Well, that's not quite the case. My understanding of the unified shaders on the ATI part is that a given pipe can either be working on a pixel or a vertex at any given time, and that the split changes based on workload. So the total floating point performance is just a straight computation based on the number of pipes - no doubling.
On the other hand, I don't see where Sony get to twice the MS number for GPU perf, unless the total number of pipes in the NVidia part is approaching twice the total number of pipes in the ATI part. Which from what I hear is not the case. (Supporting evidence: if that were true, you can be sure Sony would be saying it.)
Also, consider that if you max out the throughput on the vertex pipes, that's typically only in some theoretical case where you are pumping out tiny 1-pixel triangles. Conversely, if you max out the throughput on pixel pipes, that's in a theoretical case where you only have a few very large triangles in your scene. So it's likely that the NVidia part never really gets to max out its throughput all at the same time. Whereas the ATI part can apply all of its power to whatever is needed at the moment.
The likely reality is that Sony have some totally bogus math for computing their performance, and added in a whole bunch of stuff which it would be questionable to count. I'm sure that MS did the same to some degree, but of course, they announced their bullshit number first, giving Sony a clear target to aim at with their bullshit rebuttal.
I've talked to game developers who have knowledge of both systems. And while they say that the Cell genuinely does have more horsepower than the Xbox 360 processor (at least theoretically), when it comes to GPU, it's going to be a wash. This is a much more plausible picture, but of course it's very different than the one Sony is painting (and journalists are swallowing whole).
I'd be interested in seeing the math on this one, to be honest. Notice how Sony have quoted numbers on absolutely everything, except the number of pixel/vertex pipelines on the GPU. Perhaps that's because that particular number isn't really any better than the Xbox 360?
And yet, they're claiming their GPU pumps out twice the floating point performance of the Xbox 360 GPU (if you work backwards from their claims for total system performance). The clock rate of the two GPU is pretty close (500MHz vs. 600MHz), so that extra power must be coming from somewhere else - why don't they say what it is?
Something is fishy here. My guess is that they have done some slightly suspect counting to come up with their total performance number, and in fact the GPU power of the two machines is virtually the same.
The reality is that Sony talks a great talk, just like they did with the PS2. And they have impressive sounding numbers and pre-rendered videos to "prove" their system is vastly superior to the competition. And yet, I bet when the thing finally ships, any difference visible to the end-user is going to be marginal at best.
Y'know, if being hard to program for is a serious issue, then Sony are fucked. Talk to any game developer - three cores is more work, but at least they're symmetrical and have real random access to memory via a cache. The Cell has only one general purpose core, and you have to explicitly shuttle data to and from the embedded memory on the other processing elements, which is hugely annoying. Sure, it's fine for obvious streaming applications like video decoding, but definitely more problematic for other work. Nobody in their right mind thinks the next Playstation is going to be anything but much more difficult to work with than Xbox 360.
Of course, maybe that doesn't matter. After all, the PS2 is generally considered to be the hardest of the current generation to work with, and yet it's still the market leader by a comfortable margin.
Bottom line is that it's all going to come down to having the best games at the right time, and a superior marketing plan. The hardware is almost irrelevant.
Pi crops up in all sorts of mathematical settings, so you can reverse engineer one of those to get back the value of Pi, typically via some sort of series expansion. Here's one way:
For example, consider the trignometric function cosine, and its inverse arccosine. cosine(pi) is -1, and therefore arccosine(-1) is pi (taking the usual principal value for arccosine).
Rearrange (subtract pi/2 from each side), and you get:
pi/2 = 1 + 3/6 + 3/40 + 5/112...
Pump out as many terms of the series as you need for your desired precision, multply by 2, and you have pi.
This is actually a pretty *bad* way of computing pi, because it converges very slowly (there are faster converging series expansions which yield other convenient rational multiples of pi), but it will work eventually, and it illustrates the idea. There are other fancier techniques too, but they all work on the basic idea that pi creeps into a lot of basic mathematics, and you can 'unwind' those identities to get pi out with careful rearrangement.
From the other side, homosexual 'activists' (term used loosely here) need to remember that religion does not 'hate' gay people. They need to accept that religions say that homosexuality isn't part of a religious life. I get pissed when I see activists blasting religious people, and I get equally pissed when the reverse happens.
It's an easily observable fact that there are many people in the US (and elsewhere) who lobby quite hard to deny homosexuals rights afforded to other citizens, and who cite a religious motivation for those efforts. I'm not saying that you, personally, fall into those groups, but it's pretty hard to deny that they do exist.
To those on the receiving end of such efforts, hair-splitting distinctions along the lines of "we don't hate you, we just hate what you do" appear self-serving at best. It's a cheap get-out clause, allowing the perpetrators to sidestep otherwise awkward moral questions, at least to themselves.
It is faith plain and simple. I say that gay is wrong because it says so in the Bible. It does not say that black or any other skin color is wrong.
It's a thought experiment. I realise that the Bible doesn't actually say that. I'm asking you to imagine how your argument would run if you substituted 'gay' for 'black'. Would the mere fact of its appearance in the Bible exempt it from being bigotry in that case too? I would argue not.
Just because it's your faith, doesn't automatically exempt it from also being bigotry. Quite the reverse, in fact.
2. A person obstinately and unreasonably wedded to a particular religious creed, opinion, or ritual.
Being obstinately wedded to the opinion that being gay is 'wrong' is, I would argue, unreasonable (i.e. not supported by a rational basis), and therefore bigoted, by dictionary definition.
It's an accurate word to use. If it makes you uncomfortable to be described that way then the solution is not to attempt to redefine the word to include a 'religion exemption', but instead to question your own position.
Yeah, you're probably right. I wouldn't know, I'm not a Christian. Don't see how it affects my argument, though.
My point is that bigotry is orthogonal to religious belief. GP poster was essentially saying "I'm not bigoted, I'm just a good Christian". I'm merely pointing out that a religious motivation doesn't magically turn a bigoted belief into a non-bigoted one.
I do however still believe that being gay is wrong. I have not discriminated against anyone by saying that being gay is wrong. Even a gay person. No one says that someone discriminates if someone says stealing is wrong. The analogy is the same. People falsly believe that tolerance is the same thing as acceptance. You can tolerate and still not accept. I would fight for a gay persons right to do the same things that I do, but I STILL would tell a gay person to their face that what they do is wrong, please change your ways.
In the first place, your stealing analogy is broken. Theft negatively affects other individuals and society as a whole, which is why we criminalize it. Being gay does not adversely affect others, and isn't really anyone elses's damn business.
In the second place, I have a quick question for you. What happens if you substitute "black" for "gay" in your little rant? Would it be okay to go around telling black people that you tolerate but don't accept them, and that they should change their ways? Or would that be bigoted? If you believe, as most people do, that it would in fact be bigoted, please explain why you believe it's not bigoted to treat gays in that way?
Before the eyes of God, if we have done something that the Bible says not to do then we are WRONG. That is not bigoted or anything else.
Yes, it is. You are trying to argue that because your motivation is religious, that your bigotry isn't bigoted. This is an invalid argument; the origin of your belief is irrelevant. A bigoted belief is bigoted whether it's based on religious doctrine, personal whim, or the phase of the moon.
Being gay is wrong. It says so in the Bible. Just because "your" views may differ doesn't mean that I am a bigot, homophobe, or anything else. It simply means that I am Christian.
Just because it's your religious belief, doesn't mean it's not bigoted. Being a Christian is not a free pass. The origin of your belief is irrelevant - if you believe that being gay is wrong, then are are a bigot, more or less by definition.
This is one problem with religions in general. They teach people to discriminate - homosexuality is a frequent target, but some religions also discriminate against women or ethnic minorities. When they do those things, it's bigotry. Just because it's religiously motivated doesn't make it any less repellant.
Your analogy with stealing is flawed. Theft affects others, which is why we consider it wrong (and make it criminal). Being gay does not, and frankly shouldn't be anyone else's business.
Actually, I wouldn't describe Chris Satchell as a marketing jackass. I happen to know Chris personally, and his background is as a games programmer (you know, actually writing code). He's worked his way up through the ranks, and is now a general manager at Microsoft (deservedly so, in my opinion).
Having said that, I'm definitely going to give him some shit about that interview, because I was drowning in all the buzzwords...
I guess I was responding to the mass of posts that were making the poor case I was arguing against, and lumped yours in with them by mistake when selecting one to reply to. Apologies.
No. The argument that because MS have already made a ton of money on Halo 2 they have therefore already 'paid for' future expansion development is completely specious.
The game has already been developed, and people have already bought it. That money is already in the coffers, regardless of whether they do an expansion or not. It's like the concept of 'sunk costs', only in reverse.
The question is, if MS/Bungie spend the incremental cost to produce expansion levels, do they get enough additional revenue as a result to justify it. The expansion has to pay for itself on those terms, not based on revenue already banked.
That additional revenue might come directly, or might come in the form of additional sales of the parent product. You can also factor in the value of customer goodwill and perpetuation of the brand. It's still probably a good calculation for them, but that's not directly related to the amount of money they have already made.
Sure, I thought about mentioning Perl as another productive language which isn't C++, but I just wanted to focus on refuting the notion that STL somehow makes C/C++ a safe language.
Of course, Perl has other drawbacks. The syntax and built-in stuff has evolved over time to provide quick ways of solving particular problems, but it's not really that intuitive. It's possible to write readable code in Perl, but there's something about the language which appears to discourage it.
I think Perl is great when you stick to its strengths. As a shell-scripting language on steroids, with awesome text-munging capabilities, it's hard to beat. For building large component-based systems, I would say it's a little unwieldy.
Overall, I think that C# is probably a better thought-out general purpose language. But Perl definitely has its uses in some cases.
Either case is, as you note, going to be safer than C/C++.
With C#, you can use this using keyword and the IDisposable pattern:
public void f() {
using ( MyResource resource = new MyResource() )
{
resource.SomethingThatMightThrow();
resource.SomethingElseThatMightThrow();
} }
So long as MyResource supports the IDisposable interface, then the using block calls Dispose on the object and the end of the block, including if an exception is thrown.
And then I want to pass my std::string to some legacy library, so I get a char* from it. And some other piece of code uses that pointer to start doing 'interesting' stuff.
Or I forget what the semantics are for iterator invalidation when traversing a STL container, and start using an invalid iterator somewhere in my code. (Quick: without looking it up, which containers invalidate iterators on deletion and which don't?)
What happens? Maybe an exception gets thrown? Maybe you get unlucky and start stomping on something else's memory? Maybe you'll notice the bug right away, or maybe it won't surface for weeks? Who knows?
As an experienced C++ programmer (I've been writing C++ code since the Zortech compilers in the early nineties), who's done a fair amount of STL programming, I can absolutely tell you that it's quite possible to write buggy memory-stomping code in C++, STL or no. With C# or Java, that's simply not the case (or, at least it's orders of magnitude harder).
Except that if your third party library doesn't work with your garbage collected extension to C++ then you're hosed..Net and Java have complete environments, including fully featured library sets which are designed to work this way.
More modern languages like Java and C# also tend to be just a little bit easier to work with, with handy constructs like foreach, properties, delegates and events. Sure, it's ultimately just syntactic sugar, but then aren't all languages just syntactic sugar?
Yes, it's true that there are other resources to worry about besides memory. C# has the IDisposable pattern and the 'using' keyword to help here (I'm not sure what the Java equivalent is, but I assume there is one). These environments are not a silver bullet, but they're not any worse than C/C++.
Your point about deterministically called destructors is valid, but I think you're looking at things from a C++-centric viewpoint. In C++ that's the canoncial way to manage resource lifetime, so you try to use the same mechanism in C# and discover that it's not the right thing to do because the behaviour is different. That doesn't mean that C# is broken, just that you might need to use a different approach.
Programmer productivity is higher in garbage-collected (MS calles them "managed") environments such as Java and.Net/C#.
In addition, it's much harder to make programming blunders such as overstepping the bounds of an array or string, which can in turn lead to security vulnerabilities.
Those two reasons alone are enough to favour a Java/C#-type approach in situations where absolutely bleeding-edge performance isn't a requirement (i.e. almost all of them).
I was rather thinking that imagination might come up with different kinds of gameplay.
How much processing power is needed for the gameplay portion of Katamari Damacy, Tetris, or Parapper the Rapper? The genuine innovations in gameplay have not, as far as I can tell, really come from doing more sophisticated AI, but rather from a designer (not a programmer) using their imagination to come up with a new idea.
In addition, remember that you are getting something back for being in-order: To tackle your traditional AI worry, recall that some AI problems do lend themselves to parallelisation. The tradeoff for losing out-of-order execution is that you get to have multiple-cores. If you have 100 agents that all need to do pathfinding, well, that's an embarrasingly parallel problem and this new hardware is probably well suited for it.
Basically, I think that traditional single-threaded spaghetti code is going to be no slower than before (due to higher clock-rates). Straight-line number crunching is going to be massively faster. And any problem you can make parallel will benefit from multiple cores. That's not to terrible a situation to be in, when you consider the alternative choices.
What kind of hardware are you supposed to build, then? Hardware that you know solves today's problems better than before, and that might solve new problems okay too. Or hardware which doesn't solve today's problems any better, and which doesn't guarantee better solutions to tomorrow's problems either.
His argument is basically that the new 'in-order' chips are not going to be any faster on spaghetti game code, and that all they're really going to be better at is high-volume number-crunching for graphics and physics. And that this is somehow going to lead to worse games.
Obviously he hasn't looked at the performance profile for a game recently. The gnarly game logic doesn't really take up much of the time. The heavy-duty number-crunching is where all the cycles go. So, in fact, it's exactly the correct tradeoff to design hardware which makes those bits get faster, because those are the performance sensitive bits.
In addition, it's clear how throwing more processing power at graphics and physics makes for better graphics and physics. Whereas it's not at all clear that more processing power leads to better gamplay. You don't need more clock-cycles to make a more interesting game, just more imagination. So complaining that there isn't any more processing power available for 'gameplay-type' code is kind of a pointless complaint.
It was actually pretty interesting, showing some new DS "games" their working on. (I say "games", because they were really interactive...things, rather than games in the traditional sense - one was a pet dog simulator, and another was an abstract music play toy.)
They also showed Mario Kart multiplayer wirelessly on the DS, and a preview video of the new Zelda game.
Slashdot Mirror
Well, that's not quite the case. My understanding of the unified shaders on the ATI part is that a given pipe can either be working on a pixel or a vertex at any given time, and that the split changes based on workload. So the total floating point performance is just a straight computation based on the number of pipes - no doubling.
On the other hand, I don't see where Sony get to twice the MS number for GPU perf, unless the total number of pipes in the NVidia part is approaching twice the total number of pipes in the ATI part. Which from what I hear is not the case. (Supporting evidence: if that were true, you can be sure Sony would be saying it.)
Also, consider that if you max out the throughput on the vertex pipes, that's typically only in some theoretical case where you are pumping out tiny 1-pixel triangles. Conversely, if you max out the throughput on pixel pipes, that's in a theoretical case where you only have a few very large triangles in your scene. So it's likely that the NVidia part never really gets to max out its throughput all at the same time. Whereas the ATI part can apply all of its power to whatever is needed at the moment.
The likely reality is that Sony have some totally bogus math for computing their performance, and added in a whole bunch of stuff which it would be questionable to count. I'm sure that MS did the same to some degree, but of course, they announced their bullshit number first, giving Sony a clear target to aim at with their bullshit rebuttal.
I've talked to game developers who have knowledge of both systems. And while they say that the Cell genuinely does have more horsepower than the Xbox 360 processor (at least theoretically), when it comes to GPU, it's going to be a wash. This is a much more plausible picture, but of course it's very different than the one Sony is painting (and journalists are swallowing whole).
I'd be interested in seeing the math on this one, to be honest. Notice how Sony have quoted numbers on absolutely everything, except the number of pixel/vertex pipelines on the GPU. Perhaps that's because that particular number isn't really any better than the Xbox 360?
And yet, they're claiming their GPU pumps out twice the floating point performance of the Xbox 360 GPU (if you work backwards from their claims for total system performance). The clock rate of the two GPU is pretty close (500MHz vs. 600MHz), so that extra power must be coming from somewhere else - why don't they say what it is?
Something is fishy here. My guess is that they have done some slightly suspect counting to come up with their total performance number, and in fact the GPU power of the two machines is virtually the same.
The reality is that Sony talks a great talk, just like they did with the PS2. And they have impressive sounding numbers and pre-rendered videos to "prove" their system is vastly superior to the competition. And yet, I bet when the thing finally ships, any difference visible to the end-user is going to be marginal at best.
Y'know, if being hard to program for is a serious issue, then Sony are fucked. Talk to any game developer - three cores is more work, but at least they're symmetrical and have real random access to memory via a cache. The Cell has only one general purpose core, and you have to explicitly shuttle data to and from the embedded memory on the other processing elements, which is hugely annoying. Sure, it's fine for obvious streaming applications like video decoding, but definitely more problematic for other work. Nobody in their right mind thinks the next Playstation is going to be anything but much more difficult to work with than Xbox 360.
Of course, maybe that doesn't matter. After all, the PS2 is generally considered to be the hardest of the current generation to work with, and yet it's still the market leader by a comfortable margin.
Bottom line is that it's all going to come down to having the best games at the right time, and a superior marketing plan. The hardware is almost irrelevant.
Or, here's another empirical way of doing it.
Step 1: Get a big box of needles
Step 2: Get a big piece of paper, and draw parallel lines on it, with spacing equal to the length of a needle.
Step 3: Randomly drop all the needles on the paper.
Step 4: Count how many needles lie crossing a line. Divide by total number of needles.
Step 5: Multiply the result you got in step 4 by 4.
Pi crops up in all sorts of mathematical settings, so you can reverse engineer one of those to get back the value of Pi, typically via some sort of series expansion. Here's one way:
...
...
For example, consider the trignometric function cosine, and its inverse arccosine. cosine(pi) is -1, and therefore arccosine(-1) is pi (taking the usual principal value for arccosine).
Now form the series expansion for arccosine:
arccos(x) = pi/2 - x - x^3/6 - 3x^5/40 - 5x^7/112
Rearrange (subtract pi/2 from each side), and you get:
pi/2 = 1 + 3/6 + 3/40 + 5/112
Pump out as many terms of the series as you need for your desired precision, multply by 2, and you have pi.
This is actually a pretty *bad* way of computing pi, because it converges very slowly (there are faster converging series expansions which yield other convenient rational multiples of pi), but it will work eventually, and it illustrates the idea. There are other fancier techniques too, but they all work on the basic idea that pi creeps into a lot of basic mathematics, and you can 'unwind' those identities to get pi out with careful rearrangement.
From the other side, homosexual 'activists' (term used loosely here) need to remember that religion does not 'hate' gay people. They need to accept that religions say that homosexuality isn't part of a religious life. I get pissed when I see activists blasting religious people, and I get equally pissed when the reverse happens.
It's an easily observable fact that there are many people in the US (and elsewhere) who lobby quite hard to deny homosexuals rights afforded to other citizens, and who cite a religious motivation for those efforts. I'm not saying that you, personally, fall into those groups, but it's pretty hard to deny that they do exist.
To those on the receiving end of such efforts, hair-splitting distinctions along the lines of "we don't hate you, we just hate what you do" appear self-serving at best. It's a cheap get-out clause, allowing the perpetrators to sidestep otherwise awkward moral questions, at least to themselves.
It is faith plain and simple. I say that gay is wrong because it says so in the Bible. It does not say that black or any other skin color is wrong.
It's a thought experiment. I realise that the Bible doesn't actually say that. I'm asking you to imagine how your argument would run if you substituted 'gay' for 'black'. Would the mere fact of its appearance in the Bible exempt it from being bigotry in that case too? I would argue not.
Just because it's your faith, doesn't automatically exempt it from also being bigotry. Quite the reverse, in fact.
2. A person obstinately and unreasonably wedded to a particular religious creed, opinion, or ritual.
Being obstinately wedded to the opinion that being gay is 'wrong' is, I would argue, unreasonable (i.e. not supported by a rational basis), and therefore bigoted, by dictionary definition.
It's an accurate word to use. If it makes you uncomfortable to be described that way then the solution is not to attempt to redefine the word to include a 'religion exemption', but instead to question your own position.
Yeah, you're probably right. I wouldn't know, I'm not a Christian. Don't see how it affects my argument, though.
My point is that bigotry is orthogonal to religious belief. GP poster was essentially saying "I'm not bigoted, I'm just a good Christian". I'm merely pointing out that a religious motivation doesn't magically turn a bigoted belief into a non-bigoted one.
I do however still believe that being gay is wrong. I have not discriminated against anyone by saying that being gay is wrong. Even a gay person. No one says that someone discriminates if someone says stealing is wrong. The analogy is the same. People falsly believe that tolerance is the same thing as acceptance. You can tolerate and still not accept. I would fight for a gay persons right to do the same things that I do, but I STILL would tell a gay person to their face that what they do is wrong, please change your ways.
In the first place, your stealing analogy is broken. Theft negatively affects other individuals and society as a whole, which is why we criminalize it. Being gay does not adversely affect others, and isn't really anyone elses's damn business.
In the second place, I have a quick question for you. What happens if you substitute "black" for "gay" in your little rant? Would it be okay to go around telling black people that you tolerate but don't accept them, and that they should change their ways? Or would that be bigoted? If you believe, as most people do, that it would in fact be bigoted, please explain why you believe it's not bigoted to treat gays in that way?
Before the eyes of God, if we have done something that the Bible says not to do then we are WRONG. That is not bigoted or anything else.
Yes, it is. You are trying to argue that because your motivation is religious, that your bigotry isn't bigoted. This is an invalid argument; the origin of your belief is irrelevant. A bigoted belief is bigoted whether it's based on religious doctrine, personal whim, or the phase of the moon.
Being gay is wrong. It says so in the Bible. Just because "your" views may differ doesn't mean that I am a bigot, homophobe, or anything else. It simply means that I am Christian.
Just because it's your religious belief, doesn't mean it's not bigoted. Being a Christian is not a free pass. The origin of your belief is irrelevant - if you believe that being gay is wrong, then are are a bigot, more or less by definition.
This is one problem with religions in general. They teach people to discriminate - homosexuality is a frequent target, but some religions also discriminate against women or ethnic minorities. When they do those things, it's bigotry. Just because it's religiously motivated doesn't make it any less repellant.
Your analogy with stealing is flawed. Theft affects others, which is why we consider it wrong (and make it criminal). Being gay does not, and frankly shouldn't be anyone else's business.
Actually, I wouldn't describe Chris Satchell as a marketing jackass. I happen to know Chris personally, and his background is as a games programmer (you know, actually writing code). He's worked his way up through the ranks, and is now a general manager at Microsoft (deservedly so, in my opinion).
Having said that, I'm definitely going to give him some shit about that interview, because I was drowning in all the buzzwords...
Oops, my bad.
I guess I was responding to the mass of posts that were making the poor case I was arguing against, and lumped yours in with them by mistake when selecting one to reply to. Apologies.
No. The argument that because MS have already made a ton of money on Halo 2 they have therefore already 'paid for' future expansion development is completely specious.
The game has already been developed, and people have already bought it. That money is already in the coffers, regardless of whether they do an expansion or not. It's like the concept of 'sunk costs', only in reverse.
The question is, if MS/Bungie spend the incremental cost to produce expansion levels, do they get enough additional revenue as a result to justify it. The expansion has to pay for itself on those terms, not based on revenue already banked.
That additional revenue might come directly, or might come in the form of additional sales of the parent product. You can also factor in the value of customer goodwill and perpetuation of the brand. It's still probably a good calculation for them, but that's not directly related to the amount of money they have already made.
Sure, I thought about mentioning Perl as another productive language which isn't C++, but I just wanted to focus on refuting the notion that STL somehow makes C/C++ a safe language.
Of course, Perl has other drawbacks. The syntax and built-in stuff has evolved over time to provide quick ways of solving particular problems, but it's not really that intuitive. It's possible to write readable code in Perl, but there's something about the language which appears to discourage it.
I think Perl is great when you stick to its strengths. As a shell-scripting language on steroids, with awesome text-munging capabilities, it's hard to beat. For building large component-based systems, I would say it's a little unwieldy.
Overall, I think that C# is probably a better thought-out general purpose language. But Perl definitely has its uses in some cases.
Either case is, as you note, going to be safer than C/C++.
And then I want to pass my std::string to some legacy library, so I get a char* from it. And some other piece of code uses that pointer to start doing 'interesting' stuff.
Or I forget what the semantics are for iterator invalidation when traversing a STL container, and start using an invalid iterator somewhere in my code. (Quick: without looking it up, which containers invalidate iterators on deletion and which don't?)
What happens? Maybe an exception gets thrown? Maybe you get unlucky and start stomping on something else's memory? Maybe you'll notice the bug right away, or maybe it won't surface for weeks? Who knows?
As an experienced C++ programmer (I've been writing C++ code since the Zortech compilers in the early nineties), who's done a fair amount of STL programming, I can absolutely tell you that it's quite possible to write buggy memory-stomping code in C++, STL or no. With C# or Java, that's simply not the case (or, at least it's orders of magnitude harder).
Except that if your third party library doesn't work with your garbage collected extension to C++ then you're hosed. .Net and Java have complete environments, including fully featured library sets which are designed to work this way.
More modern languages like Java and C# also tend to be just a little bit easier to work with, with handy constructs like foreach, properties, delegates and events. Sure, it's ultimately just syntactic sugar, but then aren't all languages just syntactic sugar?
Yes, it's true that there are other resources to worry about besides memory. C# has the IDisposable pattern and the 'using' keyword to help here (I'm not sure what the Java equivalent is, but I assume there is one). These environments are not a silver bullet, but they're not any worse than C/C++.
Your point about deterministically called destructors is valid, but I think you're looking at things from a C++-centric viewpoint. In C++ that's the canoncial way to manage resource lifetime, so you try to use the same mechanism in C# and discover that it's not the right thing to do because the behaviour is different. That doesn't mean that C# is broken, just that you might need to use a different approach.
Programmer productivity is higher in garbage-collected (MS calles them "managed") environments such as Java and .Net/C#.
In addition, it's much harder to make programming blunders such as overstepping the bounds of an array or string, which can in turn lead to security vulnerabilities.
Those two reasons alone are enough to favour a Java/C#-type approach in situations where absolutely bleeding-edge performance isn't a requirement (i.e. almost all of them).
I was rather thinking that imagination might come up with different kinds of gameplay.
How much processing power is needed for the gameplay portion of Katamari Damacy, Tetris, or Parapper the Rapper? The genuine innovations in gameplay have not, as far as I can tell, really come from doing more sophisticated AI, but rather from a designer (not a programmer) using their imagination to come up with a new idea.
In addition, remember that you are getting something back for being in-order: To tackle your traditional AI worry, recall that some AI problems do lend themselves to parallelisation. The tradeoff for losing out-of-order execution is that you get to have multiple-cores. If you have 100 agents that all need to do pathfinding, well, that's an embarrasingly parallel problem and this new hardware is probably well suited for it.
Basically, I think that traditional single-threaded spaghetti code is going to be no slower than before (due to higher clock-rates). Straight-line number crunching is going to be massively faster. And any problem you can make parallel will benefit from multiple cores. That's not to terrible a situation to be in, when you consider the alternative choices.
What kind of hardware are you supposed to build, then? Hardware that you know solves today's problems better than before, and that might solve new problems okay too. Or hardware which doesn't solve today's problems any better, and which doesn't guarantee better solutions to tomorrow's problems either.
His argument is basically that the new 'in-order' chips are not going to be any faster on spaghetti game code, and that all they're really going to be better at is high-volume number-crunching for graphics and physics. And that this is somehow going to lead to worse games.
Obviously he hasn't looked at the performance profile for a game recently. The gnarly game logic doesn't really take up much of the time. The heavy-duty number-crunching is where all the cycles go. So, in fact, it's exactly the correct tradeoff to design hardware which makes those bits get faster, because those are the performance sensitive bits.
In addition, it's clear how throwing more processing power at graphics and physics makes for better graphics and physics. Whereas it's not at all clear that more processing power leads to better gamplay. You don't need more clock-cycles to make a more interesting game, just more imagination. So complaining that there isn't any more processing power available for 'gameplay-type' code is kind of a pointless complaint.
FYI, Nintendo's rah-rah keynote was today.
It was actually pretty interesting, showing some new DS "games" their working on. (I say "games", because they were really interactive...things, rather than games in the traditional sense - one was a pet dog simulator, and another was an abstract music play toy.)
They also showed Mario Kart multiplayer wirelessly on the DS, and a preview video of the new Zelda game.
Nintendo's marketing machine is alive and well.