<strong>What are Vector Units?</strong><hr></blockquote>
A Vector processing unit (Altivec/Velocity Engine, MMX) is a processing unit capable of perform a Single Instruction on Multiple Data, SIMD. Altivec is a form of this, and it is capable of pushing through 128 bits of data on every tick of the CPU clock. Sony used vector processing to make the two chips used in the Playstation 2. Vector processing works very well in the case of graphics since ussually you are performing the same operation on every pixel in the screen.
it seems obvious that they are being very opptimistic but they aren't lying. While they may have been overly opptimistic about the G3 it does in fact exist. This article says that some type of community processing that could be used for a gaming consol does exist. It isn't talking about a multiprocessor, or about batch processing over a local network, or about improved mutliplayer games.
I want to know if anyone has any idea about how they can do this. Not if they can do it because it seems they are, regardless of when it comes out. It seems like the idea behind SETI only in real-time. I want to know how. There are some obvious problems with bandwidth. But wouldn't it be cool if Apple brought Marxism to the Mac.
<strong>It has been reported that Playstation 3 due out in 2004 will have some form of "distributed computing"
If IBM and Sony are doing R&D could Apple be doing the same?</strong><hr></blockquote>
Because Apple (NeXt) based OS X on Mach it is entirely possible to support distributed computing at the kernel level. Because the Mach kernel uses a messaging system to compunicate between the micro-kernel and the OS Layer it is possible to have multiple distributed micro-kernels on many different machines. This would make the OS appear to programmers as normal, running on one computer, but actually underneath it could actually be running on a few hundred.
This was only in development at CMU, and never really saw full fruition before the Mach research project there was canceled. Apple could pick up on this research again, and perhaps with XServe they will have a reason. But the framework is there.
<strong>it seems obvious that they are being very opptimistic but they aren't lying. While they may have been overly opptimistic about the G3 it does in fact exist. This article says that some type of community processing that could be used for a gaming consol does exist. It isn't talking about a multiprocessor, or about batch processing over a local network, or about improved mutliplayer games.</strong><hr></blockquote>
They might not be lying outright so much as they are erring on the side of rampant optimism. They have plenty of time to backpedal, and there will be no shortage of perfectly legitimate reasons for them to do so.
"Whoops! We were pretty sure broadband would be more widespread by 2005!"
[quote]<strong>I want to know if anyone has any idea about how they can do this. Not if they can do it because it seems they are, regardless of when it comes out. It seems like the idea behind SETI only in real-time. I want to know how. There are some obvious problems with bandwidth. But wouldn't it be cool if Apple brought Marxism to the Mac.</strong><hr></blockquote>
From a technological standpoint, implementing it isn't that difficult. It's been a solved problem for years, in fact. The difficulty lies in determining what benefit distributed computing would offer to gaming consoles. There is also a possible sticking point involving broadband providers: Almost all consumer broadband services throttle upload bandwidth and prohibit servers.
Consider your statement, "like SETI only in real time." There is a vast gulf there. Since SETI is not real time, the computer it's run on only has to be online sporadically, and SETI can upload in the background because there's no hurry. SETI can run on spare cycles and behind screensavers because there's no hurry. The computer can be put to sleep or powered down or even taken apart and reassembled and the central SETI servers will never know or care, because there's no hurry.
Now, consider a game console. Most multiplayer games are real time. All massively multiplayer games are real time, all the time. The immediate implications are that if your console joins this network: you can't shut it off or take it offline; it becomes a server in the strictest sense of the word, violating the terms of service of every cable and ADSL provider I'm aware of; and it has to provide real-time data to both you and some arbitrary number of other people simultaneously. If your PS/3 is responsible for some part of a castle, and your knight is ambushed in a forest just as a party of ten decides to enter the castle, your console is going to get hammered, and something will have to give. Bottom line: It's not a good idea to have the same machine running something as demanding as a game and something as demanding as a game server. I can think of a few ways around this problem that use more than one machine, but there's still the problem that your upload speed is throttled to 2 or 3 56K modems' worth, and you'll get disconnected if your ISP sees to much traffic coming upstream. And, of course, there's always the much greater possibility that one of the console-servers loses power, and - whoops! - there goes the castle, and the PCs in it.
I can see no way that distributed consoles would yield a thousandfold increase in anything, unless the game developers were using them to compile their applications. Programmer?
None of this has anything to do with the viability of an Apple-standard distributed computing API for Macs. I suppose that Macs would use it in the same way that SETI does: To farm out compute-bound work which can be broken up into chunks, and which does not have to be completed in real time (3D rendering, PostScript RIPping, compositing, scientific simulations) to other machines. In fact, somewhere on these boards there is a photo of an early version of Shake(?) for OS X doing exactly this.
I don't buy into any of the FUD that Sony spews to the media (or that the media mis-interprets). The hand waving that happens around this stuff is just crazy, and it is usually highly inaccurate by the time it is translated into English and massaged for the average consumer.
I do not believe that the foundation of the PS3 will be an Internet based distributed computing model. Much more likely is a highly parallel machine with support for tight networks of these machines (i.e. clusters) to (hopefully transparently) distribute this parallelism outside of a single box. To the home user with a single machine, however, he will be limited to however many boxes he buys. I suppose they could sell "upgrades" which allow you to plug in more power, but that sort of thing has never succeeded in the console market in the past and Sony knows it.
<strong>IBM invented FUD, which is the tactic of making early claims of incredible technology in order to preempt other companies' announcements and introduce uncertainty about their products</strong><hr></blockquote>
I believe the term you're looking for is "vaporware", and the intention is not to cause anyone to question the legitimacy of another company's attempt at the same thing. Rather, it's to get consumers to wait "a little bit longer" to get something "from a name they know and trust."
Consider this Apple-based analogy.
<Steve> Hey guess what boys and girls, we're pretty close to getting a G7 out.
Sales of all PowerMacs drop to absolutely zero as people wait for the completely fictional G7.
Comments
<strong>What are Vector Units?</strong><hr></blockquote>
A Vector processing unit (Altivec/Velocity Engine, MMX) is a processing unit capable of perform a Single Instruction on Multiple Data, SIMD. Altivec is a form of this, and it is capable of pushing through 128 bits of data on every tick of the CPU clock. Sony used vector processing to make the two chips used in the Playstation 2. Vector processing works very well in the case of graphics since ussually you are performing the same operation on every pixel in the screen.
I want to know if anyone has any idea about how they can do this. Not if they can do it because it seems they are, regardless of when it comes out. It seems like the idea behind SETI only in real-time. I want to know how. There are some obvious problems with bandwidth. But wouldn't it be cool if Apple brought Marxism to the Mac.
<strong>It has been reported that Playstation 3 due out in 2004 will have some form of "distributed computing"
If IBM and Sony are doing R&D could Apple be doing the same?</strong><hr></blockquote>
Because Apple (NeXt) based OS X on Mach it is entirely possible to support distributed computing at the kernel level. Because the Mach kernel uses a messaging system to compunicate between the micro-kernel and the OS Layer it is possible to have multiple distributed micro-kernels on many different machines. This would make the OS appear to programmers as normal, running on one computer, but actually underneath it could actually be running on a few hundred.
This was only in development at CMU, and never really saw full fruition before the Mach research project there was canceled. Apple could pick up on this research again, and perhaps with XServe they will have a reason. But the framework is there.
[ 07-03-2002: Message edited by: BlueJekyll ]</p>
<strong>it seems obvious that they are being very opptimistic but they aren't lying. While they may have been overly opptimistic about the G3 it does in fact exist. This article says that some type of community processing that could be used for a gaming consol does exist. It isn't talking about a multiprocessor, or about batch processing over a local network, or about improved mutliplayer games.</strong><hr></blockquote>
They might not be lying outright so much as they are erring on the side of rampant optimism. They have plenty of time to backpedal, and there will be no shortage of perfectly legitimate reasons for them to do so.
"Whoops! We were pretty sure broadband would be more widespread by 2005!"
[quote]<strong>I want to know if anyone has any idea about how they can do this. Not if they can do it because it seems they are, regardless of when it comes out. It seems like the idea behind SETI only in real-time. I want to know how. There are some obvious problems with bandwidth. But wouldn't it be cool if Apple brought Marxism to the Mac.</strong><hr></blockquote>
From a technological standpoint, implementing it isn't that difficult. It's been a solved problem for years, in fact. The difficulty lies in determining what benefit distributed computing would offer to gaming consoles. There is also a possible sticking point involving broadband providers: Almost all consumer broadband services throttle upload bandwidth and prohibit servers.
Consider your statement, "like SETI only in real time." There is a vast gulf there. Since SETI is not real time, the computer it's run on only has to be online sporadically, and SETI can upload in the background because there's no hurry. SETI can run on spare cycles and behind screensavers because there's no hurry. The computer can be put to sleep or powered down or even taken apart and reassembled and the central SETI servers will never know or care, because there's no hurry.
Now, consider a game console. Most multiplayer games are real time. All massively multiplayer games are real time, all the time. The immediate implications are that if your console joins this network: you can't shut it off or take it offline; it becomes a server in the strictest sense of the word, violating the terms of service of every cable and ADSL provider I'm aware of; and it has to provide real-time data to both you and some arbitrary number of other people simultaneously. If your PS/3 is responsible for some part of a castle, and your knight is ambushed in a forest just as a party of ten decides to enter the castle, your console is going to get hammered, and something will have to give. Bottom line: It's not a good idea to have the same machine running something as demanding as a game and something as demanding as a game server. I can think of a few ways around this problem that use more than one machine, but there's still the problem that your upload speed is throttled to 2 or 3 56K modems' worth, and you'll get disconnected if your ISP sees to much traffic coming upstream. And, of course, there's always the much greater possibility that one of the console-servers loses power, and - whoops! - there goes the castle, and the PCs in it.
I can see no way that distributed consoles would yield a thousandfold increase in anything, unless the game developers were using them to compile their applications.
None of this has anything to do with the viability of an Apple-standard distributed computing API for Macs. I suppose that Macs would use it in the same way that SETI does: To farm out compute-bound work which can be broken up into chunks, and which does not have to be completed in real time (3D rendering, PostScript RIPping, compositing, scientific simulations) to other machines. In fact, somewhere on these boards there is a photo of an early version of Shake(?) for OS X doing exactly this.
<strong>
</strong><hr></blockquote>
I don't buy into any of the FUD that Sony spews to the media (or that the media mis-interprets). The hand waving that happens around this stuff is just crazy, and it is usually highly inaccurate by the time it is translated into English and massaged for the average consumer.
I do not believe that the foundation of the PS3 will be an Internet based distributed computing model. Much more likely is a highly parallel machine with support for tight networks of these machines (i.e. clusters) to (hopefully transparently) distribute this parallelism outside of a single box. To the home user with a single machine, however, he will be limited to however many boxes he buys. I suppose they could sell "upgrades" which allow you to plug in more power, but that sort of thing has never succeeded in the console market in the past and Sony knows it.
<strong>IBM invented FUD, which is the tactic of making early claims of incredible technology in order to preempt other companies' announcements and introduce uncertainty about their products</strong><hr></blockquote>
I believe the term you're looking for is "vaporware", and the intention is not to cause anyone to question the legitimacy of another company's attempt at the same thing. Rather, it's to get consumers to wait "a little bit longer" to get something "from a name they know and trust."
Consider this Apple-based analogy.
<Steve> Hey guess what boys and girls, we're pretty close to getting a G7 out.
Sales of all PowerMacs drop to absolutely zero as people wait for the completely fictional G7.