing, but would be less likely to be manipulated or tampered with. A system such as yours sounds like it would need a set of equations to work properly, and that may need to be adjusted depending on the number of candidates. This system could work no matter how many candidates there are, and would also help to keep certain parties from garnering the majority of the votes. People who vote Democrat would be likely to unvote Republican. However, voters that vote for one of the other parties, could also unvote Bush or Gore. Another system would be a two vote system, but this has the problem of giving more votes to the parties in power. By allowing the voter to have a negative vote, it will take some of the power away from the Republicans/Democrats.
I think what most bussinesses will do, is blatantly pass on this fee. I'd like to see AT&T explain their way out of this one when customers realize they are paying extra just for using AT&T networks.
I think the whole point is, is that dot com employers don't want less drug use. They could care less, that's why that guy's father refused to test for drugs even after his son died of an overdose. They're looking for young fresh minds to use and abuse. And anything that will help these minds to get more work done is fair game to them. The whole lifestyle is built around a manic race to earn as much money as possible, risk and consequences be damned. The only way to keep from living this lifestyle, is to refuse it, to not allow employers to dangle that carrot in front of you. Don't depend on them to look out for you, because they won't. It's not worth it.
Let's see here:
100mhz at 64 bits, is 768MB/sec of bandwidth
266mhz at 32 bits is 1024MB/sec of bandwidth
AGP 4xwins hands down. You also have to look at the bandwidth on the graphics cards, the elites are notoriously slow. Ok, here we go...
Dude, you're retarded, no fucking shit the USIII is a processor, I was referring to the architecture that's designed around the USIII. Go ahead and buy a USIII, I'll save about $10,000 and have a PC that's twice as fast in opengl. The nice thing about slashdot, is that most of these dumbshits (like you)recommending SUN's for opengl aren't in the field, and most of the guys dumb enough to believe they are fast at opengl, don't have enough money to buy one. So, at least people aren't losing any money. The only thing that's annoying, is the spread of misinformation. Do me a favor, look up the specs on the elite3d or whatever SUN's "top of the line" graphics card and compare it to what's available for PC's, it's no comparison. You're friend even agreed with me when I told him that the PC had faster graphics cards. He comes off saying that the SUN kicks ass in opengl, and then I correct him, and so he fucking caves in and says, yeah, but, if the SUN had an nvidia card, THEN it would be faster. Well, who cares? That's like recommending that someone buy a car because in 10 years, someone might produce a better engine for it. Who gives a shit, by then it will be ready for the junk heap. Oh, please enlighten me about the way you make a living, since you seem to know SOOOO much about computers. I would be willing to bet from your know-it-all attitude that you're an NT admin.
And your other argument. Your saying that the SUN's proprietary port is a good thing. You're even dumber than I thought. How in the fuck does SUN expect any graphics card manufacturer to waste their time with a proprietary hardware port for such a small market segment? No wonder the selection of graphics cards sucks. That lends even less credit to the idea of nvidia EVER bringing their cards to the SUN. And no, elite3d's are slow, I've used them, they suck. So do the Creator's. They aren't even 50% as fast.
Actually, if the faster card is out is on x86, then it has everything to do with x86 architecture. Granted, Nvidia could release this card for SUN, but tell me, does the USIII have an AGP port? That would be a pretty big thing hindering performance right there. It's like I tell my mac zealot friend, sure the G4 can do more fp per clock cycle than my Athlon, but it's a moot point when the graphics cards that are on my x86 are orders of magnitude faster, and my PC has a high bandwidth dedicated port for the card. There's no doubt that a SUN with AGP and a good graphics card would be a great high end 3D workstation, but don't hold your breath. And until then, there is no way that I will recommend them to customers. By the time Nvidia brings there cards to SUN workstations, there will probably already be a new sparc architecture release. Investing in a SUN at this point in time in the hope that something comes out would be an incredibly stupid decision. Computers go obsolete way too fast, and you're better off waiting until the support is there before dropping $10,000+ when by the time a card that's good comes out for it, you'll be able to get a 2.5 Ghz athlon that would be just as fast, for about $2,000.
Thankyou!, I couldn't agree more. I doubt these people spouting off about SUN opengl performance have even used one. OTOH, I have one in my office, and the opengl performance is less than stellar to say the least. Please read my other reply for more info.
It's http://slashdot.org/comments.pl?sid=00/09/27/14262 07&threshold=0&commentsort=0&mode=thread &pid=107#294
I just thought I'd make sure my post got the rating it deserved. I don't write that much to have my time wasted. And, I can't stand people spreading misinformation about what SUN boxes are good for.
I program opengl for a living, and have an ultrasparc II in my office. Let me tell you what our experience has been. The PC is blowing everything out of the water. Even the 'high end' Onyx2's are starting to pale in comparison. Why? One word, Nvidia. You need to go study the Opengl pipeline a little bit more, boy. Opengl requires high memory bandwidth and low latency. That's why, for many (read, most) applications, an nvidia card with 460 mhz DDR memory and 50 billion flops of geometry processing power will be much faster than a $100,000 SGI box with only 100mhz of bandwidth to the memory. Think about it.... Believe me, we have an Onyx2 sitting in our computer room, that's the size of a small fridge, with 4 processors, and 2 rastermans, and my machine at home with the new nvidia quadro pro beta test board blows it away. The machine at home has an 800 mhz Athlon, BTW. Our application is a 3D Virtual world application, it uses quite a few polygons, enough to give the Onyx2 a hard time, but my PC at home is doing just fine. In order to build a comparable Onyx2 configuration (if it's possible) one would have to spend 10's of thousands of dollars. I don't believe that it's possible, there is too much latency in multi-processor architectures, the more CPU's you add, the more likely you will have bus congestion or have to wait for some data from another CPU before you can render the frame. For pure number crunching apps like RC5, that's fine, but for a real-time app like 3D graphics, that bandwidth is crucial, and the cards on the SUN's just don't have it. In the world of graphics, sometimes smaller, specialized processors will beat out ultra expensive boxes with tons of processors on board. And I'm really unsure of what you mean by 'professional' OpenGL performance. Please come up with a professional app that runs faster on a sun than on a modern PC with a good graphics card. I know that ours certainly doesn't. It's not even close. Don't get me wrong, SUN's make great servers, but I would never recommend them for opengl.
Again, this is assuming that we're talking about real-time performance. Not setting up renderfarms for movies. Obviously, in the latter case, setting up lots of boxes in parallel is faster.
Gee, maybe what they are saying is that a teletype terminal emulator, which is what a VT100 terminal is, is not the best way to access your computer. Maybe you should change your myopic point of view, and realize that just because it can do some things better than windows (which at 20 years old, is also an arcance interface) does not mean that it isn't a primitive interface. We are running computers that are much more powerful than when UNIX was designed, yet we still are using a communication system that is based on a QWERTY keyboard (about 100+ years old) and line printer emulator (yourmonitor) which is about 50+ years old. But, I guess like you are saying, it's not primitive, whatever. The biggest bottleneck in modern technology today isn't your RAM or your hard drive, it's the primitive interface, which not only fails to use the power of the computer effeciently, it also fails to take advantage of the most powerful computer in the room, you, by forcing you to do the things that you are bad at in order to communicate with the machine. What are humans bad at? 1. Remmembering facts that don't have a context. i.e. command lines switches, the spellings of certain commands, directory structure, etc. 2. Translating ideas into a language, the essence of the idea is often lost, and many times quite a bit of work goes into translating an idea correctly. what are we good at? 1. Remembering ideas that have a context. Think about it, you have memorized every object in your house, about 5,000 separate items, and you can locate most of them without any effort, because your brain is very good at visualizing and is a powerful, context sensitive computer. 2. Forming connections and associations between ideas. The less arbitrary the OS design, and the more intuitive it is, the easier it will be to learn. We have a long way to go. Hopefully someday, a machine will be able to tap into the part of my brain that forms images, and rather than having to say something, I will remember what it looked like, or rememeber a diagram, and the machine will guess at what I am thinking, this will bypass the QUERTY keyboard, and allow me to think in a more pictoral, context senstive way. At the least, the computer could be programmed to intelligently interpret natural language, which we have already invest years of our lives in learning, and would make us use more precise descriptions only when necessary, instead of always having to tell it EXACTLY what we want it to do, and how to do it. But hey, if you like UNIX, don't let me stop you. I think the power and transparency are nice, the cryptic interface has to go.
Slashdot Mirror
Ok, here's a question. Why should they design a low power desktop when: 1. Your lightbulbs, AC, refrigerator consumer about 5x as much electricity. 2. The first thing the consumer does is shut it off....hmmmm.... Maybe you should pick a different cause..
Actually, that's not the point. The problem is that your algorithm is recursive and scales very poorly. The point of the article was to find a way to solve NP problems in a a way that's similar to P problems. In order to solve the problem the way you are doing it, the computer has to brute force it, by looking at ALL the data. The challenge is to find a way to do it that doesn't involve traversing every path, which hasn't been solved yet, but will need to be solved yet. Many algorithms, are brute forcing things, and will fail to scale. Some examples are designing the fastest route through the internet so that traffic can be routed in the quickest way possible. Using the brute force approach of testing all possible paths, will not scale, and will eventually slow things such as the internet to a crawl as the number of paths increases.
As someone who programs. I disagree completely. One thing that you have to keep in mind, is that modern software is enormously complex. To say that a modern programmer is lazy, is to misunderstand what it is that they do. Think about this, the software that I program on for a living, takes up close to a million lines of code, has it's own windowing API, as well as cross platform stdio functions that replace C's functions, and many, many tools. So, how do you go about programming something like that in assembly? Remember, you need portability, so assembly would be stupid choice for something so large and complex. It's nice to think that we could all go back and write our stuff in assembly, but I think that when you start thinking like that, you're overlooking that simple programs today are so large and complex, and do so much, that they would easily overflow the memory banks of 'advanced' machines of two decades ago. That's why programmers used to be able to get away with assembly, the stuff they were doing wasn't nearly as complex, and they didn't have to worry about portability as much.
How long do you think it will be before marathon runners start using this stuff instead of blood doping? It sounds like it would be much safer than blood doping since the user wouldn't have extra blood driving up blood pressure, but instead would be replacing the blood with a more efficient oxygen carrier.
Hmmmm.., I guess it depends on what number n is. If you pick n to be 10, and there are 10 candidates, then the 10th choice would only get 1. However, if you pick n to be 100, then the tenth choice would get 90, so I guess it depends on how you pick n. You could acheive a similar effect to what I was proposing by picking n to be 5 for 10 candidates, and not allowing the number 0, then the 10th choice would get -5. The reason that I thought my system was a good idea, is that a candidate that has 25% of the votes, but is desperately despised by the other 75%, shouldn't be made president just because he happens to get more votes than the next candidate in line, who may have only 20% of the votes. This way, a candidate who is disliked by more people than he is liked, will not be able to become president, therefore, not only must you be chosen by more people than any other candidate, but you must also make sure that your views aren't so extreme(or conservative for that matter), that you alienate the other voters. I only think that my way of doing things would fail for the main parties if alot of people don't want them in office, and in that case, if there are so many people that don't want them, that they can't get in office, then they shouldn't have a way of getting there.
How about instead, a system where each voter gets one positive vote, and another negative vote. This would have a similar effect to what you are proposing, but would be less likely to be manipulated or tampered with. A system such as yours sounds like it would need a set of equations to work properly, and that may need to be adjusted depending on the number of candidates. This system could work no matter how many candidates there are, and would also help to keep certain parties from garnering the majority of the votes. People who vote Democrat would be likely to unvote Republican. However, voters that vote for one of the other parties, could also unvote Bush or Gore. Another system would be a two vote system, but this has the problem of giving more votes to the parties in power. By allowing the voter to have a negative vote, it will take some of the power away from the Republicans/Democrats.
I think what most bussinesses will do, is blatantly pass on this fee. I'd like to see AT&T explain their way out of this one when customers realize they are paying extra just for using AT&T networks.
I think the whole point is, is that dot com employers don't want less drug use. They could care less, that's why that guy's father refused to test for drugs even after his son died of an overdose. They're looking for young fresh minds to use and abuse. And anything that will help these minds to get more work done is fair game to them. The whole lifestyle is built around a manic race to earn as much money as possible, risk and consequences be damned. The only way to keep from living this lifestyle, is to refuse it, to not allow employers to dangle that carrot in front of you. Don't depend on them to look out for you, because they won't. It's not worth it.
Let's see here: 100mhz at 64 bits, is 768MB/sec of bandwidth 266mhz at 32 bits is 1024MB/sec of bandwidth AGP 4xwins hands down. You also have to look at the bandwidth on the graphics cards, the elites are notoriously slow. Ok, here we go... Dude, you're retarded, no fucking shit the USIII is a processor, I was referring to the architecture that's designed around the USIII. Go ahead and buy a USIII, I'll save about $10,000 and have a PC that's twice as fast in opengl. The nice thing about slashdot, is that most of these dumbshits (like you)recommending SUN's for opengl aren't in the field, and most of the guys dumb enough to believe they are fast at opengl, don't have enough money to buy one. So, at least people aren't losing any money. The only thing that's annoying, is the spread of misinformation. Do me a favor, look up the specs on the elite3d or whatever SUN's "top of the line" graphics card and compare it to what's available for PC's, it's no comparison. You're friend even agreed with me when I told him that the PC had faster graphics cards. He comes off saying that the SUN kicks ass in opengl, and then I correct him, and so he fucking caves in and says, yeah, but, if the SUN had an nvidia card, THEN it would be faster. Well, who cares? That's like recommending that someone buy a car because in 10 years, someone might produce a better engine for it. Who gives a shit, by then it will be ready for the junk heap. Oh, please enlighten me about the way you make a living, since you seem to know SOOOO much about computers. I would be willing to bet from your know-it-all attitude that you're an NT admin. And your other argument. Your saying that the SUN's proprietary port is a good thing. You're even dumber than I thought. How in the fuck does SUN expect any graphics card manufacturer to waste their time with a proprietary hardware port for such a small market segment? No wonder the selection of graphics cards sucks. That lends even less credit to the idea of nvidia EVER bringing their cards to the SUN. And no, elite3d's are slow, I've used them, they suck. So do the Creator's. They aren't even 50% as fast.
Actually, if the faster card is out is on x86, then it has everything to do with x86 architecture. Granted, Nvidia could release this card for SUN, but tell me, does the USIII have an AGP port? That would be a pretty big thing hindering performance right there. It's like I tell my mac zealot friend, sure the G4 can do more fp per clock cycle than my Athlon, but it's a moot point when the graphics cards that are on my x86 are orders of magnitude faster, and my PC has a high bandwidth dedicated port for the card. There's no doubt that a SUN with AGP and a good graphics card would be a great high end 3D workstation, but don't hold your breath. And until then, there is no way that I will recommend them to customers. By the time Nvidia brings there cards to SUN workstations, there will probably already be a new sparc architecture release. Investing in a SUN at this point in time in the hope that something comes out would be an incredibly stupid decision. Computers go obsolete way too fast, and you're better off waiting until the support is there before dropping $10,000+ when by the time a card that's good comes out for it, you'll be able to get a 2.5 Ghz athlon that would be just as fast, for about $2,000.
The athlon 450 doesn't exist. Find a clue.
Thankyou!, I couldn't agree more. I doubt these people spouting off about SUN opengl performance have even used one. OTOH, I have one in my office, and the opengl performance is less than stellar to say the least. Please read my other reply for more info. It's http://slashdot.org/comments.pl?sid=00/09/27/14262 07&threshold=0&commentsort=0&mode=thread &pid=107#294
I just thought I'd make sure my post got the rating it deserved. I don't write that much to have my time wasted. And, I can't stand people spreading misinformation about what SUN boxes are good for. I program opengl for a living, and have an ultrasparc II in my office. Let me tell you what our experience has been. The PC is blowing everything out of the water. Even the 'high end' Onyx2's are starting to pale in comparison. Why? One word, Nvidia. You need to go study the Opengl pipeline a little bit more, boy. Opengl requires high memory bandwidth and low latency. That's why, for many (read, most) applications, an nvidia card with 460 mhz DDR memory and 50 billion flops of geometry processing power will be much faster than a $100,000 SGI box with only 100mhz of bandwidth to the memory. Think about it.... Believe me, we have an Onyx2 sitting in our computer room, that's the size of a small fridge, with 4 processors, and 2 rastermans, and my machine at home with the new nvidia quadro pro beta test board blows it away. The machine at home has an 800 mhz Athlon, BTW. Our application is a 3D Virtual world application, it uses quite a few polygons, enough to give the Onyx2 a hard time, but my PC at home is doing just fine. In order to build a comparable Onyx2 configuration (if it's possible) one would have to spend 10's of thousands of dollars. I don't believe that it's possible, there is too much latency in multi-processor architectures, the more CPU's you add, the more likely you will have bus congestion or have to wait for some data from another CPU before you can render the frame. For pure number crunching apps like RC5, that's fine, but for a real-time app like 3D graphics, that bandwidth is crucial, and the cards on the SUN's just don't have it. In the world of graphics, sometimes smaller, specialized processors will beat out ultra expensive boxes with tons of processors on board. And I'm really unsure of what you mean by 'professional' OpenGL performance. Please come up with a professional app that runs faster on a sun than on a modern PC with a good graphics card. I know that ours certainly doesn't. It's not even close. Don't get me wrong, SUN's make great servers, but I would never recommend them for opengl. Again, this is assuming that we're talking about real-time performance. Not setting up renderfarms for movies. Obviously, in the latter case, setting up lots of boxes in parallel is faster.
Gee, maybe what they are saying is that a teletype terminal emulator, which is what a VT100 terminal is, is not the best way to access your computer. Maybe you should change your myopic point of view, and realize that just because it can do some things better than windows (which at 20 years old, is also an arcance interface) does not mean that it isn't a primitive interface. We are running computers that are much more powerful than when UNIX was designed, yet we still are using a communication system that is based on a QWERTY keyboard (about 100+ years old) and line printer emulator (yourmonitor) which is about 50+ years old. But, I guess like you are saying, it's not primitive, whatever. The biggest bottleneck in modern technology today isn't your RAM or your hard drive, it's the primitive interface, which not only fails to use the power of the computer effeciently, it also fails to take advantage of the most powerful computer in the room, you, by forcing you to do the things that you are bad at in order to communicate with the machine. What are humans bad at? 1. Remmembering facts that don't have a context. i.e. command lines switches, the spellings of certain commands, directory structure, etc. 2. Translating ideas into a language, the essence of the idea is often lost, and many times quite a bit of work goes into translating an idea correctly. what are we good at? 1. Remembering ideas that have a context. Think about it, you have memorized every object in your house, about 5,000 separate items, and you can locate most of them without any effort, because your brain is very good at visualizing and is a powerful, context sensitive computer. 2. Forming connections and associations between ideas. The less arbitrary the OS design, and the more intuitive it is, the easier it will be to learn. We have a long way to go. Hopefully someday, a machine will be able to tap into the part of my brain that forms images, and rather than having to say something, I will remember what it looked like, or rememeber a diagram, and the machine will guess at what I am thinking, this will bypass the QUERTY keyboard, and allow me to think in a more pictoral, context senstive way. At the least, the computer could be programmed to intelligently interpret natural language, which we have already invest years of our lives in learning, and would make us use more precise descriptions only when necessary, instead of always having to tell it EXACTLY what we want it to do, and how to do it. But hey, if you like UNIX, don't let me stop you. I think the power and transparency are nice, the cryptic interface has to go.