Come back when you actually understand something about the technology that you're spouting off about.</strong><hr></blockquote>
I ask that you PROVE that you understand the technology that you are referring to. And to that end I have a program to write. Much more fun than reading MacLuvs' attempts at thinking his way out of a paperbag.
Think and surprise all of us! <img src="graemlins/lol.gif" border="0" alt="[Laughing]" /> <img src="graemlins/lol.gif" border="0" alt="[Laughing]" /> <img src="graemlins/lol.gif" border="0" alt="[Laughing]" />
[quote]Because IBM stated that they will start sampling 970 in Q2, that means nothing is going to be available earlier. IBM can say anything they want but what they do can be different from time to time.<hr></blockquote>
Exactly, and has anyone thought of this: Those release dates may be for a "general" release of the 970. Certainly there could be a "private" contract between the two providing Apple with chips at an earlier date. Might be called something other than a 970, but could still be the same, or even customized more specifically for Apple.
"Exactly, and has anyone thought of this: Those release dates may be for a "general" release of the 970. Certainly there could be a "private" contract between the two providing Apple with chips at an earlier date. Might be called something other than a 970, but could still be the same, or even customized more specifically for Apple."
It would be nice to have surprise...
Didn't G4 surprise everybody when it was first launched. Said it was coming much later than arrived. Similar occurence..? That would buy Apple 970 launch to shipping time. A stretch...but in reach of a San Fran launch?
I still think Kid Red's dual 1.4 is more likely. Dual 1 gig at low end? Tasty Mac?
Didn't G4 surprise everybody when it was first launched. Said it was coming much later than arrived.</strong><hr></blockquote>
Yeah, and shortly afterward Apple found themselves wishing they'd waited another few months, the way everyone thought they would.
The G4 was introduced to the world with a "speed dump," followed by stagnation apparently caused by a coherency problem with the L1 cache. This alone should be enough to convince people that the last thing the 970 needs is a similar rush job. We all want the thing yesterday, but we want it finished, tested, and working, right?
1. This thread is about the next generation chip that Apple will place it its pro lineup - better known by the meaningless vapor-brand "G5". This thread is not about OS X, Quack Xless, or any other application.
2. That said, while Clive's understanding of Apple's current market status (and what's driving that status) isn't technically anymore grounded in fact than the arguments of his counterparts, he is generally correct I think.
No one ever said better technology means automatic marketshare or profitability improvements. If they did, the existing number of Quark users would be drastically smaller by now, and the number of switchers drastically higher - among other things.
I frequently hear comments that go something like this from would-be Apple users: "I've always thought Apple's technology was better, but the simple fact is their machines always cost too much relative to higher end PCs". So its not superior technology Apple has to convince people of, so much as they have to deliver serious bang for the buck - especially during hard economic times - and they simply do not do this.
3. This statement
"When the 970 comes, Quark will release a mostly 32-bit carbon product, while Indesign will be mostly be 64-bit Cocoa."
is absurd to the point of being funny.
A) Applications aren't going to be "Mostly 64-bit" or "Partly 32-bit" - even if they are built on modular designs. They will either be compiled as 64-bit apps, or they won't be. 3rd party plugins not withstanding of course.
Unless you work there, you have no idea what Adobe's plans are for releasing 64-bit PPC products down the road, much less what the timeframes for release are.
C) NONE of their products are going to be Cocoa-based. I won't go into why because it's been plastered on these boards over and over for almost 2 years now. Adobe chose Carbon for a reason. If you care to know why, do some research - it's not hard to find explanations.
4) I think the only changes to the Power Mac line that we could possibly see on Jan 7th are mild speed-ups with a 200MHz bus. No major architectural changes, no "G5". The evidence overwhelmingly supports G4-based Power Macs well into next year, if not into 2004.
I'm not even sure why we're sweating this so much (although it seems expectations are much more reasonable than they usually are going into MWSF).
I say that because frankly, I agree with the point that for most users, the current bandwidth-choked PM line still performs very well on most tasks. Granted video, 3D and music production people could use a big boost for the money - but most people who buy G4 towers use them for things like running Photoshop and Dreamweaver in tandem, or playing UT, or doing MS Office stuff. Even though we're focusing on all these professional production issues (publishing in this case), I think those people are likely outnumbered by the "typical user scenario" noted above.
I'm using an old G4/500, and while the app launch times and certain render-sensitive tasks are sluggish under Jaguar, I can still get a lot done during the course of my free days. I could get quite a bit more done with even the Dual 867 vs what I have now. Point being, if Apple could release Power Macs with a 200MHz bus in January, and are smart enough to drop the prices another $150 - the machines would be a big hit with most individual G4 owners IMO.
Not ideal and not what we really need for the long haul, but enough to boost sales quite a bit while giving users "more bang for less bucks" - and isn't that the point?
You are right maybe with heavy disk intensive tasks the DDR Power Macs pull ahead...not by much but they probably do. But I am talking about every day tasks. I use FCP a LOT. I had a sdram Power Mac and I almost chose the DDR GHz Power Mac to replace it but after I saw benchmarks I went with the Dual 867.
My point is this: I had to spend extra for the DDR ram. Does the current G4 processor correctly support that DDR ram? Is there stil a large bottleneck between RAM and the G4 as there has been all along? Yes. These machines are the result of Apple Marketing trying to put a positive spin on a bad situation.
and who was talking about the Xserve? I thought we were discussing Power Macs..
NONE of [Adobe's] products are going to be Cocoa-based. I won't go into why because it's been plastered on these boards over and over for almost 2 years now. Adobe chose Carbon for a reason. If you care to know why, do some research - it's not hard to find explanations.
<hr></blockquote>
Moogs:
You could have gestured at an explanation in the same amount of space you used for ranting.
Anyhow, I searched for "cocoa" in frameusers, and came across a statement from an Adobe rep (dated jan 2001) to the effect that cocoa is unsuitable for cross-platform development. Is this the substance of your objection?
Others:
Despite this being a FH thread, I'd like to ask: is cocoa suitable for cross-platform development?
You could have gestured at an explanation in the same amount of space you used for ranting.
Anyhow, I searched for "cocoa" in frameusers, and came across a statement from an Adobe rep (dated jan 2001) to the effect that cocoa is unsuitable for cross-platform development. Is this the substance of your objection?
Others:
Despite this being a FH thread, I'd like to ask: is cocoa suitable for cross-platform development?</strong><hr></blockquote>
NeXT Step has been on Alpha(maybe), Moto 68000, Intel, Spark, and was the idea germ for Java. The idea was about portability. I believe that what they are referring to is that developing a project for MacOSX, which is NextStep kind of, would require allot of work to port over to Windows, even if they were both on Intel hardware. Which is what Programmer, and others have said. Cocoa is mapped to a series of calls to Mach/BSD. Once you have the kernal working on some hardware the frameworks are much easier to port over. I believe that the NeXT guys had this done for PPC within a few months with NeXT 4. I would like to point out that Apple had some guys fly to Apple for a weekend to port their software over to Next4 on PPC, might have been the guys from Omni, I can't remember. That's right one weekend, but this was from NeXT to NeXT, just different hardware. History, ramble But there was an agreement that what Apple would buy NeXT 5 and so as the development went along Steve took charge, the plan was broadened. Adobe decided to drop display PostScript at Microsofts' request, which ment that Apple had to develope display PDF which delayed the launch of OSX by more than a year but Apple released it in a year. But we all lived through the 10.0 release and could see that we were all living through that delay. In a way it is better, we have Quartz and display PDF. Had it not been for MS we would have been using display postscript. Hence Apples' strong embrace of open standards, MS cannot request that OpenGL be dropped to the OpenGL group and have much effect.
Back on track: It seems that January would be rushed for the 970, I hope it to be early, I hope it is here in July.
My appologies for the ramble. I was attempting to answer a question that appeared to be running.
[ 12-30-2002: Message edited by: Brendon ] Added part about weekend development.
Despite this being a FH thread, I'd like to ask: is cocoa suitable for cross-platform development?</strong><hr></blockquote>
Not really. There are exceptions: Omni Group ported Giants: Citizen Kabuto to Cocoa. But then, games are their own thing, since they take over the screen and provide custom user interfaces.
The big problem is that it's a framework (actually, several frameworks), which means that you get tremendous amounts of functionality from a few lines of code ... which means that the code is well-nigh useless on a platform that doesn't have those frameworks. Currently, the closest equivalent to Cocoa on another platform is GNUStep, and that's not near close enough.
The other problem was Objective-C, which does run on every platform gcc runs on, but which is still an unheard-of choice for cross-platform work. That's generally done in C++ or Java. Almost all Windows development is VB or C++, neither of which looks or acts like Objective-C. Objective-C++ blunts that objection somewhat, as does Cocoa for Java. But it doesn't change the fact that anyone looking to share code with Windows is going to prefer C++.
Carbon, on the other hand, can be used in straight C++ (for that matter, straight C), and it's almost immediately familiar to seasoned Mac programmers, so it's the more obvious choice for companies like Adobe which have millions of lines of existing C and C++ code, support for Mac and Windows alike, and lots of seasoned Mac programmers.
<strong>It's amazing how defensive some people can be...
You are right maybe with heavy disk intensive tasks the DDR Power Macs pull ahead...not by much but they probably do. But I am talking about every day tasks. I use FCP a LOT. I had a sdram Power Mac and I almost chose the DDR GHz Power Mac to replace it but after I saw benchmarks I went with the Dual 867.
My point is this: I had to spend extra for the DDR ram. Does the current G4 processor correctly support that DDR ram? Is there stil a large bottleneck between RAM and the G4 as there has been all along? Yes. These machines are the result of Apple Marketing trying to put a positive spin on a bad situation.
and who was talking about the Xserve? I thought we were discussing Power Macs.. </strong><hr></blockquote>
Xserves and MDD Macs share the same tech.
Most benchmarks only test a fraction of a program - eg. renders and things like that.
They don't test for network speed and so on, and I'll bet that most pros are on a network.
Regarding bandwidth: The bandwidth between a G4 and RAM is around 1GB/sec, and on a SDRAM machine the bandwidth of the RAM is 1GB/sec.
That 1GB/sec needs to be shared among the G4 and the DMA components, but on a DDR RAM Mac the G4 has 1GB/sec. to itself and the DMA components share 1GB/sec.
<strong>Adobe decided to drop display PostScript at Microsofts' request, which ment that Apple had to develope display PDF which delayed the launch of OSX by more than a year but Apple released it in a year.</strong><hr></blockquote>
Apple dumped Display PS because of a not so nice license fee (PDF is free), and Adobe wasn't happy about Apple dumping Display PS.
Most benchmarks only test a fraction of a program - eg. renders and things like that.
They don't test for network speed and so on, and I'll bet that most pros are on a network.
Regarding bandwidth: The bandwidth between a G4 and RAM is around 1GB/sec, and on a SDRAM machine the bandwidth of the RAM is 1GB/sec.
That 1GB/sec needs to be shared among the G4 and the DMA components, but on a DDR RAM Mac the G4 has 1GB/sec. to itself and the DMA components share 1GB/sec.</strong><hr></blockquote>
Just to contribute a little to your discussion about the value of the DDR in the MDD PowerMacs -- the standard benchmarking methodology is to isolate exactly one thing so that you can time that. Unfortunately that often isn't how these machines are used and thus doesn't attempt to test important cases in practical use.
Case in point: there are two major bandwidth consumers in the system, the CPU and the GPU bus. The CPU's MPX can use about 1 GB/sec, and the GPU's AGP can use about 1 GB/sec. On non-DDR Macs there is about 1 GB/sec of memory bandwidth so either the CPU or the GPU can max out, but if they are both operating at once they need to share which impacts their performance. On the DDR machines there is 2.7 GB/sec of streaming bandwidth (theoretically) so there ought to be enough to go around.
In reality we don't know how close Apple's memory controller gets to the theoretical 2.7 GB/sec, but its probably >1.8 GB/sec. As a result, in situations where both the CPU and the GPU are accessing the system's main memory at high speed the DDR machines ought to do better than the non-DDR machines. Quartz Extreme should make this a valuable feature, but you need to be using the GUI actively while running a computational test in the background -- and you have to figure out when QE is using the AGP bus because it might outsmart you and cache things in VRAM. So its a difficult benchmark test to do, but its the kind of thing that will benefit you in day-to-day work, rather than just in benchmarks.
The bus speed is higher too, remember, (133 -> 167 MHz) so even the CPU's memory bandwidth by itself is higher and this has been measured in some benchmarks.
The downside of the MDD machines is that Apple shrunk the L3 cache to 1 MB/processor instead of 2. There will be some cases where this negatively affects performance slightly. Most of time the improvement in bus performance and greater available system bandwidth outweighs the loss of cache.
In a nutshell: In these Macs the DDR RAM bandwidth is being used, just not by the CPUs directly.
<a href="http://www-3.ibm.com/chips/news/2002/1216_xilinx.html"; target="_blank">Here</a>, we have IBM and Xilinx announcing 90nm production in 2nd half of 2003. IBM has stated in the past that the 970 will be initially released as a 130nm product. So if we have 90nm in 2nd half of 2003, who's to say we won't have 130nm product (and new Powermacs, yum!) much sooner, say 1Q 2003?
I have to agree with Rhumgod, the IBM article states that IBM plans to manufacture the new product line in volume in the second half of 2003 at the company's state-of-the-art 300mm plant in East Fishkill, N.Y. . so if the 90nm will be available in the second half of 2003, what about the 130nm chips? Well MacWorld SF will be all consumer iMac, eMac, iApps and may be a digital device. Not sure about the portables? But since Apple looks like it wants out of the big MacWorld trade shows 4 times a year and may want to try to organize its own, smaller, regional conferences. Apple already produces Macworld in Paris. So may be there first one in North America will be a new PowerMac introduced late January to March.
It says what I've been saying for a few months. A Moto G4+ of sorts in Jan/Feb and the 970 by Sept/Aug. I figure the G4 will be up to about 1.4ghz and the 970 starts at 1.4ghz so the high end becomes the low end as norm when the 970 comes out. The 1.8ghz will probably be to top end. I can't wait till the fall
Comments
I think I'd settle for that. If 2003 gets a 970 and Quark X Mac then I'm sure Apple will be happy for the time being...
Lemon Bon Bon
[quote]Originally posted by Programmer:
<strong>
Come back when you actually understand something about the technology that you're spouting off about.</strong><hr></blockquote>
I ask that you PROVE that you understand the technology that you are referring to. And to that end I have a program to write. Much more fun than reading MacLuvs' attempts at thinking his way out of a paperbag.
Think and surprise all of us! <img src="graemlins/lol.gif" border="0" alt="[Laughing]" /> <img src="graemlins/lol.gif" border="0" alt="[Laughing]" /> <img src="graemlins/lol.gif" border="0" alt="[Laughing]" />
Exactly, and has anyone thought of this: Those release dates may be for a "general" release of the 970. Certainly there could be a "private" contract between the two providing Apple with chips at an earlier date. Might be called something other than a 970, but could still be the same, or even customized more specifically for Apple.
What the hey?
It would be nice to have surprise...
Didn't G4 surprise everybody when it was first launched. Said it was coming much later than arrived. Similar occurence..? That would buy Apple 970 launch to shipping time. A stretch...but in reach of a San Fran launch?
I still think Kid Red's dual 1.4 is more likely. Dual 1 gig at low end? Tasty Mac?
LEmBONB
[ 12-29-2002: Message edited by: Lemon Bon Bon ]</p>
<strong>It would be nice to have surprise...
Didn't G4 surprise everybody when it was first launched. Said it was coming much later than arrived.</strong><hr></blockquote>
Yeah, and shortly afterward Apple found themselves wishing they'd waited another few months, the way everyone thought they would.
The G4 was introduced to the world with a "speed dump," followed by stagnation apparently caused by a coherency problem with the L1 cache. This alone should be enough to convince people that the last thing the 970 needs is a similar rush job. We all want the thing yesterday, but we want it finished, tested, and working, right?
[ 12-29-2002: Message edited by: Amorph ]</p>
1. This thread is about the next generation chip that Apple will place it its pro lineup - better known by the meaningless vapor-brand "G5". This thread is not about OS X, Quack Xless, or any other application.
2. That said, while Clive's understanding of Apple's current market status (and what's driving that status) isn't technically anymore grounded in fact than the arguments of his counterparts, he is generally correct I think.
No one ever said better technology means automatic marketshare or profitability improvements. If they did, the existing number of Quark users would be drastically smaller by now, and the number of switchers drastically higher - among other things.
I frequently hear comments that go something like this from would-be Apple users: "I've always thought Apple's technology was better, but the simple fact is their machines always cost too much relative to higher end PCs". So its not superior technology Apple has to convince people of, so much as they have to deliver serious bang for the buck - especially during hard economic times - and they simply do not do this.
3. This statement
"When the 970 comes, Quark will release a mostly 32-bit carbon product, while Indesign will be mostly be 64-bit Cocoa."
is absurd to the point of being funny.
A) Applications aren't going to be "Mostly 64-bit" or "Partly 32-bit" - even if they are built on modular designs. They will either be compiled as 64-bit apps, or they won't be. 3rd party plugins not withstanding of course.
C) NONE of their products are going to be Cocoa-based. I won't go into why because it's been plastered on these boards over and over for almost 2 years now. Adobe chose Carbon for a reason. If you care to know why, do some research - it's not hard to find explanations.
4) I think the only changes to the Power Mac line that we could possibly see on Jan 7th are mild speed-ups with a 200MHz bus. No major architectural changes, no "G5". The evidence overwhelmingly supports G4-based Power Macs well into next year, if not into 2004.
I'm not even sure why we're sweating this so much (although it seems expectations are much more reasonable than they usually are going into MWSF).
I say that because frankly, I agree with the point that for most users, the current bandwidth-choked PM line still performs very well on most tasks. Granted video, 3D and music production people could use a big boost for the money - but most people who buy G4 towers use them for things like running Photoshop and Dreamweaver in tandem, or playing UT, or doing MS Office stuff. Even though we're focusing on all these professional production issues (publishing in this case), I think those people are likely outnumbered by the "typical user scenario" noted above.
I'm using an old G4/500, and while the app launch times and certain render-sensitive tasks are sluggish under Jaguar, I can still get a lot done during the course of my free days. I could get quite a bit more done with even the Dual 867 vs what I have now. Point being, if Apple could release Power Macs with a 200MHz bus in January, and are smart enough to drop the prices another $150 - the machines would be a big hit with most individual G4 owners IMO.
Not ideal and not what we really need for the long haul, but enough to boost sales quite a bit while giving users "more bang for less bucks" - and isn't that the point?
You are right maybe with heavy disk intensive tasks the DDR Power Macs pull ahead...not by much but they probably do. But I am talking about every day tasks. I use FCP a LOT. I had a sdram Power Mac and I almost chose the DDR GHz Power Mac to replace it but after I saw benchmarks I went with the Dual 867.
My point is this: I had to spend extra for the DDR ram. Does the current G4 processor correctly support that DDR ram? Is there stil a large bottleneck between RAM and the G4 as there has been all along? Yes. These machines are the result of Apple Marketing trying to put a positive spin on a bad situation.
and who was talking about the Xserve? I thought we were discussing Power Macs..
NONE of [Adobe's] products are going to be Cocoa-based. I won't go into why because it's been plastered on these boards over and over for almost 2 years now. Adobe chose Carbon for a reason. If you care to know why, do some research - it's not hard to find explanations.
<hr></blockquote>
Moogs:
You could have gestured at an explanation in the same amount of space you used for ranting.
Anyhow, I searched for "cocoa" in frameusers, and came across a statement from an Adobe rep (dated jan 2001) to the effect that cocoa is unsuitable for cross-platform development. Is this the substance of your objection?
Others:
Despite this being a FH thread, I'd like to ask: is cocoa suitable for cross-platform development?
<strong>
Moogs:
You could have gestured at an explanation in the same amount of space you used for ranting.
Anyhow, I searched for "cocoa" in frameusers, and came across a statement from an Adobe rep (dated jan 2001) to the effect that cocoa is unsuitable for cross-platform development. Is this the substance of your objection?
Others:
Despite this being a FH thread, I'd like to ask: is cocoa suitable for cross-platform development?</strong><hr></blockquote>
NeXT Step has been on Alpha(maybe), Moto 68000, Intel, Spark, and was the idea germ for Java. The idea was about portability. I believe that what they are referring to is that developing a project for MacOSX, which is NextStep kind of, would require allot of work to port over to Windows, even if they were both on Intel hardware. Which is what Programmer, and others have said. Cocoa is mapped to a series of calls to Mach/BSD. Once you have the kernal working on some hardware the frameworks are much easier to port over. I believe that the NeXT guys had this done for PPC within a few months with NeXT 4. I would like to point out that Apple had some guys fly to Apple for a weekend to port their software over to Next4 on PPC, might have been the guys from Omni, I can't remember. That's right one weekend, but this was from NeXT to NeXT, just different hardware. History, ramble But there was an agreement that what Apple would buy NeXT 5 and so as the development went along Steve took charge, the plan was broadened. Adobe decided to drop display PostScript at Microsofts' request, which ment that Apple had to develope display PDF which delayed the launch of OSX by more than a year but Apple released it in a year. But we all lived through the 10.0 release and could see that we were all living through that delay. In a way it is better, we have Quartz and display PDF. Had it not been for MS we would have been using display postscript. Hence Apples' strong embrace of open standards, MS cannot request that OpenGL be dropped to the OpenGL group and have much effect.
Back on track: It seems that January would be rushed for the 970, I hope it to be early, I hope it is here in July.
My appologies for the ramble. I was attempting to answer a question that appeared to be running.
[ 12-30-2002: Message edited by: Brendon ] Added part about weekend development.
[ 12-30-2002: Message edited by: Brendon ]
[ 12-30-2002: Message edited by: Brendon ]</p>
<strong>
Despite this being a FH thread, I'd like to ask: is cocoa suitable for cross-platform development?</strong><hr></blockquote>
Not really. There are exceptions: Omni Group ported Giants: Citizen Kabuto to Cocoa. But then, games are their own thing, since they take over the screen and provide custom user interfaces.
The big problem is that it's a framework (actually, several frameworks), which means that you get tremendous amounts of functionality from a few lines of code ... which means that the code is well-nigh useless on a platform that doesn't have those frameworks. Currently, the closest equivalent to Cocoa on another platform is GNUStep, and that's not near close enough.
The other problem was Objective-C, which does run on every platform gcc runs on, but which is still an unheard-of choice for cross-platform work. That's generally done in C++ or Java. Almost all Windows development is VB or C++, neither of which looks or acts like Objective-C. Objective-C++ blunts that objection somewhat, as does Cocoa for Java. But it doesn't change the fact that anyone looking to share code with Windows is going to prefer C++.
Carbon, on the other hand, can be used in straight C++ (for that matter, straight C), and it's almost immediately familiar to seasoned Mac programmers, so it's the more obvious choice for companies like Adobe which have millions of lines of existing C and C++ code, support for Mac and Windows alike, and lots of seasoned Mac programmers.
[ 12-30-2002: Message edited by: Amorph ]</p>
<strong>It's amazing how defensive some people can be...
You are right maybe with heavy disk intensive tasks the DDR Power Macs pull ahead...not by much but they probably do. But I am talking about every day tasks. I use FCP a LOT. I had a sdram Power Mac and I almost chose the DDR GHz Power Mac to replace it but after I saw benchmarks I went with the Dual 867.
My point is this: I had to spend extra for the DDR ram. Does the current G4 processor correctly support that DDR ram? Is there stil a large bottleneck between RAM and the G4 as there has been all along? Yes. These machines are the result of Apple Marketing trying to put a positive spin on a bad situation.
and who was talking about the Xserve? I thought we were discussing Power Macs..
Xserves and MDD Macs share the same tech.
Most benchmarks only test a fraction of a program - eg. renders and things like that.
They don't test for network speed and so on, and I'll bet that most pros are on a network.
Regarding bandwidth: The bandwidth between a G4 and RAM is around 1GB/sec, and on a SDRAM machine the bandwidth of the RAM is 1GB/sec.
That 1GB/sec needs to be shared among the G4 and the DMA components, but on a DDR RAM Mac the G4 has 1GB/sec. to itself and the DMA components share 1GB/sec.
<strong>Adobe decided to drop display PostScript at Microsofts' request, which ment that Apple had to develope display PDF which delayed the launch of OSX by more than a year but Apple released it in a year.</strong><hr></blockquote>
Apple dumped Display PS because of a not so nice license fee (PDF is free), and Adobe wasn't happy about Apple dumping Display PS.
<strong>
Xserves and MDD Macs share the same tech.
Most benchmarks only test a fraction of a program - eg. renders and things like that.
They don't test for network speed and so on, and I'll bet that most pros are on a network.
Regarding bandwidth: The bandwidth between a G4 and RAM is around 1GB/sec, and on a SDRAM machine the bandwidth of the RAM is 1GB/sec.
That 1GB/sec needs to be shared among the G4 and the DMA components, but on a DDR RAM Mac the G4 has 1GB/sec. to itself and the DMA components share 1GB/sec.</strong><hr></blockquote>
Just to contribute a little to your discussion about the value of the DDR in the MDD PowerMacs -- the standard benchmarking methodology is to isolate exactly one thing so that you can time that. Unfortunately that often isn't how these machines are used and thus doesn't attempt to test important cases in practical use.
Case in point: there are two major bandwidth consumers in the system, the CPU and the GPU bus. The CPU's MPX can use about 1 GB/sec, and the GPU's AGP can use about 1 GB/sec. On non-DDR Macs there is about 1 GB/sec of memory bandwidth so either the CPU or the GPU can max out, but if they are both operating at once they need to share which impacts their performance. On the DDR machines there is 2.7 GB/sec of streaming bandwidth (theoretically) so there ought to be enough to go around.
In reality we don't know how close Apple's memory controller gets to the theoretical 2.7 GB/sec, but its probably >1.8 GB/sec. As a result, in situations where both the CPU and the GPU are accessing the system's main memory at high speed the DDR machines ought to do better than the non-DDR machines. Quartz Extreme should make this a valuable feature, but you need to be using the GUI actively while running a computational test in the background -- and you have to figure out when QE is using the AGP bus because it might outsmart you and cache things in VRAM. So its a difficult benchmark test to do, but its the kind of thing that will benefit you in day-to-day work, rather than just in benchmarks.
The bus speed is higher too, remember, (133 -> 167 MHz) so even the CPU's memory bandwidth by itself is higher and this has been measured in some benchmarks.
The downside of the MDD machines is that Apple shrunk the L3 cache to 1 MB/processor instead of 2. There will be some cases where this negatively affects performance slightly. Most of time the improvement in bus performance and greater available system bandwidth outweighs the loss of cache.
In a nutshell: In these Macs the DDR RAM bandwidth is being used, just not by the CPUs directly.
<a href="http://www-3.ibm.com/chips/news/2002/1216_xilinx.html"; target="_blank">Here</a>, we have IBM and Xilinx announcing 90nm production in 2nd half of 2003. IBM has stated in the past that the 970 will be initially released as a 130nm product. So if we have 90nm in 2nd half of 2003, who's to say we won't have 130nm product (and new Powermacs, yum!) much sooner, say 1Q 2003?
Only time will tell
<strong>Mmm? new rumor: <a href="http://www.geek.com/news/geeknews/2002Dec/bch20021230017949.htm"; target="_blank">http://www.geek.com/news/geeknews/2002Dec/bch20021230017949.htm</a></strong><hr></blockquote>;
It says what I've been saying for a few months. A Moto G4+ of sorts in Jan/Feb and the 970 by Sept/Aug. I figure the G4 will be up to about 1.4ghz and the 970 starts at 1.4ghz so the high end becomes the low end as norm when the 970 comes out. The 1.8ghz will probably be to top end. I can't wait till the fall