The business market is very difficult.....etc etc etc .....
All they have to do to entice gamers is to add a few more video cards.
I'd say having all the top games on release day is what enticers gamers. Apple doesn't, and hasn't done anything to to bring in game developers to the Mac. The switch to Intel has made it somewhat easier, but all those direct X games have to re-written specifically for OGL, and that's just piling more work on developers. Mac is a small market for games. I wish Apple would start a gaming strategy because that would bring in more users, but there is no indication that Apple will ever break out of the constraints they put on themselves, by saying we are interested in "this" and "that", but "these" sales, developers, and users are not for us.
So back on topic...which of the various OS's (Linux, Windows, OSX, Solaris) is best for a 8 core Mac Pro? Just from the perspective of SMP support...
Right now it would be Linux (depending on which version), and Solaris.
Solaris is designed for that, and the commercial versions of Linux are as well. So is IBM's AIX, and their other OS's.
Windows, so far, becomes less efficient above 4 sockets, or cores, and while X does pretty well at 4, we don't really know much about it with 8. The one test in Anands showed no problems with it running with 8 cores, but he didn't have time to do useful tests on performance, other than to say it seemed to do well with multitasking.
Hopefully, Leopard will have some optimizations in Mach that will eliminate some of the reported problems in threading.
I'd say having all the top games on release day is what enticers gamers. Apple doesn't, and hasn't done anything to to bring in game developers to the Mac. The switch to Intel has made it somewhat easier, but all those direct X games have to re-written specifically for OGL, and that's just piling more work on developers. Mac is a small market for games. I wish Apple would start a gaming strategy because that would bring in more users, but there is no indication that Apple will ever break out of the constraints they put on themselves, by saying we are interested in "this" and "that", but "these" sales, developers, and users are not for us.
That's true. Apple seems to run hot and cold over games. Remember in the '90's when they made such a big deal about "Game Sprockets"? Then they did little to move further with it.
Right now, I see gamers as being excited about BootCamp. They don't seem to mind re-booting for games, as long as the performance is there.**graphics cards**
Unlike many other computer users, gamers are willing to put up with more bother for their hobby.
It really isn't such a big bother to re-boot. If you're turning your machine from a computer (OS X), to a game machine (XP or Vista), the two minutes isn't really such a big deal.
And from the word that is out there right now, Vista runs current games 15% to 20% slower than XP does. That will change over time, but it could be an incentive for gamers over the next year or so to move over and use XP for games, rather than to get a new Vista machine.
I don't know about that. Isn't Leopard supposed to be a multi-core-friendly OS?
It is. But that doesn't mean it scales infinitely well.
The requirements change slightly as the number of cores go up. What may be optimised for two cores may be very inefficient on four. You can work it out from there.
The more cores, the more threads needed. If the OS can't spread out with a lot of cores then the speed will hardly increase as the core count goes up. That's way oversimplified, but it's a problem that the OS teams have to deal with.
It is. But that doesn't mean it scales infinitely well.
The requirements change slightly as the number of cores go up. What may be optimised for two cores may be very inefficient on four. You can work it out from there.
The more cores, the more threads needed. If the OS can't spread out with a lot of cores then the speed will hardly increase as the core count goes up. That's way oversimplified, but it's a problem that the OS teams have to deal with.
So as I understand it, the speed of chips wont have to go up as long as the number of cores breaks down the number calculations the OS has to do..er I mean each individual core. Like a huge graphic or 3d render job spread over 8 cores that run at say 2ghz is much more efficient as doing all the computational work on one or 2 cores runing at say 3-4 ghz?
So as I understand it, the speed of chips wont have to go up as long as the number of cores breaks down the number calculations the OS has to do..er I mean each individual core. Like a huge graphic or 3d render job spread over 8 cores that run at say 2ghz is much more efficient as doing all the computational work on one or 2 cores runing at say 3-4 ghz?
This is what they're hoping. But it doesn't really work that way.
Only some programs can work with more than two cores and gain more speed.
Look at PS. Two cores give almost twice the speed, but going to four does nothing. it must be rewrittwn for more cores.
An example: An OS on one core might do a task in, say for ease of arithmetic, 10 seconds. with two cores, it might take 6 seconds. With 4, it might take 4 seconds, with 8, it might take 2.5 seconds. You don't get linear scaling.
The problem is threefold.
The chips don't scale linearly either. They get more inefficient as the number goes up. Then the OS loses efficiency as well.
Then, finally, the program has to scale well.
There will be many cases where a program works 25% faster on an 8 core machine then it does on a one core machine.
For most users, the gain will be in multitasking. Using several programs at once.
That was supposedly the advantage of the OS with one cpu as well. But, it never lived up to its hype.
OS X works much better with two cores, even for OS tasks alone.
I guess the quesiton isn't just support but efficiency at the kernel level for SMP. Is xnu any better than say Linux? Is Solaris king? How well locking and synchconization is implemented will impact the OS's performance in SMP environments. How well IOKit handles synchonization determines how well drivers work in a SMP environment.
I know just enough to ask the question but not enough to figure out the right answer. Perhaps I need to start reading Ars again.
I guess the quesiton isn't just support but efficiency at the kernel level for SMP. Is xnu any better than say Linux? Is Solaris king? How well locking and synchconization is implemented will impact the OS's performance in SMP environments. How well IOKit handles synchonization determines how well drivers work in a SMP environment.
I know just enough to ask the question but not enough to figure out the right answer. Perhaps I need to start reading Ars again.
This is what they're hoping. But it doesn't really work that way.
Of course you're generalizing somewhat. What problem the software is trying to solve and how its written play a huge part in the efficiency numbers. I've seen (and written) cases where a linear speed up was achieved on an increasing number of cores. Its even better to get better than linear in some cases. What it comes down to is where the bottleneck is, and how the software partitions the work. Unfortunately most software quickly becomes bandwidth and/or memory latency bound, and then adding cores typically doesn't improve the situation much. Software the has to communicate between the cores runs into other bottlenecks. And so on. As time goes on and developers get better at this (and the tools improve) then the performance scalability should generally be better.
Keep in mind also that nothing scales linearly... clock rate, cache sizes, number of cores, etc. all suffer from diminishing returns. Even mindlessly increasing bandwidth or decreasing latency involves tradeoffs that mean its not worth pushing them to extremes.
Of course you're generalizing somewhat. What problem the software is trying to solve and how its written play a huge part in the efficiency numbers. I've seen (and written) cases where a linear speed up was achieved on an increasing number of cores. Its even better to get better than linear in some cases. What it comes down to is where the bottleneck is, and how the software partitions the work. Unfortunately most software quickly becomes bandwidth and/or memory latency bound, and then adding cores typically doesn't improve the situation much. Software the has to communicate between the cores runs into other bottlenecks. And so on. As time goes on and developers get better at this (and the tools improve) then the performance scalability should generally be better.
Keep in mind also that nothing scales linearly... clock rate, cache sizes, number of cores, etc. all suffer from diminishing returns. Even mindlessly increasing bandwidth or decreasing latency involves tradeoffs that mean its not worth pushing them to extremes.
I was going to mention that a very few specialized programs can scale better than linearly, but not by much, and it is rare enough so that is doesn't affect most people.
If we take a look at the top 500 supercomputer list and specs, we can see just how poorly some cpu's scale.
Well I don't know if this has been said, but couldn't Apple put SLi or Crossfire in their Mac Pro, which would obviously increase the graphics power. It seems like they could easily do that, am I right?
Well I don't know if this has been said, but couldn't Apple put SLi or Crossfire in their Mac Pro, which would obviously increase the graphics power. It seems like they could easily do that, am I right?
I'd say having all the top games on release day is what enticers gamers. Apple doesn't, and hasn't done anything to to bring in game developers to the Mac. The switch to Intel has made it somewhat easier, but all those direct X games have to re-written specifically for OGL, and that's just piling more work on developers. Mac is a small market for games. I wish Apple would start a gaming strategy because that would bring in more users, but there is no indication that Apple will ever break out of the constraints they put on themselves, by saying we are interested in "this" and "that", but "these" sales, developers, and users are not for us.
That's true. Apple seems to run hot and cold over games. Remember in the '90's when they made such a big deal about "Game Sprockets"? Then they did little to move further with it.
Right now, I see gamers as being excited about BootCamp. They don't seem to mind re-booting for games, as long as the performance is there.**graphics cards**
Unlike many other computer users, gamers are willing to put up with more bother for their hobby.
It really isn't such a big bother to re-boot. If you're turning your machine from a computer (OS X), to a game machine (XP or Vista), the two minutes isn't really such a big deal.
And from the word that is out there right now, Vista runs current games 15% to 20% slower than XP does. That will change over time, but it could be an incentive for gamers over the next year or so to move over and use XP for games, rather than to get a new Vista machine.
Nods. Apple does seem to run hot and cold. And they do seem to have some unnecessary self imposed barriers to gaming.
3 gaming towers under the Quad range ie some Conroe towers would be a great range between the mini Mac and the Workstation. They'd sell more towers that way. And get a nice cut. Sure they'd lose 'some' cannibal iMac sales. But those who want a gaming rig probably aren't going to buy an iMac anyway?
Apple seem to make a fuss about games on its gaming pages...but it's not at the fore of Apple's business. But they don't mind 'boasting' about Doom and Quake framerates on the Mac Pro page. And they DO support Open GL...the open(?) standard. So...there's something(!) there. Isn't the technologies to port Direct X to GL/Macs out there now? There have been a couple of such technologies boasting Press Releases in the last few years?
Apple aren't far away now from selling 2 million Macs per quarter. That's 8 million a year. After four years? That's 32 million Macs on top of installed base of 30 million?
Don't tell me people can walk away from 64 million of anything? That's comparable to any amount of sales the PS3, 360 or Wii may get over the next four years?
Apple aren't far away now from selling 2 million Macs per quarter. That's 8 million a year. After four years? That's 32 million Macs on top of installed base of 30 million?
Don't tell me people can walk away from 64 million of anything? That's comparable to any amount of sales the PS3, 360 or Wii may get over the next four years?
What you are missing is that most of the people buying Macs aren't buying any (or at least not many) games for them, whereas the gaming console owners are buying on average 10 games per year (and that's conservative). So if you compare the game sales for the 360 to Mac, and assume both will be about 64 million machines in 4 years, then you'll see ~50 times the game sales on the console. And developing for a console is easier/cheaper than for the Mac (there is exactly 1 version of the console hardware, as opposed to many flavors of Mac).
This doesn't mean there won't be any games for the Mac, just that it isn't going to be notable as a significant gaming platform. You may see Apple put a bit of increased emphasis on it in the near future in order to add games to iTV much like the iPod now has games. They'll typically be the big hits and a lot of casual games.
What you are missing is that most of the people buying Macs aren't buying any (or at least not many) games for them, whereas the gaming console owners are buying on average 10 games per year (and that's conservative). So if you compare the game sales for the 360 to Mac, and assume both will be about 64 million machines in 4 years, then you'll see ~50 times the game sales on the console. And developing for a console is easier/cheaper than for the Mac (there is exactly 1 version of the console hardware, as opposed to many flavors of Mac).
This doesn't mean there won't be any games for the Mac, just that it isn't going to be notable as a significant gaming platform. You may see Apple put a bit of increased emphasis on it in the near future in order to add games to iTV much like the iPod now has games. They'll typically be the big hits and a lot of casual games.
I agree with the idea of problems with Mac game development.
But, I'm'talking about the interes in the Mac platform from PC gamers.
We wouldn't need much game development for the Mac, if they could play PC games well.
Crossover allows some games to play very well on Linux. That's another area to look to.
When Crossover Mac is officially out, we will see how many games it plays, and how well.
Comments
That's what I'm saying.
The business market is very difficult.....etc etc etc .....
All they have to do to entice gamers is to add a few more video cards.
I'd say having all the top games on release day is what enticers gamers. Apple doesn't, and hasn't done anything to to bring in game developers to the Mac. The switch to Intel has made it somewhat easier, but all those direct X games have to re-written specifically for OGL, and that's just piling more work on developers. Mac is a small market for games. I wish Apple would start a gaming strategy because that would bring in more users, but there is no indication that Apple will ever break out of the constraints they put on themselves, by saying we are interested in "this" and "that", but "these" sales, developers, and users are not for us.
So back on topic...which of the various OS's (Linux, Windows, OSX, Solaris) is best for a 8 core Mac Pro? Just from the perspective of SMP support...
Right now it would be Linux (depending on which version), and Solaris.
Solaris is designed for that, and the commercial versions of Linux are as well. So is IBM's AIX, and their other OS's.
Windows, so far, becomes less efficient above 4 sockets, or cores, and while X does pretty well at 4, we don't really know much about it with 8. The one test in Anands showed no problems with it running with 8 cores, but he didn't have time to do useful tests on performance, other than to say it seemed to do well with multitasking.
Hopefully, Leopard will have some optimizations in Mach that will eliminate some of the reported problems in threading.
I'd say having all the top games on release day is what enticers gamers. Apple doesn't, and hasn't done anything to to bring in game developers to the Mac. The switch to Intel has made it somewhat easier, but all those direct X games have to re-written specifically for OGL, and that's just piling more work on developers. Mac is a small market for games. I wish Apple would start a gaming strategy because that would bring in more users, but there is no indication that Apple will ever break out of the constraints they put on themselves, by saying we are interested in "this" and "that", but "these" sales, developers, and users are not for us.
That's true. Apple seems to run hot and cold over games. Remember in the '90's when they made such a big deal about "Game Sprockets"? Then they did little to move further with it.
Right now, I see gamers as being excited about BootCamp. They don't seem to mind re-booting for games, as long as the performance is there.**graphics cards**
Unlike many other computer users, gamers are willing to put up with more bother for their hobby.
It really isn't such a big bother to re-boot. If you're turning your machine from a computer (OS X), to a game machine (XP or Vista), the two minutes isn't really such a big deal.
And from the word that is out there right now, Vista runs current games 15% to 20% slower than XP does. That will change over time, but it could be an incentive for gamers over the next year or so to move over and use XP for games, rather than to get a new Vista machine.
I don't know about that. Isn't Leopard supposed to be a multi-core-friendly OS?
It is. But that doesn't mean it scales infinitely well.
The requirements change slightly as the number of cores go up. What may be optimised for two cores may be very inefficient on four. You can work it out from there.
The more cores, the more threads needed. If the OS can't spread out with a lot of cores then the speed will hardly increase as the core count goes up. That's way oversimplified, but it's a problem that the OS teams have to deal with.
It is. But that doesn't mean it scales infinitely well.
The requirements change slightly as the number of cores go up. What may be optimised for two cores may be very inefficient on four. You can work it out from there.
The more cores, the more threads needed. If the OS can't spread out with a lot of cores then the speed will hardly increase as the core count goes up. That's way oversimplified, but it's a problem that the OS teams have to deal with.
So as I understand it, the speed of chips wont have to go up as long as the number of cores breaks down the number calculations the OS has to do..er I mean each individual core. Like a huge graphic or 3d render job spread over 8 cores that run at say 2ghz is much more efficient as doing all the computational work on one or 2 cores runing at say 3-4 ghz?
So as I understand it, the speed of chips wont have to go up as long as the number of cores breaks down the number calculations the OS has to do..er I mean each individual core. Like a huge graphic or 3d render job spread over 8 cores that run at say 2ghz is much more efficient as doing all the computational work on one or 2 cores runing at say 3-4 ghz?
This is what they're hoping. But it doesn't really work that way.
Only some programs can work with more than two cores and gain more speed.
Look at PS. Two cores give almost twice the speed, but going to four does nothing. it must be rewrittwn for more cores.
An example: An OS on one core might do a task in, say for ease of arithmetic, 10 seconds. with two cores, it might take 6 seconds. With 4, it might take 4 seconds, with 8, it might take 2.5 seconds. You don't get linear scaling.
The problem is threefold.
The chips don't scale linearly either. They get more inefficient as the number goes up. Then the OS loses efficiency as well.
Then, finally, the program has to scale well.
There will be many cases where a program works 25% faster on an 8 core machine then it does on a one core machine.
For most users, the gain will be in multitasking. Using several programs at once.
That was supposedly the advantage of the OS with one cpu as well. But, it never lived up to its hype.
OS X works much better with two cores, even for OS tasks alone.
I know just enough to ask the question but not enough to figure out the right answer. Perhaps I need to start reading Ars again.
Vinea
I guess the quesiton isn't just support but efficiency at the kernel level for SMP. Is xnu any better than say Linux? Is Solaris king? How well locking and synchconization is implemented will impact the OS's performance in SMP environments. How well IOKit handles synchonization determines how well drivers work in a SMP environment.
I know just enough to ask the question but not enough to figure out the right answer. Perhaps I need to start reading Ars again.
Vinea
Just don't post. If you think I'm bad...
This is what they're hoping. But it doesn't really work that way.
Of course you're generalizing somewhat. What problem the software is trying to solve and how its written play a huge part in the efficiency numbers. I've seen (and written) cases where a linear speed up was achieved on an increasing number of cores. Its even better to get better than linear in some cases. What it comes down to is where the bottleneck is, and how the software partitions the work. Unfortunately most software quickly becomes bandwidth and/or memory latency bound, and then adding cores typically doesn't improve the situation much. Software the has to communicate between the cores runs into other bottlenecks. And so on. As time goes on and developers get better at this (and the tools improve) then the performance scalability should generally be better.
Keep in mind also that nothing scales linearly... clock rate, cache sizes, number of cores, etc. all suffer from diminishing returns. Even mindlessly increasing bandwidth or decreasing latency involves tradeoffs that mean its not worth pushing them to extremes.
Of course you're generalizing somewhat. What problem the software is trying to solve and how its written play a huge part in the efficiency numbers. I've seen (and written) cases where a linear speed up was achieved on an increasing number of cores. Its even better to get better than linear in some cases. What it comes down to is where the bottleneck is, and how the software partitions the work. Unfortunately most software quickly becomes bandwidth and/or memory latency bound, and then adding cores typically doesn't improve the situation much. Software the has to communicate between the cores runs into other bottlenecks. And so on. As time goes on and developers get better at this (and the tools improve) then the performance scalability should generally be better.
Keep in mind also that nothing scales linearly... clock rate, cache sizes, number of cores, etc. all suffer from diminishing returns. Even mindlessly increasing bandwidth or decreasing latency involves tradeoffs that mean its not worth pushing them to extremes.
I was going to mention that a very few specialized programs can scale better than linearly, but not by much, and it is rare enough so that is doesn't affect most people.
If we take a look at the top 500 supercomputer list and specs, we can see just how poorly some cpu's scale.
Well I don't know if this has been said, but couldn't Apple put SLi or Crossfire in their Mac Pro, which would obviously increase the graphics power. It seems like they could easily do that, am I right?
Arrrggg!
But, welcome anyway.
If they do have a 8 core macpro, it should have 32gb of memory. That would cost $20000 more.
Ummmmmm yeah. Not for quite a while my friend...
I'd say having all the top games on release day is what enticers gamers. Apple doesn't, and hasn't done anything to to bring in game developers to the Mac. The switch to Intel has made it somewhat easier, but all those direct X games have to re-written specifically for OGL, and that's just piling more work on developers. Mac is a small market for games. I wish Apple would start a gaming strategy because that would bring in more users, but there is no indication that Apple will ever break out of the constraints they put on themselves, by saying we are interested in "this" and "that", but "these" sales, developers, and users are not for us.
That's true. Apple seems to run hot and cold over games. Remember in the '90's when they made such a big deal about "Game Sprockets"? Then they did little to move further with it.
Right now, I see gamers as being excited about BootCamp. They don't seem to mind re-booting for games, as long as the performance is there.**graphics cards**
Unlike many other computer users, gamers are willing to put up with more bother for their hobby.
It really isn't such a big bother to re-boot. If you're turning your machine from a computer (OS X), to a game machine (XP or Vista), the two minutes isn't really such a big deal.
And from the word that is out there right now, Vista runs current games 15% to 20% slower than XP does. That will change over time, but it could be an incentive for gamers over the next year or so to move over and use XP for games, rather than to get a new Vista machine.
Nods. Apple does seem to run hot and cold. And they do seem to have some unnecessary self imposed barriers to gaming.
3 gaming towers under the Quad range ie some Conroe towers would be a great range between the mini Mac and the Workstation. They'd sell more towers that way. And get a nice cut. Sure they'd lose 'some' cannibal iMac sales. But those who want a gaming rig probably aren't going to buy an iMac anyway?
Apple seem to make a fuss about games on its gaming pages...but it's not at the fore of Apple's business. But they don't mind 'boasting' about Doom and Quake framerates on the Mac Pro page. And they DO support Open GL...the open(?) standard. So...there's something(!) there. Isn't the technologies to port Direct X to GL/Macs out there now? There have been a couple of such technologies boasting Press Releases in the last few years?
Apple aren't far away now from selling 2 million Macs per quarter. That's 8 million a year. After four years? That's 32 million Macs on top of installed base of 30 million?
Don't tell me people can walk away from 64 million of anything? That's comparable to any amount of sales the PS3, 360 or Wii may get over the next four years?
Lemon Bon Bon
Hmm. Gaming iMac is starting to look a reality...
But that doesn't rule out a Conroe tower with an Ati 1900...
Lemon Bon Bon
Apple aren't far away now from selling 2 million Macs per quarter. That's 8 million a year. After four years? That's 32 million Macs on top of installed base of 30 million?
Don't tell me people can walk away from 64 million of anything? That's comparable to any amount of sales the PS3, 360 or Wii may get over the next four years?
What you are missing is that most of the people buying Macs aren't buying any (or at least not many) games for them, whereas the gaming console owners are buying on average 10 games per year (and that's conservative). So if you compare the game sales for the 360 to Mac, and assume both will be about 64 million machines in 4 years, then you'll see ~50 times the game sales on the console. And developing for a console is easier/cheaper than for the Mac (there is exactly 1 version of the console hardware, as opposed to many flavors of Mac).
This doesn't mean there won't be any games for the Mac, just that it isn't going to be notable as a significant gaming platform. You may see Apple put a bit of increased emphasis on it in the near future in order to add games to iTV much like the iPod now has games. They'll typically be the big hits and a lot of casual games.
What you are missing is that most of the people buying Macs aren't buying any (or at least not many) games for them, whereas the gaming console owners are buying on average 10 games per year (and that's conservative). So if you compare the game sales for the 360 to Mac, and assume both will be about 64 million machines in 4 years, then you'll see ~50 times the game sales on the console. And developing for a console is easier/cheaper than for the Mac (there is exactly 1 version of the console hardware, as opposed to many flavors of Mac).
This doesn't mean there won't be any games for the Mac, just that it isn't going to be notable as a significant gaming platform. You may see Apple put a bit of increased emphasis on it in the near future in order to add games to iTV much like the iPod now has games. They'll typically be the big hits and a lot of casual games.
I agree with the idea of problems with Mac game development.
But, I'm'talking about the interes in the Mac platform from PC gamers.
We wouldn't need much game development for the Mac, if they could play PC games well.
Crossover allows some games to play very well on Linux. That's another area to look to.
When Crossover Mac is officially out, we will see how many games it plays, and how well.