The only timeline I ever heard was part of the news.com story that broke the transition news. The timing might make perfect sense, but that isn't the same as being confirmed by Apple.
whatever happens it is clear to me that we are in for quite a ride. The message to me here is that we don't 100% what to expect from Apple. They must have seen something very cool on the Intel roadmap to make this shift. I also think that Apple is positioning itself to get out of its box one way or another. I am continually amazed at the way Apple engineers and designs products and delivers experiences to consumers. Moving into the Intel camp may be a break with a lot of cool stuff that came with the ppc, but it may also signal the beginning of an entirely new adventure for the platform.
Well, I can't speak for everyone, but not having end notes is a non-starter for me and my wife (academic writing). When I played around with it in my local Apple store, I liked a lot of it, but felt that it hadn't quite figured out how it was going to handle certain issues given its dtp-like model (mostly, deleting pages and linking/flowing text to boxes). I'm hopeful that Pages 2 will take care of those issues.
Also, I've read that pdf exporting isn't perfect across platforms (drop shadows and the like).
They fixed the 'Delete Page' thing in the v1.01 update although I still think it's badly done since a page could actually be 100 pages if you've just kept typing without sticking in Page Breaks. It's more of a section really. It's just badly termed/implemented but it's now there. How they missed 'Delete Page' in v1.0 though is beyond belief. The developers must have been too close to their perfect ideal and not saw the error.
Text flow works for me - In fact it's one of the things that works stupidly well in comparison to Word. I love it. Click outside the page (in the margin, Click Insert->Text and a box appears. Keep typing and a + appears on the bottom of the box. Click off the box, then on again and small blue boxes appear left/right on the box allowing you to place overflow boxes before or after. And it all happens in realtime flowing between them.
PDF problems are because Apple used the very latest PDF spec that supports drop shadows fully. Some printer drivers and for that matter some print shops don't support that spec. Even Adobe tools don't all support it. It's simply fixed. Instead of using export, use Print and 'Save As...' then select PDF-X as the format. This is a lowest common denominator PDF standard that renders the drop shadows as bitmaps.
Quote:
Originally posted by DCQ
OK. I admit I don't know exactly what problems they'd face in a corporate intranet environment. (I'm imagining my old employer: 12,000 employees globally; 56 offices; 22 countries.) They may work well. But I'd guess that things like integrating iCal functionality into both Mail and iChat would be a minimal necessity (i.e., sending a Calendar invitation as an "attachment"). I don't know...that was just off the top of my head.
[/B]
You can send iCal invites already. Been there for years.
Put an event in to iCal, right click on it, click 'Mail Event'. It opens a new email with an invite bit of text, plus an iCal .ics file that the person can click on to add it to their calendar.
Again, I think this is just one of those things where because it doesn't work like Office or Outlook, people don't find these things, or don't try.
How about AddressBook and iCal - did you know you can drag contacts from address book to an event's attendees list or just dropping a group onto a spare slot in iCal automatically sets up a meeting with all those in the group?
It's not perfect though I'll give you that. It'd be nice to share calendars without having to set up an internal webdav server or use .Mac and it's not as simple as Outlook to publish/subscribe calendars but given some reasonable training or semi-intelligent users it's perfectly practical IME.
Only problem is, like Outlook, everyone has to use the same calendar application so that kind of means everyone has to use a Mac in the office.
I've seen this comment a couple of times, but I don't believe it to be true. The OSX thread implementation is based on the Mach threaded kernel and so ought to be very efficient. The pthreads are a thin API wrapper on that, and the Cocoa & Carbon threads are a API wrapper on pthreads.
[/B]
The following predates Tiger but comparing Linux v OSX on the same G5...
Look at the LMBench scores, some of which are just shocking.
Then there was also the recent Anandtech MySQL and Apache shootout which had similar problems although I hate to quote that as proof as it was fundamentally flawed.
A lot of it was concluded to be the file system and Dominic Giampolo said it was because HFS+ was 'right' whereas EXT3 had greater potential to lose data but having run webservers for a number of years, seeing performance not just a few percent better but a few 100 percent better, I'll stick with Linux thanks there.
In Tiger, I believe they got rid of the kernel funnel and network funnel and made the kernel lock at a finer level.
A lot of it was concluded to be the file system and Dominic Giampolo said it was because HFS+ was 'right' whereas EXT3 had greater potential to lose data but having run webservers for a number of years, seeing performance not just a few percent better but a few 100 percent better, I'll stick with Linux thanks there.
For those purposes there are faster file systems around. Hell, Dominic knows all too well just how frickin fast and nice BFS was, but that is long gone, alas. HFS+ may get there, but I think some changes are underway. I read an article on ARS about Spotlight and how it uses an external database for the purpose of storing the metadata, rather than the file system itself. Not that that is any reason to draw any conclusions, but it seems sorta stop gap to me. I guess time will tell.
Look at the LMBench scores, some of which are just shocking.
When you see results like this there is typically "something strange going on" that is being exposed by the particular test. Even the author acknowledges this:
Quote:
However, a few stunningly bad results on OSX make me wonder if all parts of the suite were running accurately.
He also goes on to state that he's not a benchmarking expert and doesn't properly understand what he's doing. So I'd be inclined to ignore the 500x difference on the one test because my Macs sure as heck aren't 500x slower than my PCs or Linux boxes. The 1-2x differences are certainly believeable, but every release of OSX has been getting faster and that is because Apple is addressing these kinds of issues over time.
Of all those LMBench tests, however, none of them really say much about MacOSX "threading" per-se. Most reflect on the file and memory systems. The context switch time for the smallest unit of work is about the closest to a measurement of threading, and it is about 10-20% slower. The larger amounts of work start reflecting on the VM setup of MacOS X. It would be interesting to see these tests re-run to compare 10.3.4 vs. 10.4.1 to see how much progress Apple is making on the low-level issues.
None of these really says much about how effectively MacOS X could take advantage of more cores compared to other OSes. These test are measure context switching times on a single processor... more cores means fewer context switches which means that MacOS X might actually benefit more from more cores than an OS that is more optimized.
Quote:
Then there was also the recent Anandtech MySQL and Apache shootout which had similar problems although I hate to quote that as proof as it was fundamentally flawed.
A lot of it was concluded to be the file system and Dominic Giampolo said it was because HFS+ was 'right' whereas EXT3 had greater potential to lose data but having run webservers for a number of years, seeing performance not just a few percent better but a few 100 percent better, I'll stick with Linux thanks there.
Yes, the flush in OSX actually works. That's like writing a sort function that is 50 times faster but doesn't generate a sorted array as output. And, FWIW, Dominic said you could disable the flush and performance would increase dramatically. The real bugs are in Linux and the server software -- (a) for having a non-functional flush and (b) for calling flush inappropriately. Apple is working to fix the server software so that they don't have to break the OS to make it performant.
Quote:
In Tiger, I believe they got rid of the kernel funnel and network funnel and made the kernel lock at a finer level.
Yes, plus they have put more internal APIs in place which will allow them to make deeper changes in how the OS works in the future.
How efficient would a dual dual-core be? I'm guessing that most users get little benefit right now from duals, let alone quads. I'm sure there are some who use apps that take advantage of them, or use multiple apps simultaneously. But quads?
Do not look at it this way. Mac OS X handles perfectly multiple processors and will distribute the load accordingly. This means that you can fully utilise the four cores by running, for example, three or four independent CPU-intensive applications. Each one of them will take one processor to go.
Now telling about the efficiency is not that simple. There has been some speculation (I don't remember if it was based on info from the leaked IBM document on the 970MP) that the 970MP chip will sit on the G5 bus we have today and that there will be a special "bridge" to allow a very fast inter-chip communication between the two cores. Or something like that. So, in a quad system, we would have two chips with two independent FSB as of today, but each chip will have two processing cores, each capable to speak to the other using the special bridge without needing the FSB. What the efficiency of such a design is, remains to be seen.
A note of interest about quads coming. Apple boldly emphasizes the fine grain locking feature in Tiger (the first in the "New in Tiger" section). Now read this about funnels and Mac OS X (Programmer already made a remark on that). And then remember the 4 CPU reference in the CHUD tools that leaked some time ago. Oh, add to that the 970MP announcement. Oh, and the fact that the Power Mac need to be updated at least once in the coming two years with PowerPC processors, before the Intel switch. Does it start to make sense now ?
Do not look at it this way. Mac OS X handles perfectly multiple processors and will distribute the load accordingly. This means that you can fully utilise the four cores by running, for example, three or four independent CPU-intensive applications. Each one of them will take one processor to go.
Now telling about the efficiency is not that simple. There has been some speculation (I don't remember if it was based on info from the leaked IBM document on the 970MP) that the 970MP chip will sit on the G5 bus we have today and that there will be a special "bridge" to allow a very fast inter-chip communication between the two cores. Or something like that. So, in a quad system, we would have two chips with two independent FSB as of today, but each chip will have two processing cores, each capable to speak to the other using the special bridge without needing the FSB. What the efficiency of such a design is, remains to be seen.
A note of interest about quads coming. Apple boldly emphasizes the fine grain locking feature in Tiger (the first in the "New in Tiger" section). Now read this about funnels and Mac OS X (Programmer already made a remark on that). And then remember the 4 CPU reference in the CHUD tools that leaked some time ago. Oh, add to that the 970MP announcement. Oh, and the fact that the Power Mac need to be updated at least once in the coming two years with PowerPC processors, before the Intel switch. Does it start to make sense now ?
I would also imagine that the new 1MB per core L2 is also because both cores will ride on the same FSB, unlike the older single core cpu's. This would lessen the dependence on the bus throughput and contention that would result from the two cores sharing that bus.
Of course, if what I remember is correct, the 970 can also use a full 1:1 speed FSB as well. It would certainly be interesting to see if Apple considered going to that, or whether it would simply be to expensive with the changes in the support chips that would be required. That would sure shake things up!
I would also imagine that the new 1MB per core L2 is also because both cores will ride on the same FSB, unlike the older single core cpu's. This would lessen the dependence on the bus throughput and contention that would result from the two cores sharing that bus.
I don't know, more L2 helps in many things. Maybe inter-chip communication is one of them.
Quote:
Of course, if what I remember is correct, the 970 can also use a full 1:1 speed FSB as well. It would certainly be interesting to see if Apple considered going to that, or whether it would simply be to expensive with the changes in the support chips that would be required. That would sure shake things up!
Yeah, you are right, I forgot about that. I doubt it though, because of power dissipation (with everything this implies) and cost considerations.
I don't know, more L2 helps in many things. Maybe inter-chip communication is one of them.
Yeah, you are right, I forgot about that. I doubt it though, because of power dissipation (with everything this implies) and cost considerations.
Something else that we don't know at this point is what similarities and differences exist between the cores in the MP and the FX low power(LP).
We know that the LP has a 512KB L2 vs a 1MB L2. But that's it! I'm assuming that otherwise they are pretty much the same, except for whatever is neccessary for the interchip communication. Is the MP core PL like the FX? I would think so. Any other differences???
We don't know far more about these chips than we do. Support chips???
If IBM intends the FX LP to be used for mobile machines, they MUST have some reason as to why this would be possible.
If it's really just intended for embedded operations, it might not be as important, depending on what they are.
Right now we're blowing a lot of hot air over cold coals.
We know that the LP has a 512KB L2 vs a 1MB L2. But that's it! I'm assuming that otherwise they are pretty much the same, except for whatever is neccessary for the interchip communication. Is the MP core PL like the FX? I would think so. Any other differences???
We don't know far more about these chips than we do. Support chips???
Quite true. And I wonder why IBM do not just reveal the specifications of these chips.
Quote:
If IBM intends the FX LP to be used for mobile machines, they MUST have some reason as to why this would be possible.
If it's really just intended for embedded operations, it might not be as important, depending on what they are.
Right now we're blowing a lot of hot air over cold coals.
Equally true, but as far as the Apple notebooks are concerned, I don't believe that we will see the low power 970FX chips in them. This would require an overhaul at hardware level and with the Intel transition in sight, another major hardware shift, it just does not make sense. Besides, there is this 7448 from Freescale (I think it is already out at limited quantities) which can give the Powerbooks a reasonable update until the Intel transition starts, making at the same time room for an iBook update with the current Powerbook G4 (or close to that). It seems to me that Apple just waits for the 7448 to be available in sufficient quantities before going on and update the iBook and the Powerbook.
Well, I've said this before and I'll say it again. Don't expect Macs on Intel to magically bring on a whole new hardware market we don't already have. Your example about PCIe cards is a good example. There's nothing about apple going intel that implies or aides in getting more video card support than apple has now. Nothing. Drivers still need to be written for the OS, not the chip-set. And Apple will NOT distribute drivers for cards they don't sell bundled. You're not going to get an ATI x300 from NewEgg.com, and then get a magical driver from Apple that will let it run. ATI will need to provide those (just like they have to now).
Actually, you will be able use any PCIe graphics card available on Newegg as long as there are drivers. An Apple engineer confirmed this at WWDC. With the endian issue gone, there's nothing stopping it. Apple is also apparently going to not write drivers anymore.
ATI and NVidia will now write drivers and so they should be downloadable from the website, new updates monthly, like their Linux and Windows drivers.
Actually, you will be able use any PCIe graphics card available on Newegg as long as there are drivers. An Apple engineer confirmed this at WWDC. With the endian issue gone, there's nothing stopping it. Apple is also apparently going to not write drivers anymore.
ATI and NVidia will now write drivers and so they should be downloadable from the website, new updates monthly, like their Linux and Windows drivers.
Well that is a huge hurdle if it is true! I can't wait to get the next generation Quadro up, and running on OS X. Shit! Now I hope they have SLI as well. That would be stellar for a Mac 3d station. I doubt you would hear gamers complian about it either.
Actually, you will be able use any PCIe graphics card available on Newegg as long as there are drivers. An Apple engineer confirmed this at WWDC. With the endian issue gone, there's nothing stopping it. Apple is also apparently going to not write drivers anymore.
ATI and NVidia will now write drivers and so they should be downloadable from the website, new updates monthly, like their Linux and Windows drivers.
I agree with most of this, except for one thing. PPC 601's and every chip version after them could read either little endian or big endian. Except for one - the G5. So, if that was the problem, we should have had no problem all along.
Actually, you will be able use any PCIe graphics card available on Newegg as long as there are drivers. An Apple engineer confirmed this at WWDC. With the endian issue gone, there's nothing stopping it. Apple is also apparently going to not write drivers anymore.
ATI and NVidia will now write drivers and so they should be downloadable from the website, new updates monthly, like their Linux and Windows drivers.
Smells like Scientology-grade B.S. to me, Existence. The Stinktel switch will magically transfer the burden of OS X GPU drivers from Apple to ATI and Nvidia?
Smells like Scientology-grade B.S. to me, Existence. The Stinktel switch will magically transfer the burden of OS X GPU drivers from Apple to ATI and Nvidia?
i would say that Existence's idea is not complete rubbish, in that right now you can only purchase specific ATI- or Nvidia- fully branded cards that Apple supports.
if the Stinktel switcheroo allows for using third-party ati/nvidia cards (eg, MSI, Asus, Chaintech) then i think there would be a huge incentive for the graphics card industry, and video game manufacturers, as a whole to start catering for the Macintel crowd.
also in this case Apple not having to do drivers leaves that stuff to the "experts" themselves, ATi and nVidia, i'm sure apple has a good enough relationship with them to ensure that ATi and nVidia code for Macintel to Apple standards.
i would say that Existence's idea is not complete rubbish, in that right now you can only purchase specific ATI- or Nvidia- fully branded cards that Apple supports.
if the Stinktel switcheroo allows for using third-party ati/nvidia cards (eg, MSI, Asus, Chaintech) then i think there would be a huge incentive for the graphics card industry, and video game manufacturers, as a whole to start catering for the Macintel crowd.
also in this case Apple not having to do drivers leaves that stuff to the "experts" themselves, ATi and nVidia, i'm sure apple has a good enough relationship with them to ensure that ATi and nVidia code for Macintel to Apple standards.
We have to stop meeting like this!
Right now, Apple only does drivers for its own version of the video boards. I'm still not sure why this switch would make much of a difference. Drivers and software still have to be written for the features of each variation of a board. That wouldn't change. We would still need Mac OS drivers and software. If the market size is still too small for them to bother with, then they won't. we won't be able to use Win drivers, and the boards are the same except for the firmware, and that's no big deal.
Right now, Apple only does drivers for its own version of the video boards. I'm still not sure why this switch would make much of a difference. Drivers and software still have to be written for the features of each variation of a board. That wouldn't change. We would still need Mac OS drivers and software. If the market size is still too small for them to bother with, then they won't. we won't be able to use Win drivers, and the boards are the same except for the firmware, and that's no big deal.
Comments
Originally posted by TigerWoods99
haha.....Apple are idiots
I thought IBM had "nothing to offer"
This isn't a brilliant post. Why don't you follow the posts and learn something?
Originally posted by melgross
It was stated.
The only timeline I ever heard was part of the news.com story that broke the transition news. The timing might make perfect sense, but that isn't the same as being confirmed by Apple.
Originally posted by DCQ
Well, I can't speak for everyone, but not having end notes is a non-starter for me and my wife (academic writing). When I played around with it in my local Apple store, I liked a lot of it, but felt that it hadn't quite figured out how it was going to handle certain issues given its dtp-like model (mostly, deleting pages and linking/flowing text to boxes). I'm hopeful that Pages 2 will take care of those issues.
Also, I've read that pdf exporting isn't perfect across platforms (drop shadows and the like).
They fixed the 'Delete Page' thing in the v1.01 update although I still think it's badly done since a page could actually be 100 pages if you've just kept typing without sticking in Page Breaks. It's more of a section really. It's just badly termed/implemented but it's now there. How they missed 'Delete Page' in v1.0 though is beyond belief. The developers must have been too close to their perfect ideal and not saw the error.
Text flow works for me - In fact it's one of the things that works stupidly well in comparison to Word. I love it. Click outside the page (in the margin, Click Insert->Text and a box appears. Keep typing and a + appears on the bottom of the box. Click off the box, then on again and small blue boxes appear left/right on the box allowing you to place overflow boxes before or after. And it all happens in realtime flowing between them.
PDF problems are because Apple used the very latest PDF spec that supports drop shadows fully. Some printer drivers and for that matter some print shops don't support that spec. Even Adobe tools don't all support it. It's simply fixed. Instead of using export, use Print and 'Save As...' then select PDF-X as the format. This is a lowest common denominator PDF standard that renders the drop shadows as bitmaps.
Originally posted by DCQ
OK. I admit I don't know exactly what problems they'd face in a corporate intranet environment. (I'm imagining my old employer: 12,000 employees globally; 56 offices; 22 countries.) They may work well. But I'd guess that things like integrating iCal functionality into both Mail and iChat would be a minimal necessity (i.e., sending a Calendar invitation as an "attachment"). I don't know...that was just off the top of my head.
[/B]
You can send iCal invites already. Been there for years.
Put an event in to iCal, right click on it, click 'Mail Event'. It opens a new email with an invite bit of text, plus an iCal .ics file that the person can click on to add it to their calendar.
Again, I think this is just one of those things where because it doesn't work like Office or Outlook, people don't find these things, or don't try.
How about AddressBook and iCal - did you know you can drag contacts from address book to an event's attendees list or just dropping a group onto a spare slot in iCal automatically sets up a meeting with all those in the group?
It's not perfect though I'll give you that. It'd be nice to share calendars without having to set up an internal webdav server or use .Mac and it's not as simple as Outlook to publish/subscribe calendars but given some reasonable training or semi-intelligent users it's perfectly practical IME.
Only problem is, like Outlook, everyone has to use the same calendar application so that kind of means everyone has to use a Mac in the office.
Originally posted by Programmer
I've seen this comment a couple of times, but I don't believe it to be true. The OSX thread implementation is based on the Mach threaded kernel and so ought to be very efficient. The pthreads are a thin API wrapper on that, and the Cocoa & Carbon threads are a API wrapper on pthreads.
[/B]
The following predates Tiger but comparing Linux v OSX on the same G5...
http://www-128.ibm.com/developerwork...r-mw11LinxOnG5
Look at the LMBench scores, some of which are just shocking.
Then there was also the recent Anandtech MySQL and Apache shootout which had similar problems although I hate to quote that as proof as it was fundamentally flawed.
A lot of it was concluded to be the file system and Dominic Giampolo said it was because HFS+ was 'right' whereas EXT3 had greater potential to lose data but having run webservers for a number of years, seeing performance not just a few percent better but a few 100 percent better, I'll stick with Linux thanks there.
In Tiger, I believe they got rid of the kernel funnel and network funnel and made the kernel lock at a finer level.
Originally posted by aegisdesign
A lot of it was concluded to be the file system and Dominic Giampolo said it was because HFS+ was 'right' whereas EXT3 had greater potential to lose data but having run webservers for a number of years, seeing performance not just a few percent better but a few 100 percent better, I'll stick with Linux thanks there.
For those purposes there are faster file systems around. Hell, Dominic knows all too well just how frickin fast and nice BFS was, but that is long gone, alas. HFS+ may get there, but I think some changes are underway. I read an article on ARS about Spotlight and how it uses an external database for the purpose of storing the metadata, rather than the file system itself. Not that that is any reason to draw any conclusions, but it seems sorta stop gap to me. I guess time will tell.
Originally posted by aegisdesign
The following predates Tiger but comparing Linux v OSX on the same G5...
http://www-128.ibm.com/developerwork...r-mw11LinxOnG5
Look at the LMBench scores, some of which are just shocking.
When you see results like this there is typically "something strange going on" that is being exposed by the particular test. Even the author acknowledges this:
However, a few stunningly bad results on OSX make me wonder if all parts of the suite were running accurately.
He also goes on to state that he's not a benchmarking expert and doesn't properly understand what he's doing. So I'd be inclined to ignore the 500x difference on the one test because my Macs sure as heck aren't 500x slower than my PCs or Linux boxes. The 1-2x differences are certainly believeable, but every release of OSX has been getting faster and that is because Apple is addressing these kinds of issues over time.
Of all those LMBench tests, however, none of them really say much about MacOSX "threading" per-se. Most reflect on the file and memory systems. The context switch time for the smallest unit of work is about the closest to a measurement of threading, and it is about 10-20% slower. The larger amounts of work start reflecting on the VM setup of MacOS X. It would be interesting to see these tests re-run to compare 10.3.4 vs. 10.4.1 to see how much progress Apple is making on the low-level issues.
None of these really says much about how effectively MacOS X could take advantage of more cores compared to other OSes. These test are measure context switching times on a single processor... more cores means fewer context switches which means that MacOS X might actually benefit more from more cores than an OS that is more optimized.
Then there was also the recent Anandtech MySQL and Apache shootout which had similar problems although I hate to quote that as proof as it was fundamentally flawed.
A lot of it was concluded to be the file system and Dominic Giampolo said it was because HFS+ was 'right' whereas EXT3 had greater potential to lose data but having run webservers for a number of years, seeing performance not just a few percent better but a few 100 percent better, I'll stick with Linux thanks there.
Yes, the flush in OSX actually works. That's like writing a sort function that is 50 times faster but doesn't generate a sorted array as output. And, FWIW, Dominic said you could disable the flush and performance would increase dramatically. The real bugs are in Linux and the server software -- (a) for having a non-functional flush and (b) for calling flush inappropriately. Apple is working to fix the server software so that they don't have to break the OS to make it performant.
In Tiger, I believe they got rid of the kernel funnel and network funnel and made the kernel lock at a finer level.
Yes, plus they have put more internal APIs in place which will allow them to make deeper changes in how the OS works in the future.
Originally posted by BRussell
How efficient would a dual dual-core be? I'm guessing that most users get little benefit right now from duals, let alone quads. I'm sure there are some who use apps that take advantage of them, or use multiple apps simultaneously. But quads?
Do not look at it this way. Mac OS X handles perfectly multiple processors and will distribute the load accordingly. This means that you can fully utilise the four cores by running, for example, three or four independent CPU-intensive applications. Each one of them will take one processor to go.
Now telling about the efficiency is not that simple. There has been some speculation (I don't remember if it was based on info from the leaked IBM document on the 970MP) that the 970MP chip will sit on the G5 bus we have today and that there will be a special "bridge" to allow a very fast inter-chip communication between the two cores. Or something like that. So, in a quad system, we would have two chips with two independent FSB as of today, but each chip will have two processing cores, each capable to speak to the other using the special bridge without needing the FSB. What the efficiency of such a design is, remains to be seen.
A note of interest about quads coming. Apple boldly emphasizes the fine grain locking feature in Tiger (the first in the "New in Tiger" section). Now read this about funnels and Mac OS X (Programmer already made a remark on that). And then remember the 4 CPU reference in the CHUD tools that leaked some time ago. Oh, add to that the 970MP announcement. Oh, and the fact that the Power Mac need to be updated at least once in the coming two years with PowerPC processors, before the Intel switch. Does it start to make sense now
Originally posted by PB
Do not look at it this way. Mac OS X handles perfectly multiple processors and will distribute the load accordingly. This means that you can fully utilise the four cores by running, for example, three or four independent CPU-intensive applications. Each one of them will take one processor to go.
Now telling about the efficiency is not that simple. There has been some speculation (I don't remember if it was based on info from the leaked IBM document on the 970MP) that the 970MP chip will sit on the G5 bus we have today and that there will be a special "bridge" to allow a very fast inter-chip communication between the two cores. Or something like that. So, in a quad system, we would have two chips with two independent FSB as of today, but each chip will have two processing cores, each capable to speak to the other using the special bridge without needing the FSB. What the efficiency of such a design is, remains to be seen.
A note of interest about quads coming. Apple boldly emphasizes the fine grain locking feature in Tiger (the first in the "New in Tiger" section). Now read this about funnels and Mac OS X (Programmer already made a remark on that). And then remember the 4 CPU reference in the CHUD tools that leaked some time ago. Oh, add to that the 970MP announcement. Oh, and the fact that the Power Mac need to be updated at least once in the coming two years with PowerPC processors, before the Intel switch. Does it start to make sense now
I would also imagine that the new 1MB per core L2 is also because both cores will ride on the same FSB, unlike the older single core cpu's. This would lessen the dependence on the bus throughput and contention that would result from the two cores sharing that bus.
Of course, if what I remember is correct, the 970 can also use a full 1:1 speed FSB as well. It would certainly be interesting to see if Apple considered going to that, or whether it would simply be to expensive with the changes in the support chips that would be required. That would sure shake things up!
Originally posted by melgross
I would also imagine that the new 1MB per core L2 is also because both cores will ride on the same FSB, unlike the older single core cpu's. This would lessen the dependence on the bus throughput and contention that would result from the two cores sharing that bus.
I don't know, more L2 helps in many things. Maybe inter-chip communication is one of them.
Of course, if what I remember is correct, the 970 can also use a full 1:1 speed FSB as well. It would certainly be interesting to see if Apple considered going to that, or whether it would simply be to expensive with the changes in the support chips that would be required. That would sure shake things up!
Yeah, you are right, I forgot about that. I doubt it though, because of power dissipation (with everything this implies) and cost considerations.
Originally posted by PB
I don't know, more L2 helps in many things. Maybe inter-chip communication is one of them.
Yeah, you are right, I forgot about that. I doubt it though, because of power dissipation (with everything this implies) and cost considerations.
Something else that we don't know at this point is what similarities and differences exist between the cores in the MP and the FX low power(LP).
We know that the LP has a 512KB L2 vs a 1MB L2. But that's it! I'm assuming that otherwise they are pretty much the same, except for whatever is neccessary for the interchip communication. Is the MP core PL like the FX? I would think so. Any other differences???
We don't know far more about these chips than we do. Support chips???
If IBM intends the FX LP to be used for mobile machines, they MUST have some reason as to why this would be possible.
If it's really just intended for embedded operations, it might not be as important, depending on what they are.
Right now we're blowing a lot of hot air over cold coals.
Originally posted by melgross
We know that the LP has a 512KB L2 vs a 1MB L2. But that's it! I'm assuming that otherwise they are pretty much the same, except for whatever is neccessary for the interchip communication. Is the MP core PL like the FX? I would think so. Any other differences???
We don't know far more about these chips than we do. Support chips???
Quite true. And I wonder why IBM do not just reveal the specifications of these chips.
If IBM intends the FX LP to be used for mobile machines, they MUST have some reason as to why this would be possible.
If it's really just intended for embedded operations, it might not be as important, depending on what they are.
Right now we're blowing a lot of hot air over cold coals.
Equally true, but as far as the Apple notebooks are concerned, I don't believe that we will see the low power 970FX chips in them. This would require an overhaul at hardware level and with the Intel transition in sight, another major hardware shift, it just does not make sense. Besides, there is this 7448 from Freescale (I think it is already out at limited quantities) which can give the Powerbooks a reasonable update until the Intel transition starts, making at the same time room for an iBook update with the current Powerbook G4 (or close to that). It seems to me that Apple just waits for the 7448 to be available in sufficient quantities before going on and update the iBook and the Powerbook.
Originally posted by Louzer
Well, I've said this before and I'll say it again. Don't expect Macs on Intel to magically bring on a whole new hardware market we don't already have. Your example about PCIe cards is a good example. There's nothing about apple going intel that implies or aides in getting more video card support than apple has now. Nothing. Drivers still need to be written for the OS, not the chip-set. And Apple will NOT distribute drivers for cards they don't sell bundled. You're not going to get an ATI x300 from NewEgg.com, and then get a magical driver from Apple that will let it run. ATI will need to provide those (just like they have to now).
Actually, you will be able use any PCIe graphics card available on Newegg as long as there are drivers. An Apple engineer confirmed this at WWDC. With the endian issue gone, there's nothing stopping it. Apple is also apparently going to not write drivers anymore.
ATI and NVidia will now write drivers and so they should be downloadable from the website, new updates monthly, like their Linux and Windows drivers.
Originally posted by Existence
Actually, you will be able use any PCIe graphics card available on Newegg as long as there are drivers. An Apple engineer confirmed this at WWDC. With the endian issue gone, there's nothing stopping it. Apple is also apparently going to not write drivers anymore.
ATI and NVidia will now write drivers and so they should be downloadable from the website, new updates monthly, like their Linux and Windows drivers.
Well that is a huge hurdle if it is true! I can't wait to get the next generation Quadro up, and running on OS X. Shit! Now I hope they have SLI as well. That would be stellar for a Mac 3d station. I doubt you would hear gamers complian about it either.
Originally posted by Existence
Actually, you will be able use any PCIe graphics card available on Newegg as long as there are drivers. An Apple engineer confirmed this at WWDC. With the endian issue gone, there's nothing stopping it. Apple is also apparently going to not write drivers anymore.
ATI and NVidia will now write drivers and so they should be downloadable from the website, new updates monthly, like their Linux and Windows drivers.
I agree with most of this, except for one thing. PPC 601's and every chip version after them could read either little endian or big endian. Except for one - the G5. So, if that was the problem, we should have had no problem all along.
Originally posted by Existence
Actually, you will be able use any PCIe graphics card available on Newegg as long as there are drivers. An Apple engineer confirmed this at WWDC. With the endian issue gone, there's nothing stopping it. Apple is also apparently going to not write drivers anymore.
ATI and NVidia will now write drivers and so they should be downloadable from the website, new updates monthly, like their Linux and Windows drivers.
Smells like Scientology-grade B.S. to me, Existence. The Stinktel switch will magically transfer the burden of OS X GPU drivers from Apple to ATI and Nvidia?
Originally posted by Big Mac
Smells like Scientology-grade B.S. to me, Existence. The Stinktel switch will magically transfer the burden of OS X GPU drivers from Apple to ATI and Nvidia?
i would say that Existence's idea is not complete rubbish, in that right now you can only purchase specific ATI- or Nvidia- fully branded cards that Apple supports.
if the Stinktel
also in this case Apple not having to do drivers leaves that stuff to the "experts" themselves, ATi and nVidia, i'm sure apple has a good enough relationship with them to ensure that ATi and nVidia code for Macintel to Apple standards.
Originally posted by sunilraman
i would say that Existence's idea is not complete rubbish, in that right now you can only purchase specific ATI- or Nvidia- fully branded cards that Apple supports.
if the Stinktel
also in this case Apple not having to do drivers leaves that stuff to the "experts" themselves, ATi and nVidia, i'm sure apple has a good enough relationship with them to ensure that ATi and nVidia code for Macintel to Apple standards.
We have to stop meeting like this!
Right now, Apple only does drivers for its own version of the video boards. I'm still not sure why this switch would make much of a difference. Drivers and software still have to be written for the features of each variation of a board. That wouldn't change. We would still need Mac OS drivers and software. If the market size is still too small for them to bother with, then they won't. we won't be able to use Win drivers, and the boards are the same except for the firmware, and that's no big deal.
Originally posted by melgross
We have to stop meeting like this!
Right now, Apple only does drivers for its own version of the video boards. I'm still not sure why this switch would make much of a difference. Drivers and software still have to be written for the features of each variation of a board. That wouldn't change. We would still need Mac OS drivers and software. If the market size is still too small for them to bother with, then they won't. we won't be able to use Win drivers, and the boards are the same except for the firmware, and that's no big deal.
ok cool..
NOW TURN OFF THE COMPUTER AND GO TO SLEEP !! heh
Originally posted by sunilraman
ok cool..
NOW TURN OFF THE COMPUTER AND GO TO SLEEP !! heh
Hey, didn't you say that already?
(It's only 11:00 P.M. here. Wait 'till it gets late.)