Obviously the poster is confused, because it was pretty clear to everyone in the room that Allard was referring to Sony when he talked about the 'science fair' approach to platform design. He was contrasting Sony's approach, which he contends is all about theoretically cool hardware that might not be practically usable (i.e. the Cell), versus the MS approach which he contends is more balanced and developer-centric.
In particular, his line about "the only emotion that will evoke is frustration", when referring to the competition's approach, is obviously a reference to the Emotion Engine (remember all the hype surrounding that, and how much of a pig to develop with it turned out to be?).
In my experience doing performance tuning, most games tend to be CPU (and/or memory-bandwidth) bound on their common configurations. Sure, you can always concoct cases where this isn't true (e.g. slow video card in super-fast PC, insane resolutions or pathological scenes), but it does tend to be broadly the case.
This is partly because it's much easier to tune to a GPU budget. On the PC you can recommend different resolutions and/or anti-aliasing modes and instantly have a dramatic impact on fill-rate requirements without sustantially altering how your game plays. You can also add or remove polygons from models, and swap out shader effects until you get something that fits your budget on your target platform.
Tuning for CPU is more difficult, because making a sweeping change is likely to have gameplay impact and is harder to do. Changing how often or how deeply the AI thinks, or the level of sophistication of your physics system, is going to have an impact on gameplay, and is certainly a lot more programmer work than just telling your artists to remove a couple of lights. Coming up with more efficient algorithms that deliver identical results requires a lot more hard thinking - and time for that is limited.
Yes, but he's a big name in the Japanese industry, and was the creative force behind FF.
A Western equivalent might be Will Wright or Sid Meier. It's still big news if one of those guys says they're working on two exclusive titles for a next-gen console. The fact that it might not be The Sims or Civilisation (because those IPs are owned by a company they used to work for) doesn't really dimish the impact.
I think given Xbox's lack of success in Japan, I can't really think of a better statement from Microsoft that they intend to get it right this time around.
Since right now you own neither a PS3 nor a Xbox2 (or whatever it will be called), why is it a slap in the face that a developer announced they are working on titles for Xbox2? Doesn't it merely inform a future purchasing decision?
Sure, but you can only do so much before the numbers don't work. At what point are you taking so much of a loss that it's just not worth it?
Clearly there is some amount of RAM at which you are making the console powerful enough to attract customers, at a competitve pricepoint, and which still enables you to make a long-term profit on the whole project. That amount is not zero, and it's not infinite. It's some number in-between, and that's the value the console vendors arrive at.
It's a balancing act. Subsidize each console too much and it doesn't matter how many you sell, because you'll never make your money back. Subsidize them not enough, and you'll either be charging too much, or they won't be powerful enough to attract sufficient market share.
We can all have our own individual opinions about where the sweet-spot really is, but frankly, this is just one of those big spreadsheet crunching exercises that MBAs are supposed to be good at.
Consoles demand fast memory, and that stuff ain't cheap. If you add $20 to the cost of each console, and you plan to sell 50 million of them, you just took a billion dollars off your bottom line!
You either have to eat that loss (ouch), or increase your prices, which costs you market share.
At some point there is a sweet spot between packing the console with more memory, and ensuring you get the market share you want. I'm guessing that Sony, Nintendo, and the rest run those numbers and it turns out that the sweet-spot is quite a bit less memory than the average PC owner is used to.
Well, that would assume everything that was said about cell processors here was also untrue.
Heh. Looking at your first link, I think you should consider the source a little bit. This is the same guy who believes he knows how to counteract gravity and travel faster than light. So if it's all the same to you, I'll consider his "analysis" of the cell processor with a large dose of salt.
The predicatable part of the schedule for movie production happens after the script has been written and directors and actors signed. Predicting how long it will take for all of the steps after that is a relatively straightforward business.
However, before you get to that stage there are a number of steps which are not predictable. Someone writes the script. Perhaps it goes through re-writes. You need to retain the right director. Casting for principal characters needs to be done. All of those steps are highly variable, and are not easy to predict accurately.
Think of all of those movies you've heard about being "in development" for years. Remember back when the rumour was that James Cameron was going to direct Spiderman? How long has the Hitch-Hiker's movie been talked about? How many attempts at scripts for various movies are abandoned before the final version is settled on?
If you count all of that time, I would argue that while the "production" phase of movie making might be highly predicatable, the whole process, start-to-finish, of making a movie really isn't.
It's the same thing with games. You can have a long "pre-production" phase where you are nailing down all of the core mechanics and getting key systems and pipelines working. This is less predictable because of the iteration required. At some point, you're into "production", where you're just churning out content, and that phase is much more predictable.
Granted, movies are more predictable than games, but it's not quite such a big gap as you might first suppose.
WoW suffers from too many players per server, at least partly as a result of some backfiring of attempts to loadbalance.
Firstly, they didn't release the entire list of server names that would be available on launch day ahead of time, only a subset. Then they are surprised that those servers are overloaded. People agreed with their friends which server they were going to play on, and obviously you can only agree to a server you know exists.
Secondly, classifying the servers by timezone was a mistake. The most overloaded servers are the central zone ones. Why? Because people have friends on both coasts, and decide to plump for the central servers as a compromise. The reality is that geographical location of the server isn't really that important (I've played WoW from the UK, no problem), so the timezone classification didn't really help and in fact probably exacerbated the clumping problem.
And if you RTFA, he says this to back up his estimate:
Launched in November 2004, this is Blizzard's much-anticipated entry into the MMOG market. Numbers are still very preliminary; not everyone who starts playing a MMOG at launch actually "subscribes" the following month. We know for sure that WoW has at least 200,000 subscribers, and as of January 2005 has sold through close to 600,000 copies. Based on a variety of factors such as likely conversion rates and server loads, it is reasonable to believe that World of Warcraft has between 350,000 and 500,000 subscribers as of January 2005. I generally do not put "best guesses" in my data, but given both the interest in and importance of this game, I have put a preliminary figure of 350,000 for now. Until that can be confirmed, I have given the data a Confidence Rating of C.
He knows about the 600,000 figure, he's just saying that the data is sketchy, and not all of those sales will turn into subscriptions. Based on his previous experience analysing these things, his conversative estimate for now is 350,000. Seems fair to me.
This is true for 3D. The 3D output from two different devices is likely to be slightly different due to floating-point rounding issues and the like affecting polygon edges, interpolants, and so forth. They also tend to fudge their anisotropic and/or mip-map sampling in different ways, with varying results. In some cases vendors 'cheat' a little to improve perf, again affecting results.
However, for basic 2D blitting, I would expect identical results. The common operation is a straightforward copy, no math involved. The next most common case is a simple copy with a 1-bit mask, again no room for rounding differences. Anti-aliased line-draws might be very slightly different, but those are pretty rare in the average desktop app. Ditto for alpha blending.
So it's really highly unlikely. If Matrox cards look 'sharper' for regular 2D desktop apps, it has nothing to do with dedicated 2D hardware (a copy is a copy), but rather better RAMDACs and the like.
The sharpness of the output image might have something to do with the quality of the RAMDAC used, but has nothing to do with the presence or otherwise of specialist 2D hardware.
The RAMDAC reads the contents of the framebuffer, which is just a chunk of memory on the card. The 2D or 3D hardware you refer to is merely the equivalent of a CPU which operates on that memory. 2D hardware is optimized for filling and copying rectangular axis-aligned arrays of pixels. 3D hardware is more flexible and in general renders arbitrary triangles which can be used to build up simple rectangles if you choose. However, in both cases they're just whacking bytes in memory. The sharpness of the output image cannot possibly be affected by how those bytes are computed.
The reason for specialist 2D hardware in the past has been that it's possible to optimize for the simple axis-aligned rectangle case, and get extremely high performance with relatively little silicon. These days, there is so much fillrate available in the 3D core, than an optimized piece of silicon to do restricted 2D operations is pointless.
First, the GPU is the processing unit, the framebuffer is the memory where the bits are stored. Both are involved in any kind of rendering operation, 2D or 3D. The GPU operates on the bits on the framebuffer.
Second, modern graphics devices don't have any dedicated 2D hardware left in them. They all just use their 3D cores to do basic blit operations. Why waste silicon on specialist 2D blitting when you've got a gajillion megapixels of fillrate sitting right there in the 3D core?
Third, you are obviously unaware of how modern shader technology works. If I want to stream down 2D coordinates then I can do that just fine. In fact, shaders don't really care what all the numbers are, they just know that they are getting a certain number of inputs. If you choose to write a shader program that interprets them as coordinates to be transformed, then that's merely the common convention. Heck, I could just stream down 1D coordinates if I wanted to (actually, this is genuinely useful, if the coordinate is time and the shader is computing, say, a particle system). So there is really no inefficiency in using the 3D core to do 2D operations, because I can just transmit the minimum amount of data necessary by means of a suitably chosen shader.
They missed out Jade Empire, currently being developed by Bioware (the developers responsible for Baldur's Gate and the original KOTOR), which I believe is due out for Xbox this year (no idea about a PC version).
That's certainly one RPG that I'm looking forward to this year.
They don't have a mandate to do anything; they are simply two private citizens running their website in a way that pleases them. It turns out that sufficient people find their website entertaining that they can turn this endeavour into a business. Good for them.
When you vote for them, or directly pay money for their services, then you get to bitch about them not following their "mandate". Until such time, recognize that Gabe and Tycho are not under any obligation to you, or anyone else, and that you can simply choose to visit their website if you like it, or not, if you don't.
It's legitmate to offer an opinion on the quality of their work, or to express hopes such as "I wish they'd get back to talking about games". However, when you use a word like "mandate" you imply that they have some kind of duty to listen to you, which clearly they don't.
But it doesn't work like that. The attacker looks at all the SSID's being broadcast. In the first place, I don't broadcast my SSID. In the second place, even if I did, don't you think the attacker is going to guess that "default" or "linksys" might be easier targets and try those first?
Two guys are out camping, when one night an angry bear starts trying to get into their tent. The first man quickly grabs his sneakers and starts lacing them up. The second man says "what the hell are you doing? You'll never outrun the bear!", to which the first replies "I don't have to outrun the bear, I just have to outrun you".
The moral of this story is that your security doesn't need to be perfect, it just needs to be 'good enough', and in this case 'good enough' is probably merely 'better than the muppet next door who hasn't secured their network at all'.
I use WEP to secure my wireless LAN. Does it bother me that it's possible to crack? Not really, because there are at least 2 other networks in my apartment building (with SSIDs of 'linksys' and 'default') which don't appear to have any kind of security at all. Which means that someone casually looking for a free connection is going to use them, not me. If someone really wants to compromise my network specifically, and has the time and skill to do so, well, then I have bigger problems...
EA have the exclusive rights, but I assume that also includes the right to sub-license to other companies. In which case you still might see other football games produced, although perhaps not head-on competitors to the Madden franchise, with the 3rd parties licensing the NFL properties from EA.
There is precident for such a thing: EA currently have (or perhaps it might have just expired, I haven't kept track for a while) the exclusive license for Ferrari cars. Other publishers of driving and racing games can and do sub-license from EA when the need arises.
This is even more true in my sport of fencing. Since you face right handed opponents much more often, you get little practice against lefties. When you face a leftie, it's all backwards, which is very confusing. Even if, like me, you're another leftie!
This is even worse than squash (I play squash too). In squash, you just mentally adjust to play to the other side. In fencing the "muscle memory" is all wrong.
This is completely incorrect. The purpose served by the HR department in any large corporation is to protect the company from its employees. Now, sometimes this goal is best served by preventing abuses which may lead to lawsuits, but never forget that at core, HR is there to serve the company not the employees.
I think the point was that although variability might be more realistic, it isn't necessarily better for gameplay, because predictable outcome allows the player to plan. The game then takes on some puzzle-game like qualities.
If you have to choose between realism and good gameplay, good gameplay should win.
It doesn't eliminate all cases, of course, but the/GS compiler flag for Visual C++ does eliminate many of them. In essence, it checks if the return address has been trashed, and throws an exception if it has. Your app still crashes, but that's probably better than being 0wn3d.
Yes, it is possible to circumvent, and there are of course other kinds of attacks/bugs which this doesn't help with. Nor is it a substitute for actually fixing those buffer overflow problems. However, all that said, it's still a good extra level of defense that does improve the security of the system and apps by substantially mitigating a large class of potential bugs.
Slashdot Mirror
Obviously the poster is confused, because it was pretty clear to everyone in the room that Allard was referring to Sony when he talked about the 'science fair' approach to platform design. He was contrasting Sony's approach, which he contends is all about theoretically cool hardware that might not be practically usable (i.e. the Cell), versus the MS approach which he contends is more balanced and developer-centric.
In particular, his line about "the only emotion that will evoke is frustration", when referring to the competition's approach, is obviously a reference to the Emotion Engine (remember all the hype surrounding that, and how much of a pig to develop with it turned out to be?).
In my experience doing performance tuning, most games tend to be CPU (and/or memory-bandwidth) bound on their common configurations. Sure, you can always concoct cases where this isn't true (e.g. slow video card in super-fast PC, insane resolutions or pathological scenes), but it does tend to be broadly the case.
This is partly because it's much easier to tune to a GPU budget. On the PC you can recommend different resolutions and/or anti-aliasing modes and instantly have a dramatic impact on fill-rate requirements without sustantially altering how your game plays. You can also add or remove polygons from models, and swap out shader effects until you get something that fits your budget on your target platform.
Tuning for CPU is more difficult, because making a sweeping change is likely to have gameplay impact and is harder to do. Changing how often or how deeply the AI thinks, or the level of sophistication of your physics system, is going to have an impact on gameplay, and is certainly a lot more programmer work than just telling your artists to remove a couple of lights. Coming up with more efficient algorithms that deliver identical results requires a lot more hard thinking - and time for that is limited.
Star Ocean was developed by TriAce. Square-Enix is the publisher. TriAce is pretty well known in Japan, for titles such as Valkyrie Profile.
(Although, to fair, even though TriAce is a separate company their offices are located in the same building as Square.)
Yes, but he's a big name in the Japanese industry, and was the creative force behind FF.
A Western equivalent might be Will Wright or Sid Meier. It's still big news if one of those guys says they're working on two exclusive titles for a next-gen console. The fact that it might not be The Sims or Civilisation (because those IPs are owned by a company they used to work for) doesn't really dimish the impact.
I think given Xbox's lack of success in Japan, I can't really think of a better statement from Microsoft that they intend to get it right this time around.
Slapped in the face? Why?
Since right now you own neither a PS3 nor a Xbox2 (or whatever it will be called), why is it a slap in the face that a developer announced they are working on titles for Xbox2? Doesn't it merely inform a future purchasing decision?
Best quote from the article:
Their case argues that EA's engineers "do not perform work that is original or creative,"
EA games have no orignality or creativity? Say it ain't so!
On the other hand, the books were already made into a TV and radio show, so it's not much of a secret, is it?
Actually, it was the original radio show which was later turned into a book, and thence a TV show.
Sure, but you can only do so much before the numbers don't work. At what point are you taking so much of a loss that it's just not worth it?
Clearly there is some amount of RAM at which you are making the console powerful enough to attract customers, at a competitve pricepoint, and which still enables you to make a long-term profit on the whole project. That amount is not zero, and it's not infinite. It's some number in-between, and that's the value the console vendors arrive at.
It's a balancing act. Subsidize each console too much and it doesn't matter how many you sell, because you'll never make your money back. Subsidize them not enough, and you'll either be charging too much, or they won't be powerful enough to attract sufficient market share.
We can all have our own individual opinions about where the sweet-spot really is, but frankly, this is just one of those big spreadsheet crunching exercises that MBAs are supposed to be good at.
Cost.
Consoles demand fast memory, and that stuff ain't cheap. If you add $20 to the cost of each console, and you plan to sell 50 million of them, you just took a billion dollars off your bottom line!
You either have to eat that loss (ouch), or increase your prices, which costs you market share.
At some point there is a sweet spot between packing the console with more memory, and ensuring you get the market share you want. I'm guessing that Sony, Nintendo, and the rest run those numbers and it turns out that the sweet-spot is quite a bit less memory than the average PC owner is used to.
Heh. Looking at your first link, I think you should consider the source a little bit. This is the same guy who believes he knows how to counteract gravity and travel faster than light. So if it's all the same to you, I'll consider his "analysis" of the cell processor with a large dose of salt.
I don't think it is true with movies either.
The predicatable part of the schedule for movie production happens after the script has been written and directors and actors signed. Predicting how long it will take for all of the steps after that is a relatively straightforward business.
However, before you get to that stage there are a number of steps which are not predictable. Someone writes the script. Perhaps it goes through re-writes. You need to retain the right director. Casting for principal characters needs to be done. All of those steps are highly variable, and are not easy to predict accurately.
Think of all of those movies you've heard about being "in development" for years. Remember back when the rumour was that James Cameron was going to direct Spiderman? How long has the Hitch-Hiker's movie been talked about? How many attempts at scripts for various movies are abandoned before the final version is settled on?
If you count all of that time, I would argue that while the "production" phase of movie making might be highly predicatable, the whole process, start-to-finish, of making a movie really isn't.
It's the same thing with games. You can have a long "pre-production" phase where you are nailing down all of the core mechanics and getting key systems and pipelines working. This is less predictable because of the iteration required. At some point, you're into "production", where you're just churning out content, and that phase is much more predictable.
Granted, movies are more predictable than games, but it's not quite such a big gap as you might first suppose.
WoW suffers from too many players per server, at least partly as a result of some backfiring of attempts to loadbalance.
Firstly, they didn't release the entire list of server names that would be available on launch day ahead of time, only a subset. Then they are surprised that those servers are overloaded. People agreed with their friends which server they were going to play on, and obviously you can only agree to a server you know exists.
Secondly, classifying the servers by timezone was a mistake. The most overloaded servers are the central zone ones. Why? Because people have friends on both coasts, and decide to plump for the central servers as a compromise. The reality is that geographical location of the server isn't really that important (I've played WoW from the UK, no problem), so the timezone classification didn't really help and in fact probably exacerbated the clumping problem.
And if you RTFA, he says this to back up his estimate:
Launched in November 2004, this is Blizzard's much-anticipated entry into the MMOG market. Numbers are still very preliminary; not everyone who starts playing a MMOG at launch actually "subscribes" the following month. We know for sure that WoW has at least 200,000 subscribers, and as of January 2005 has sold through close to 600,000 copies. Based on a variety of factors such as likely conversion rates and server loads, it is reasonable to believe that World of Warcraft has between 350,000 and 500,000 subscribers as of January 2005. I generally do not put "best guesses" in my data, but given both the interest in and importance of this game, I have put a preliminary figure of 350,000 for now. Until that can be confirmed, I have given the data a Confidence Rating of C.
He knows about the 600,000 figure, he's just saying that the data is sketchy, and not all of those sales will turn into subscriptions. Based on his previous experience analysing these things, his conversative estimate for now is 350,000. Seems fair to me.
This is true for 3D. The 3D output from two different devices is likely to be slightly different due to floating-point rounding issues and the like affecting polygon edges, interpolants, and so forth. They also tend to fudge their anisotropic and/or mip-map sampling in different ways, with varying results. In some cases vendors 'cheat' a little to improve perf, again affecting results.
However, for basic 2D blitting, I would expect identical results. The common operation is a straightforward copy, no math involved. The next most common case is a simple copy with a 1-bit mask, again no room for rounding differences. Anti-aliased line-draws might be very slightly different, but those are pretty rare in the average desktop app. Ditto for alpha blending.
So it's really highly unlikely. If Matrox cards look 'sharper' for regular 2D desktop apps, it has nothing to do with dedicated 2D hardware (a copy is a copy), but rather better RAMDACs and the like.
The sharpness of the output image might have something to do with the quality of the RAMDAC used, but has nothing to do with the presence or otherwise of specialist 2D hardware.
The RAMDAC reads the contents of the framebuffer, which is just a chunk of memory on the card. The 2D or 3D hardware you refer to is merely the equivalent of a CPU which operates on that memory. 2D hardware is optimized for filling and copying rectangular axis-aligned arrays of pixels. 3D hardware is more flexible and in general renders arbitrary triangles which can be used to build up simple rectangles if you choose. However, in both cases they're just whacking bytes in memory. The sharpness of the output image cannot possibly be affected by how those bytes are computed.
The reason for specialist 2D hardware in the past has been that it's possible to optimize for the simple axis-aligned rectangle case, and get extremely high performance with relatively little silicon. These days, there is so much fillrate available in the 3D core, than an optimized piece of silicon to do restricted 2D operations is pointless.
Okay, cluehammer time:
First, the GPU is the processing unit, the framebuffer is the memory where the bits are stored. Both are involved in any kind of rendering operation, 2D or 3D. The GPU operates on the bits on the framebuffer.
Second, modern graphics devices don't have any dedicated 2D hardware left in them. They all just use their 3D cores to do basic blit operations. Why waste silicon on specialist 2D blitting when you've got a gajillion megapixels of fillrate sitting right there in the 3D core?
Third, you are obviously unaware of how modern shader technology works. If I want to stream down 2D coordinates then I can do that just fine. In fact, shaders don't really care what all the numbers are, they just know that they are getting a certain number of inputs. If you choose to write a shader program that interprets them as coordinates to be transformed, then that's merely the common convention. Heck, I could just stream down 1D coordinates if I wanted to (actually, this is genuinely useful, if the coordinate is time and the shader is computing, say, a particle system). So there is really no inefficiency in using the 3D core to do 2D operations, because I can just transmit the minimum amount of data necessary by means of a suitably chosen shader.
They missed out Jade Empire, currently being developed by Bioware (the developers responsible for Baldur's Gate and the original KOTOR), which I believe is due out for Xbox this year (no idea about a PC version).
That's certainly one RPG that I'm looking forward to this year.
Mandate?
They don't have a mandate to do anything; they are simply two private citizens running their website in a way that pleases them. It turns out that sufficient people find their website entertaining that they can turn this endeavour into a business. Good for them.
When you vote for them, or directly pay money for their services, then you get to bitch about them not following their "mandate". Until such time, recognize that Gabe and Tycho are not under any obligation to you, or anyone else, and that you can simply choose to visit their website if you like it, or not, if you don't.
It's legitmate to offer an opinion on the quality of their work, or to express hopes such as "I wish they'd get back to talking about games". However, when you use a word like "mandate" you imply that they have some kind of duty to listen to you, which clearly they don't.
But it doesn't work like that. The attacker looks at all the SSID's being broadcast. In the first place, I don't broadcast my SSID. In the second place, even if I did, don't you think the attacker is going to guess that "default" or "linksys" might be easier targets and try those first?
The moral of this story is that your security doesn't need to be perfect, it just needs to be 'good enough', and in this case 'good enough' is probably merely 'better than the muppet next door who hasn't secured their network at all'.
I use WEP to secure my wireless LAN. Does it bother me that it's possible to crack? Not really, because there are at least 2 other networks in my apartment building (with SSIDs of 'linksys' and 'default') which don't appear to have any kind of security at all. Which means that someone casually looking for a free connection is going to use them, not me. If someone really wants to compromise my network specifically, and has the time and skill to do so, well, then I have bigger problems...
EA have the exclusive rights, but I assume that also includes the right to sub-license to other companies. In which case you still might see other football games produced, although perhaps not head-on competitors to the Madden franchise, with the 3rd parties licensing the NFL properties from EA.
There is precident for such a thing: EA currently have (or perhaps it might have just expired, I haven't kept track for a while) the exclusive license for Ferrari cars. Other publishers of driving and racing games can and do sub-license from EA when the need arises.
This is even more true in my sport of fencing. Since you face right handed opponents much more often, you get little practice against lefties. When you face a leftie, it's all backwards, which is very confusing. Even if, like me, you're another leftie!
This is even worse than squash (I play squash too). In squash, you just mentally adjust to play to the other side. In fencing the "muscle memory" is all wrong.
This is completely incorrect. The purpose served by the HR department in any large corporation is to protect the company from its employees. Now, sometimes this goal is best served by preventing abuses which may lead to lawsuits, but never forget that at core, HR is there to serve the company not the employees.
I think the point was that although variability might be more realistic, it isn't necessarily better for gameplay, because predictable outcome allows the player to plan. The game then takes on some puzzle-game like qualities.
If you have to choose between realism and good gameplay, good gameplay should win.
It doesn't eliminate all cases, of course, but the /GS compiler flag for Visual C++ does eliminate many of them. In essence, it checks if the return address has been trashed, and throws an exception if it has. Your app still crashes, but that's probably better than being 0wn3d.
Yes, it is possible to circumvent, and there are of course other kinds of attacks/bugs which this doesn't help with. Nor is it a substitute for actually fixing those buffer overflow problems. However, all that said, it's still a good extra level of defense that does improve the security of the system and apps by substantially mitigating a large class of potential bugs.