You know why you never hear about the "secular left"? Because by and large, we mind our own business. You don't have leftists putting up shrines to whomever they worship on state property. Down here in Atlanta, I don't get leftists trying to convert me to secularism a couple of times a year. I don't get idiot 16-year old leftist kids coming to my door telling me about their resurrected prophet (16! what the hell does a 16 year old know about ANYTHING, much less the underlying truth of the universe???) Secular leftists don't cripple the major throughfare near my house every weekend morning driving their giant SUVs to their place of worship. Secular leftists don't send me mail claiming that they know the path I should follow to redemption.
All in all, secular leftists are nice, happy people. We only get our panties in a bunch when the Religious Right decides to butt into everyone's lives.
Are you joking? What training do you have in evolutionary biology to be able to discredit it off-hand? What peer-reviewed research have you done to come to your conclusion? Did you spend the best years of your life in school getting a PHD, before even being allowed to do research in the field? I can just imagine how galling it must be to be a biologist right now. I know if some know-nothing came up to me talking non-sense about fluid dynamics (I'm an aerospace engineer), I'd be more than a bit pissed off.
I'm reminded of a line from A Few Good Men. To paraphrase Jack:
" I have neither the time nor the inclination to explain myself to a man who rises and sleeps under the blanket of the very technology/progress/knowledge that I provide, and then questions the manner in which I provide it! I would rather you just said, "Thank you," and went on your way."
Actually, it's true. Both are somewhere around 2%. The thing Americans often forget is that while Macs are fairly common in the US, they're almost absent in the rest of the world, while Linux has had very significant penetration, even on the desktop, in Latin America, Europe, and Asia.
OK, I think I get your point now, and you are correct in a fairly narrow use case: cases where context switching is a more significant cost, than other factors, and that means essentially server applications that spawn VERY large numbers of threads. Yup, user mode threads a-la OS X have lower performance than say linux kernel threads. On the other hand, they're much less likely to take your machine down if used innapropriately.
OS X uses kernel threads. A POSIX thread maps directly to a Mach thread.
For typical desktop situations, I dispute your claim: no one should be writing real world desktop applications that spawn huge numbers of threads.
Highly multi-threaded programs are the natural result of certain programming paradigms, such as the actor model. If you can implement very cheap threads, why limit the possibilities available to the programmer? In any case, whether or not highly multithreaded programs are useful on the desktop is irrelevent. My claim wasn't that OS X should have a highly capable SMP system, but rather that it was highly misleading to claim that it did, as the original article stated. It's downright dishonest to claim that OS X is at the leading edge when it comes to being able to take advantage of multicore technology, and even more dishonest to claim that this has been the case since "day one".
Overall, the benefit of a threading API, is not merely the implementation, it is also the convenience and protection it offers the programmer. I don't believe that applications with, say, tens of threads are significantly poorer performers on OS X than NT (especially with true SMP PPC systems vs. multicore x86). Indeed there's a reason why NT threads can switch faster, and that is largely due to the crude implementation of virtual memory.
Bullshit. Virtual memory isn't implemented any more crudely in x86/NT than it is in OS X/PPC. The "crude" implementation certainly isn't a performance advantage. Indeed, x86 lacks many of the features that other architectures have for implementing fast context switches. For example, a context switch on x86 results in a TLB flush, because x86 lacks support for tagged TLBs.
The reason OS X's multithreaded performance sucks is because it's just very ancient code that doesn't implement modern mechanisms used in highly concurrent kernels. The kernel locking is very coarse even in Tiger (it was worse in Panther, with just two locks for the whole kernel), and the locking scheme doesn't use any advanced techniques like RCU, lockless algorithms, etc. Darwin still uses the old 4BSD scheduler, which isn't O(1), and threading calls are still heavyweight because a substantial amount of the pthreads implementation results in Mach IPC, which is quite slow.
Again, not that this is really a problem on the desktop, but I never claimed that it was. OS X is an excellent desktop OS with very good performance for desktop apps. It just so happens that the performance of the underlying kernel is largely irrelevent for desktop apps, so you can get away with a pretty shitty kernel (which Darwin is), as long as you have good userspace libraries (fast toolkits, etc).
Apple's strategy is not to do what the competition does.
Not since they started shpping Intel hardware. Now, they have to deliver comparable hardware at a comparable pricepoint. You can bet they wanted to ship Core Solos all across the Mac Mini and Macbook lines, and that's what they would've done in the PowerPC era. But they can't do that anymore, because being on the same architecture makes direct comparison possible, and shipping substantially less than your competitors at a higher price does not result in favorable comparisons.
Apple shipped two Yonahs in desktop systems. No other major player (including Dell) did that.
Only because they wanted to skip the P4 entirely, which is a luxury they had given their late entry into the market.
For example, they already make an iMac with integrated graphics. It was released a couple weeks ago.
That was at a vastly lower pricepoint. A $900 education machine with integrated graphics is not ridiculous. A $1300 general desktop with integrated graphics is.
Merom is the same price as Yonah for any MHz rate, at least in the markets we see (who knows what Apple pays). I can't see how Conroe would be cheaper.
Yonah is a laptop chip. Conroe is a desktop chip. Desktop chips, by virtue of their looser power envelope requirements, are generally significantly cheaper.
First of all the iMac G5 started with the 90+W G5. Second of all, Apple completely revamped the cooling zones inside the iMac when it went Intel.
The 970FX is a 60W chip. So is Conroe. I highly doubt Apple completely revamped the iMac case cooling to make it less capable. They knew Conroe was coming out, that the Core Duo was a stopgap chip, and I'm sure they planned the case design with Conroe in mind. They'd have been stupid not to.
Yes, their all POSIX, but we're talking about the implementation, not the API. OS X's pthreads implementation is nowhere near as scalable as NT's or Linux's. It's completely wrong, thus, to say that OS X is an industry leader in thread support.
Nope. iMac will get Conroe and keep discrete graphics for three reasons:
1) Conroe is cheaper than Merom for a given MHz level; 2) Conroe sits within the 60W thermal envelope of the 970FX in the G5 iMacs; 3) Apple's competitors will have discrete graphics in the iMac's price class.
The iMac is Apple's top desktop Mac. The PowerMac is a workstation line, and will be further differentiated as such by moving to Woodcrest. The Mac Mini will take the lower end of the market. In order to be competitive with other vendors pushing Intel machines in the $1300-$1700 category, Conroe and discrete graphics are a must.
It's interesting to note how Apple users have continually low-balled Apple's configurations during the Intel transition. Everyone thought that the Mac Minis would be Core Solo-only, and that only the highest end Macbook would have a Core Duo, if a Duo was offered in the line at all. I think its a holdover from G4/G5 era "differentiation" thinking. The era of arbitrary differentiation is over at Apple. If Dell is shipping Meroms in its $1200 laptops, so will Apple. If Dell is offering discrete graphics and Conroe processors in its $1500 desktops, so will Apple.
This is precisely the strategy Apple has followed thus far with the Intel transition. The Apple products are at the high end of their price classes, with the price buoyed up with features that are usually extra-cost options for other machines, but they feature competitive configurations at each of those price-points. They know very well that they can sell an iMac that offers more features than a Dell that's $200 cheaper, but not an iMac that costs the same and offers a much lesser processor or GPU.
Even when lots of different applications, many of them efficiently multi-threaded, are run on 10.4.5 or 10.4.6 only the first two CPUs are used efficiently while the third and fourth are getting plenty of work....but aren't quite living up to their full potential. Each added core after four seems to drop off in efficiency....not because OS X doesn't handle lots of processors properly, it does. In fact it's an industry leader in terms of being ready for the next generation of multicore, multiprocessor technology. It has been since day one and Apple has consistently kept it at the leading edge since then.
OS X doesn't handle lots of processors properly, and in no way is it anything resembling an industry leader in being ready for multicore technology. That's not the case in Tiger, and it certainly wasn't the case from "day one". OS X's SMP support was deficient pre-Tiger, and its merely "barely adequate" in Tiger. Of the major kernels in use, OS X's is near the back in regards to SMP and thread support, behind NT, Solaris, AIX, HP-UX, Linux, BeOS, and FreeBSD. The only thing its ahead of are kernels like NetBSD and OpenBSD, which still use a single kernel lock.
Apple can improve this situation in Tiger, but there's no secret technology Apple is holding close to their chest that'd give them the lead, as the article claims. Threading and SMP will get better in Tiger, maybe even up to Linux 2.4 levels, but its not being to offer world-beater performance.
Bullshit, on multiple levels. First, it wasn't even until 10.4 that OS X's SMP was good enough for machines with > 2 CPUs, and even in 10.4 OS X's SMP support is far behind that of Linux, FreeBSD, or Solaris.
The Sims is the best selling game of all time... on the PC. However, the PC gaming market is only 1/4th as large as the console gaming market. Moreover, while the Sims may have outsold specific games in other genres, its sales are still small compared to the overall sales of games in other genres. That is to say its very hard to extrapolate the size of the casual gaming market just from sales of one game. The Sims may have sold 16 million copies, but hundreds of millions of other games are being sold in all the non-casual genres.
It is very possible that sales of "The Sims" might hint at a large untapped market that exists. However, until that market starts buying games at the same rate as the "hardcore" market, it's silly to say that its a larger (ie: potentially more profitable) one.
Actually, check the stats sometime. The casual gaming market (as represented by the sims, etc) is quite a bit smaller than the "hardcore" gaming market. As for non-gamers --- well, they don't play games. It's possible that Nintendo might get them to play games, but that's a big "if".
Says the Nintendo fan? Who in the last two generations (N64 and GC) had a console that could do pretty games, but with precious little good content to play on them?
2% my ass. The difference between HD and SD is night and day. I watched the World Cup in HD for the first time this summer, and it was incredible. You could actually pick out player's faces when they were on the far side of the field, without straining to hear the announcer to figure out who had the ball. Don't think the increased fidelity isn't going to translate to gaming.
Moreover, your tradeoff is imaginary. 3D is inherently a vector graphics format --- you don't need to do any more artistic work to take advantage of higher pixel resolution. Zelda TP in HD would've taken the same amount of time to develop, but you'd actually be able to make out stuff in the the expansive vistas, instead of having everything appear indistinct due to lack of resolution.
The lack of HD isn't there to make development cheaper (the more limited polygon pushing power, the lack of shaders, etc do that). The lack of HD is just so Nintendo can get away with just a die shrink of the Dolphin GPU, without having to go back and change the circuit design to get a higher clockspeed or more pipelines. The entertaining thing is that most Wii fans are perfectly happy paying $250 for the Wii, despite the fact that it probably doesn't cost Nintendo any more to make than the $99 GC.*
*) Seriously. If they make the chips on a 90nm process, they'll be able to put both Hollywood and Broadway on a chip 1/4th the size of just the Xenon in the XBox 360. The thing is going to be dirt-cheap to make.
Actually, it's not like that. First, the OpenGL driver has a large hardware-independent portion. Second, if you look at the partial specs that are available for NIVIDA cards, you'll see that the interfaces are quit esimilar between the hardware generations, all the way back to the Riva 128.
Right, because big companies like IBM, Novell, and even Apple don't use GPL'ed code as a basis for their products...
In any case, if Apple had not used said BSD code, what would be different? Apple gives almost nothing back to the BSD community as it is. How would the community have benefited from Apple's use of the code?
I think you're really missing the point. Consider the reasons why Open Darwin failed as a project. They couldn't generate the community interest and involvement necessary to further the project. This in itself is surprising, because OS X has a relatively large userbase, and is different enough from other *NIXs to be interesting.
So why was nobody interested in Open Darwin? Because it's Apple's product. There is no sense of community ownership, or community involvement, working on Open Darwin amounts do doing free R&D for Apple. Moreover, Apple won't even release the really interesting parts of OS X, and can, at any time (as they've demonstrated with the x86 release), withhold code if it is convenient for them to do so.
It's naive to believe that GPL vs BSD has nothing to do with the failure of Open Darwin. If the BSD code had been GPL'ed, Open Darwin could be a true community project. Apple wouldn't be able to withhold code at any time, it would have to release interesting kernel drivers, and they couldn't take peoples' changes and close them back up later. Of course, that is not to say that just GPL by itself would've compensated for the complete lack of tact with which Apple approaches its open source projects, or that this occurrance is necessarily the fate of all BSD licensed projects, but rather that this event is a textbook demonstration of one of the shortcomings of the BSD license.
I think you're being imprecise with the definition of "hardcore gamer" and "casual gamers". Sony's market is from people who buy Madden, Halo, GTA, Final Fantasy, MGS, etc. I wouldn't call these "casual gaming" franchises, not in the sense of what games like the Sims, Nintendogs, Brain Age, Animal Crossing, etc, are trying to target.
The gaming industry is still the domain of enthusiasts. For every game like the Sims or Civ that appeals to a "casual" gamer market, there are a dozen WoWs, Dooms, and UTs. Read some statistics about the gaming industry --- they're quite surprising. "Casual gaming", such as wireless and online gaming, is still much smaller than PC gaming, which is about a quarter of the size of console gaming. Both the latter two markets are dominated by "enthusiast" titles. That leaves the marketshare of "casual" games looking fairly small.
Actually, that's not true. It seems believable, but if you actually look at the statistics, its not the case. Casual gaming is still a very emerging marketplace. It's growing quite fast, and I think Nintendo is banking on that growth, but the gaming industry today is still owned by the GTAs, Halos, Final Fantasys, etc.
Think logically here. Consider the system as a flow, from Sony to the retailer, and from the retailer to the customer. Units shipped/month measures the flow rate to the retailer, units sold/month measured the flow rate to the customer. In practice, the two rates are roughly equal. The retailer might keep some inventory, but that's irrelevent. What's relevant is that the retailer isn't going to keep a continually growing inventory. For the size of the inventory to remain constant, the number of units shipped to the retailer wil be the same as the number of units sold to the customer.
In other words, if Sony is consistently shipping 500k units/month, that means 500k units/month are getting sold through retail channels, on average. If the systems weren't selling, retailers would drop their orders for the next month, and we'd see that in the sales figures. The only place where the size of the inventory matters is in judging total units sold, which is different from the total units shipped by the size of the inventory. Inventories are non-zero, but they're quite small in the case of modern retailers (nowhere near your 500k figure). Warehousing unsold product is very expensive, both for the storage itself and because of the capital tied up in stored merchandise. Therefore, retailers very rarely keep large inventories.
Linux users are perhaps a few percent of the general gamer market, but on the other hand, they make up a substantial percentage of the professional market. If you're using your GPU to do 3D modeling, scientific visualization, etc, there is not insubstantial chance you're on Linux.
The low-end PS3 isn't castrated any more than any other cheaper model in a product line. Is the 17" iMac "castrated"? Is the 30GB iPod "castrated"? Is a Palm E2 "castrated"?
The reason people got on Microsoft's case about the "castrated" 360 was because it was missing a component key to its basic function of gaming (the HDD). The low-end PS3 doesn't have any of those limitations. It's got enough HDD for gaming, wireless controllers, etc. The only things its missing are "extras", like a memory card reader, HDMI port, etc. If you're not interested in the PS3 as a home media device, you don't need any of these things (except maybe the WiFi).
Why not? A high-end GPU off newegg will do over a teraflop too. The difference between those teraflops and the ones in a supercomputer are that the former are not fully-programmable. Only about 128 gigaflops are programmable, and even that is 32-bit precision, compared to the 64-bit precision you find in a supercomputer. The rest is tied up in fixed function circuits that are a lot cheaper than a general purpose floating-point unit.
As for Cell, yeah, 200gigaflops is a lot, but it's also 200gigaflops from very simple in-order cores. Again, a high-end GPU has a couple of dozen quad-issue 128-bit vector cores running at hundreds of megahertz. Is it hard to believe that $400m invested in Cell manged to get 8 dual-issue 128-bit vector cores running at 3.2 GHz?
Yes, citing "2 teraflops" for the PS3 is taking advantage of people who don't know better, but its not something to be disbelieved. Its just a number you have to put into the perspective based on context.
Slashdot Mirror
Science and nonsense should be provided in equal amounts, for the sake of balance?
You know why you never hear about the "secular left"? Because by and large, we mind our own business. You don't have leftists putting up shrines to whomever they worship on state property. Down here in Atlanta, I don't get leftists trying to convert me to secularism a couple of times a year. I don't get idiot 16-year old leftist kids coming to my door telling me about their resurrected prophet (16! what the hell does a 16 year old know about ANYTHING, much less the underlying truth of the universe???) Secular leftists don't cripple the major throughfare near my house every weekend morning driving their giant SUVs to their place of worship. Secular leftists don't send me mail claiming that they know the path I should follow to redemption.
All in all, secular leftists are nice, happy people. We only get our panties in a bunch when the Religious Right decides to butt into everyone's lives.
Yay. America is more progressive than a bunch of backwards Muslims in Turkey.
Sigh...
Are you joking? What training do you have in evolutionary biology to be able to discredit it off-hand? What peer-reviewed research have you done to come to your conclusion? Did you spend the best years of your life in school getting a PHD, before even being allowed to do research in the field? I can just imagine how galling it must be to be a biologist right now. I know if some know-nothing came up to me talking non-sense about fluid dynamics (I'm an aerospace engineer), I'd be more than a bit pissed off.
I'm reminded of a line from A Few Good Men. To paraphrase Jack:
" I have neither the time nor the inclination to explain myself to a man who rises and sleeps under the blanket of the very technology/progress/knowledge that I provide, and then questions the manner in which I provide it! I would rather you just said, "Thank you," and went on your way."
Actually, it's true. Both are somewhere around 2%. The thing Americans often forget is that while Macs are fairly common in the US, they're almost absent in the rest of the world, while Linux has had very significant penetration, even on the desktop, in Latin America, Europe, and Asia.
OK, I think I get your point now, and you are correct in a fairly narrow use case: cases where context switching is a more significant cost, than other factors, and that means essentially server applications that spawn VERY large numbers of threads. Yup, user mode threads a-la OS X have lower performance than say linux kernel threads. On the other hand, they're much less likely to take your machine down if used innapropriately.
OS X uses kernel threads. A POSIX thread maps directly to a Mach thread.
For typical desktop situations, I dispute your claim: no one should be writing real world desktop applications that spawn huge numbers of threads.
Highly multi-threaded programs are the natural result of certain programming paradigms, such as the actor model. If you can implement very cheap threads, why limit the possibilities available to the programmer? In any case, whether or not highly multithreaded programs are useful on the desktop is irrelevent. My claim wasn't that OS X should have a highly capable SMP system, but rather that it was highly misleading to claim that it did, as the original article stated. It's downright dishonest to claim that OS X is at the leading edge when it comes to being able to take advantage of multicore technology, and even more dishonest to claim that this has been the case since "day one".
Overall, the benefit of a threading API, is not merely the implementation, it is also the convenience and protection it offers the programmer. I don't believe that applications with, say, tens of threads are significantly poorer performers on OS X than NT (especially with true SMP PPC systems vs. multicore x86). Indeed there's a reason why NT threads can switch faster, and that is largely due to the crude implementation of virtual memory.
Bullshit. Virtual memory isn't implemented any more crudely in x86/NT than it is in OS X/PPC. The "crude" implementation certainly isn't a performance advantage. Indeed, x86 lacks many of the features that other architectures have for implementing fast context switches. For example, a context switch on x86 results in a TLB flush, because x86 lacks support for tagged TLBs.
The reason OS X's multithreaded performance sucks is because it's just very ancient code that doesn't implement modern mechanisms used in highly concurrent kernels. The kernel locking is very coarse even in Tiger (it was worse in Panther, with just two locks for the whole kernel), and the locking scheme doesn't use any advanced techniques like RCU, lockless algorithms, etc. Darwin still uses the old 4BSD scheduler, which isn't O(1), and threading calls are still heavyweight because a substantial amount of the pthreads implementation results in Mach IPC, which is quite slow.
Again, not that this is really a problem on the desktop, but I never claimed that it was. OS X is an excellent desktop OS with very good performance for desktop apps. It just so happens that the performance of the underlying kernel is largely irrelevent for desktop apps, so you can get away with a pretty shitty kernel (which Darwin is), as long as you have good userspace libraries (fast toolkits, etc).
Apple's strategy is not to do what the competition does.
Not since they started shpping Intel hardware. Now, they have to deliver comparable hardware at a comparable pricepoint. You can bet they wanted to ship Core Solos all across the Mac Mini and Macbook lines, and that's what they would've done in the PowerPC era. But they can't do that anymore, because being on the same architecture makes direct comparison possible, and shipping substantially less than your competitors at a higher price does not result in favorable comparisons.
Apple shipped two Yonahs in desktop systems. No other major player (including Dell) did that.
Only because they wanted to skip the P4 entirely, which is a luxury they had given their late entry into the market.
For example, they already make an iMac with integrated graphics. It was released a couple weeks ago.
That was at a vastly lower pricepoint. A $900 education machine with integrated graphics is not ridiculous. A $1300 general desktop with integrated graphics is.
Merom is the same price as Yonah for any MHz rate, at least in the markets we see (who knows what Apple pays). I can't see how Conroe would be cheaper.
Yonah is a laptop chip. Conroe is a desktop chip. Desktop chips, by virtue of their looser power envelope requirements, are generally significantly cheaper.
First of all the iMac G5 started with the 90+W G5. Second of all, Apple completely revamped the cooling zones inside the iMac when it went Intel.
The 970FX is a 60W chip. So is Conroe. I highly doubt Apple completely revamped the iMac case cooling to make it less capable. They knew Conroe was coming out, that the Core Duo was a stopgap chip, and I'm sure they planned the case design with Conroe in mind. They'd have been stupid not to.
Yes, their all POSIX, but we're talking about the implementation, not the API. OS X's pthreads implementation is nowhere near as scalable as NT's or Linux's. It's completely wrong, thus, to say that OS X is an industry leader in thread support.
Nope. iMac will get Conroe and keep discrete graphics for three reasons:
1) Conroe is cheaper than Merom for a given MHz level;
2) Conroe sits within the 60W thermal envelope of the 970FX in the G5 iMacs;
3) Apple's competitors will have discrete graphics in the iMac's price class.
The iMac is Apple's top desktop Mac. The PowerMac is a workstation line, and will be further differentiated as such by moving to Woodcrest. The Mac Mini will take the lower end of the market. In order to be competitive with other vendors pushing Intel machines in the $1300-$1700 category, Conroe and discrete graphics are a must.
It's interesting to note how Apple users have continually low-balled Apple's configurations during the Intel transition. Everyone thought that the Mac Minis would be Core Solo-only, and that only the highest end Macbook would have a Core Duo, if a Duo was offered in the line at all. I think its a holdover from G4/G5 era "differentiation" thinking. The era of arbitrary differentiation is over at Apple. If Dell is shipping Meroms in its $1200 laptops, so will Apple. If Dell is offering discrete graphics and Conroe processors in its $1500 desktops, so will Apple.
This is precisely the strategy Apple has followed thus far with the Intel transition. The Apple products are at the high end of their price classes, with the price buoyed up with features that are usually extra-cost options for other machines, but they feature competitive configurations at each of those price-points. They know very well that they can sell an iMac that offers more features than a Dell that's $200 cheaper, but not an iMac that costs the same and offers a much lesser processor or GPU.
I'm claiming that this paragraph is bullshit.
Even when lots of different applications, many of them efficiently multi-threaded, are run on 10.4.5 or 10.4.6 only the first two CPUs are used efficiently while the third and fourth are getting plenty of work....but aren't quite living up to their full potential. Each added core after four seems to drop off in efficiency....not because OS X doesn't handle lots of processors properly, it does. In fact it's an industry leader in terms of being ready for the next generation of multicore, multiprocessor technology. It has been since day one and Apple has consistently kept it at the leading edge since then.
OS X doesn't handle lots of processors properly, and in no way is it anything resembling an industry leader in being ready for multicore technology. That's not the case in Tiger, and it certainly wasn't the case from "day one". OS X's SMP support was deficient pre-Tiger, and its merely "barely adequate" in Tiger. Of the major kernels in use, OS X's is near the back in regards to SMP and thread support, behind NT, Solaris, AIX, HP-UX, Linux, BeOS, and FreeBSD. The only thing its ahead of are kernels like NetBSD and OpenBSD, which still use a single kernel lock.
Apple can improve this situation in Tiger, but there's no secret technology Apple is holding close to their chest that'd give them the lead, as the article claims. Threading and SMP will get better in Tiger, maybe even up to Linux 2.4 levels, but its not being to offer world-beater performance.
Bullshit, on multiple levels. First, it wasn't even until 10.4 that OS X's SMP was good enough for machines with > 2 CPUs, and even in 10.4 OS X's SMP support is far behind that of Linux, FreeBSD, or Solaris.
After all, everything is inferior to science, isn't it?
Yep, pretty much. Everything else is just a way to unwind so we can be fresh to do more science the next day...
The Sims is the best selling game of all time... on the PC. However, the PC gaming market is only 1/4th as large as the console gaming market. Moreover, while the Sims may have outsold specific games in other genres, its sales are still small compared to the overall sales of games in other genres. That is to say its very hard to extrapolate the size of the casual gaming market just from sales of one game. The Sims may have sold 16 million copies, but hundreds of millions of other games are being sold in all the non-casual genres.
It is very possible that sales of "The Sims" might hint at a large untapped market that exists. However, until that market starts buying games at the same rate as the "hardcore" market, it's silly to say that its a larger (ie: potentially more profitable) one.
Actually, check the stats sometime. The casual gaming market (as represented by the sims, etc) is quite a bit smaller than the "hardcore" gaming market. As for non-gamers --- well, they don't play games. It's possible that Nintendo might get them to play games, but that's a big "if".
Says the Nintendo fan? Who in the last two generations (N64 and GC) had a console that could do pretty games, but with precious little good content to play on them?
2% my ass. The difference between HD and SD is night and day. I watched the World Cup in HD for the first time this summer, and it was incredible. You could actually pick out player's faces when they were on the far side of the field, without straining to hear the announcer to figure out who had the ball. Don't think the increased fidelity isn't going to translate to gaming.
Moreover, your tradeoff is imaginary. 3D is inherently a vector graphics format --- you don't need to do any more artistic work to take advantage of higher pixel resolution. Zelda TP in HD would've taken the same amount of time to develop, but you'd actually be able to make out stuff in the the expansive vistas, instead of having everything appear indistinct due to lack of resolution.
The lack of HD isn't there to make development cheaper (the more limited polygon pushing power, the lack of shaders, etc do that). The lack of HD is just so Nintendo can get away with just a die shrink of the Dolphin GPU, without having to go back and change the circuit design to get a higher clockspeed or more pipelines. The entertaining thing is that most Wii fans are perfectly happy paying $250 for the Wii, despite the fact that it probably doesn't cost Nintendo any more to make than the $99 GC.*
*) Seriously. If they make the chips on a 90nm process, they'll be able to put both Hollywood and Broadway on a chip 1/4th the size of just the Xenon in the XBox 360. The thing is going to be dirt-cheap to make.
Actually, it's not like that. First, the OpenGL driver has a large hardware-independent portion. Second, if you look at the partial specs that are available for NIVIDA cards, you'll see that the interfaces are quit esimilar between the hardware generations, all the way back to the Riva 128.
Right, because big companies like IBM, Novell, and even Apple don't use GPL'ed code as a basis for their products...
In any case, if Apple had not used said BSD code, what would be different? Apple gives almost nothing back to the BSD community as it is. How would the community have benefited from Apple's use of the code?
I think you're really missing the point. Consider the reasons why Open Darwin failed as a project. They couldn't generate the community interest and involvement necessary to further the project. This in itself is surprising, because OS X has a relatively large userbase, and is different enough from other *NIXs to be interesting.
So why was nobody interested in Open Darwin? Because it's Apple's product. There is no sense of community ownership, or community involvement, working on Open Darwin amounts do doing free R&D for Apple. Moreover, Apple won't even release the really interesting parts of OS X, and can, at any time (as they've demonstrated with the x86 release), withhold code if it is convenient for them to do so.
It's naive to believe that GPL vs BSD has nothing to do with the failure of Open Darwin. If the BSD code had been GPL'ed, Open Darwin could be a true community project. Apple wouldn't be able to withhold code at any time, it would have to release interesting kernel drivers, and they couldn't take peoples' changes and close them back up later. Of course, that is not to say that just GPL by itself would've compensated for the complete lack of tact with which Apple approaches its open source projects, or that this occurrance is necessarily the fate of all BSD licensed projects, but rather that this event is a textbook demonstration of one of the shortcomings of the BSD license.
I think you're being imprecise with the definition of "hardcore gamer" and "casual gamers". Sony's market is from people who buy Madden, Halo, GTA, Final Fantasy, MGS, etc. I wouldn't call these "casual gaming" franchises, not in the sense of what games like the Sims, Nintendogs, Brain Age, Animal Crossing, etc, are trying to target.
The gaming industry is still the domain of enthusiasts. For every game like the Sims or Civ that appeals to a "casual" gamer market, there are a dozen WoWs, Dooms, and UTs. Read some statistics about the gaming industry --- they're quite surprising. "Casual gaming", such as wireless and online gaming, is still much smaller than PC gaming, which is about a quarter of the size of console gaming. Both the latter two markets are dominated by "enthusiast" titles. That leaves the marketshare of "casual" games looking fairly small.
Actually, that's not true. It seems believable, but if you actually look at the statistics, its not the case. Casual gaming is still a very emerging marketplace. It's growing quite fast, and I think Nintendo is banking on that growth, but the gaming industry today is still owned by the GTAs, Halos, Final Fantasys, etc.
Think logically here. Consider the system as a flow, from Sony to the retailer, and from the retailer to the customer. Units shipped/month measures the flow rate to the retailer, units sold/month measured the flow rate to the customer. In practice, the two rates are roughly equal. The retailer might keep some inventory, but that's irrelevent. What's relevant is that the retailer isn't going to keep a continually growing inventory. For the size of the inventory to remain constant, the number of units shipped to the retailer wil be the same as the number of units sold to the customer.
In other words, if Sony is consistently shipping 500k units/month, that means 500k units/month are getting sold through retail channels, on average. If the systems weren't selling, retailers would drop their orders for the next month, and we'd see that in the sales figures. The only place where the size of the inventory matters is in judging total units sold, which is different from the total units shipped by the size of the inventory. Inventories are non-zero, but they're quite small in the case of modern retailers (nowhere near your 500k figure). Warehousing unsold product is very expensive, both for the storage itself and because of the capital tied up in stored merchandise. Therefore, retailers very rarely keep large inventories.
Linux users are perhaps a few percent of the general gamer market, but on the other hand, they make up a substantial percentage of the professional market. If you're using your GPU to do 3D modeling, scientific visualization, etc, there is not insubstantial chance you're on Linux.
The low-end PS3 isn't castrated any more than any other cheaper model in a product line. Is the 17" iMac "castrated"? Is the 30GB iPod "castrated"? Is a Palm E2 "castrated"?
The reason people got on Microsoft's case about the "castrated" 360 was because it was missing a component key to its basic function of gaming (the HDD). The low-end PS3 doesn't have any of those limitations. It's got enough HDD for gaming, wireless controllers, etc. The only things its missing are "extras", like a memory card reader, HDMI port, etc. If you're not interested in the PS3 as a home media device, you don't need any of these things (except maybe the WiFi).
Why not? A high-end GPU off newegg will do over a teraflop too. The difference between those teraflops and the ones in a supercomputer are that the former are not fully-programmable. Only about 128 gigaflops are programmable, and even that is 32-bit precision, compared to the 64-bit precision you find in a supercomputer. The rest is tied up in fixed function circuits that are a lot cheaper than a general purpose floating-point unit.
As for Cell, yeah, 200gigaflops is a lot, but it's also 200gigaflops from very simple in-order cores. Again, a high-end GPU has a couple of dozen quad-issue 128-bit vector cores running at hundreds of megahertz. Is it hard to believe that $400m invested in Cell manged to get 8 dual-issue 128-bit vector cores running at 3.2 GHz?
Yes, citing "2 teraflops" for the PS3 is taking advantage of people who don't know better, but its not something to be disbelieved. Its just a number you have to put into the perspective based on context.