This method ("Instant Runoff Voting") has edge case issues that are solved in a similar system called Condorcet voting. A start at an explaination of the problem is at http://condorcet.org/rp/IRV.shtml.
You can have a language that catches mistakes like this yet isn't overly verbose. In a more strongly typed language (like Haskell for example) the erroneous code wouldn't have compiled because a pointer to a function obviously isn't what you want in an integer comparison.
In case 1, movie studios and bands are stil fine. Movies make a good chunk of money in the cinema, and people will still buy "collectors edition" DVD boxed sets reguardless of free downloading. Bands will still be fine because they make most of their money at live shows anyway.
The only people who lose out significantly are big name actors (movie budgets will go down some, and $10 million is a bit high for two months of work), as well as music studio execs unless they get their pricing model right and figure out the "collector's edition" trick soon enough.
With the current physical layout of the United States, not driving isn't an especially valid choice for a significant portion of the suburban population (there are areas of suburbia where it's 3 miles to the supermarket and a mile and a half to the nearest bus stop). Car insurance is effectively "manditory" for a good chunk of the population.
A smaller but perhaps more obvious example is the fact that, in Massachussetts, health insurance is mandatory for full time college students.
This sort of thing does reduce the relitively minor issues of poor people not paying unexpected car / medical expenses, but the primary benifit is for the stockholders and executives at the insurance companies.
So... you are able to get the same functionality that you could get from a $5 used video card off eBay. Since your nVidia card probably cost more than $5, that's not the best deal ever.
The general idea of having different memory at different speeds and using your faster memory to "cache" information from the slower memory is one that will be around for a long, long time.
It's true that specifically for PCI Express using main memory to supplement on-board memory will get worse as the PCI Express bus is used more fully by future cards, I doubt that it will become completely useless before the successor to PCI express comes around with "excess" capacity that can be used in this way. It's not like main memory is getting faster any slower than graphics memory, nor are games getting less hungry for texture memory.
While using shared memory to entirely replace onboard memory on a graphics card for video games is probably not optimal, I'm going to have to disagree with you in the general case.
Using shared memory appropriately in a TurboCache archecture is no more "bad" than using system memory instead of having a gig of on-chip CPU cache is bad.
All these system archetecture choices are cost/benifit tradeoffs, and a hybrid Shared/Dedicated memory archetecture is just more flexible than the alternitives.
Well, at what point does a computer monitor start to become a home entertainment monitor?
The minute you start playing video games on it.
More seriously, there's no reason that a higher resolution monitor implies a giant monitor.
19" CRTs commonly support 1920x1440.
A 23" 1920x1200 really isn't that huge.
20" LCDs come in 1600x1200 and 1680x1050.
The point is better quality video, and resolution is an important component to that.
I already have a 42" Grand Wega HDTV
Have you considered hooking it up to your computer? I bet that would look awesome, expecially if it does 1080p (assuming you have a decent graphics card).
TurboCache is a technology that's not nessarily limited to the GeForce 6200 line of cards. It's true that those particular cards give comparitively low performance, but there's no reason why the TurboCache technology couldn't be in some future GeForce 7700 card that provides solid game performance.
In the general case, TurboCache is much more flexible for providing 3D texture memory than either 100% dedicated video RAM or 100% shared system RAM.
I would like to point out at this time that a decent $300 card has 24 pixel pipelines. Now, the X800 GT02 is decent, but you haven't managed to cheat your price point significantly.
The 5900 -> 7900 leap is pretty significant. I'm suprised you aren't impressed by it.
You're talking about the jump from "Runs HL2 decently at 800x600" to "Fully supports all the features needed to run HL2: Lost Coast beautifully at 1600x1200". If you really can just blow off four times as many pixels, higher resolution textures, TAA, and HDR lighting, then that's your loss.
Oblivion is the "current" video card abusing game. Historically, there has usually been a game that supports higher quality graphics than the top of the line cards can produce - because new cards will come out that support those graphics modes later.
I don't know about 1600x1200, but at 2500x1600 Oblivion does manage to crush any card on the market at full quality.
TurboCache is actually a great idea. They're currently only using it for trash low-end cards, but that doesn't mean that abusing the underused PCI Express bus to get a couple more megs of "fake" video RAM isn't the plan.
The only thing I'd be afraid of is the fact that it uses a "subset" of the MIPS instruction set. It's not like MIPS is a wildly complex and hard to implement instruction set...
Assuming that this processor is no slower than a 600mhz Pentium III, the machine can easily run a normal set of productivity applications and access the internet. It'll be slow, but not too bad.
As other posters have stated, $150 in rural China is like half a year's pay. Paying more than twice as much for better performance isn't nessisarily the plan. It'd be like if I had gotten a BMW C series instead of my Hyundai Accent - possible, but not nessisarily economically responsible.
So... you can get the same machine for $50 more with one difference: A Celeron instead of the non-intel processor. If the almost-MIPS thing has the power benifits that I'd expect it to have, then for the rural chinese market it could easily be the better deal.
The patent office is well built to resist denial of service attacks like that - it costs a shitload of money to apply for a patent. That does mean that few "small inventors" can ever afford a patent, but it prevents it from getting "bogged down" at least.
The problem isn't as simple as regulation vs. deregulation, it's an issue with a bad market model. Specifically, there are regulations that offer total local monopolies on data services.
The current regulations prevent competition. Either it needs to be legal for other companies to run their own lines, or it needs to be illegal for the line owners to provide data services. (Or the government needs to step in and provide decent telecom service, but that has it's own large set of problems)
No one's ever going to see the reduced accuracy. It's not like people are going at the graph with calipers to evaluate the function. I highly doubt that going from 30kb worth of curve data to 15kb worth of curve data will even be visible at any likely resolution.
How much adminining do they really need to do? Installing desktop apps and updates is trival if they have an internet connection. Even if they don't have an internet connection, working off the install CD isn't that bad. Printers and stuff mostly just work.
I'm not seeing it. What's this task they're going to be doing that'll be any more difficult than Windows XP vs. Windows 2000?
The only time there's any difference between the BSD license and the GPL licence is when some company decides that they need to keep their source code a trade secret and still redistribute the binaries.
There is only one business model where that even matters, and that business model can be emulated pretty closely with GPL software (i.e. What licence is the Quake 2 source code released under? Can you legally distribute copies of Quake 2 yet? No - the data files are still under pure copyright.) So you'd need to release some code if you made modifications to quake 2 - who cares? Your compeditors will get to start looking at your code when they buy a copy of your game. By the time they can produce a product using that code, your game will be done with its lifecycle anyway.
Slashdot Mirror
This method ("Instant Runoff Voting") has edge case issues that are solved in a similar system called Condorcet voting. A start at an explaination of the problem is at http://condorcet.org/rp/IRV.shtml.
You can have a language that catches mistakes like this yet isn't overly verbose. In a more strongly typed language (like Haskell for example) the erroneous code wouldn't have compiled because a pointer to a function obviously isn't what you want in an integer comparison.
The only people who lose out significantly are big name actors (movie budgets will go down some, and $10 million is a bit high for two months of work), as well as music studio execs unless they get their pricing model right and figure out the "collector's edition" trick soon enough.
With the current physical layout of the United States, not driving isn't an especially valid choice for a significant portion of the suburban population (there are areas of suburbia where it's 3 miles to the supermarket and a mile and a half to the nearest bus stop). Car insurance is effectively "manditory" for a good chunk of the population.
A smaller but perhaps more obvious example is the fact that, in Massachussetts, health insurance is mandatory for full time college students.
This sort of thing does reduce the relitively minor issues of poor people not paying unexpected car / medical expenses, but the primary benifit is for the stockholders and executives at the insurance companies.
Hey, If you can make a perfect copy of my girlfriend then I'm absolutely fine with you trying to convince the copy to go out with you.
So... you are able to get the same functionality that you could get from a $5 used video card off eBay. Since your nVidia card probably cost more than $5, that's not the best deal ever.
The general idea of having different memory at different speeds and using your faster memory to "cache" information from the slower memory is one that will be around for a long, long time.
It's true that specifically for PCI Express using main memory to supplement on-board memory will get worse as the PCI Express bus is used more fully by future cards, I doubt that it will become completely useless before the successor to PCI express comes around with "excess" capacity that can be used in this way. It's not like main memory is getting faster any slower than graphics memory, nor are games getting less hungry for texture memory.
While using shared memory to entirely replace onboard memory on a graphics card for video games is probably not optimal, I'm going to have to disagree with you in the general case.
Using shared memory appropriately in a TurboCache archecture is no more "bad" than using system memory instead of having a gig of on-chip CPU cache is bad.
All these system archetecture choices are cost/benifit tradeoffs, and a hybrid Shared/Dedicated memory archetecture is just more flexible than the alternitives.
The minute you start playing video games on it.
More seriously, there's no reason that a higher resolution monitor implies a giant monitor.
The point is better quality video, and resolution is an important component to that.
Have you considered hooking it up to your computer? I bet that would look awesome, expecially if it does 1080p (assuming you have a decent graphics card).
In the general case, TurboCache is much more flexible for providing 3D texture memory than either 100% dedicated video RAM or 100% shared system RAM.
Get a motherboard with intel integrated graphics. It'll run XGL fine on free software drivers.
Sounds like the video card isn't the component that's holding you back any more.
I would like to point out at this time that a decent $300 card has 24 pixel pipelines. Now, the X800 GT02 is decent, but you haven't managed to cheat your price point significantly.
The 5900 -> 7900 leap is pretty significant. I'm suprised you aren't impressed by it.
You're talking about the jump from "Runs HL2 decently at 800x600" to "Fully supports all the features needed to run HL2: Lost Coast beautifully at 1600x1200". If you really can just blow off four times as many pixels, higher resolution textures, TAA, and HDR lighting, then that's your loss.
Oblivion is the "current" video card abusing game. Historically, there has usually been a game that supports higher quality graphics than the top of the line cards can produce - because new cards will come out that support those graphics modes later.
I don't know about 1600x1200, but at 2500x1600 Oblivion does manage to crush any card on the market at full quality.
TurboCache is actually a great idea. They're currently only using it for trash low-end cards, but that doesn't mean that abusing the underused PCI Express bus to get a couple more megs of "fake" video RAM isn't the plan.
You can use any bank in the world. Especially with online banking, where your bank's physical offices are doesn't matter in the slightest.
The only thing I'd be afraid of is the fact that it uses a "subset" of the MIPS instruction set. It's not like MIPS is a wildly complex and hard to implement instruction set...
Assuming that this processor is no slower than a 600mhz Pentium III, the machine can easily run a normal set of productivity applications and access the internet. It'll be slow, but not too bad.
As other posters have stated, $150 in rural China is like half a year's pay. Paying more than twice as much for better performance isn't nessisarily the plan. It'd be like if I had gotten a BMW C series instead of my Hyundai Accent - possible, but not nessisarily economically responsible.
So... you can get the same machine for $50 more with one difference: A Celeron instead of the non-intel processor. If the almost-MIPS thing has the power benifits that I'd expect it to have, then for the rural chinese market it could easily be the better deal.
The patent office is well built to resist denial of service attacks like that - it costs a shitload of money to apply for a patent. That does mean that few "small inventors" can ever afford a patent, but it prevents it from getting "bogged down" at least.
The problem isn't as simple as regulation vs. deregulation, it's an issue with a bad market model. Specifically, there are regulations that offer total local monopolies on data services.
The current regulations prevent competition. Either it needs to be legal for other companies to run their own lines, or it needs to be illegal for the line owners to provide data services. (Or the government needs to step in and provide decent telecom service, but that has it's own large set of problems)
No one's ever going to see the reduced accuracy. It's not like people are going at the graph with calipers to evaluate the function. I highly doubt that going from 30kb worth of curve data to 15kb worth of curve data will even be visible at any likely resolution.
How much adminining do they really need to do? Installing desktop apps and updates is trival if they have an internet connection. Even if they don't have an internet connection, working off the install CD isn't that bad. Printers and stuff mostly just work.
I'm not seeing it. What's this task they're going to be doing that'll be any more difficult than Windows XP vs. Windows 2000?
There is only one business model where that even matters, and that business model can be emulated pretty closely with GPL software (i.e. What licence is the Quake 2 source code released under? Can you legally distribute copies of Quake 2 yet? No - the data files are still under pure copyright.) So you'd need to release some code if you made modifications to quake 2 - who cares? Your compeditors will get to start looking at your code when they buy a copy of your game. By the time they can produce a product using that code, your game will be done with its lifecycle anyway.