I intend to read this article but for now I have to wait until the abstract no longer annoys me as much as it does now. What difference does it make whether anyone labels this upgrade a "service pack"? I doubt if there is any universally recognized, objective definition for this term, and as such, there is no way that anyone could objectively say whether it is or is not a "service pack". It's utterly meaningless, and I don't understand why the author or anyone else would take exception to something that has no meaning. In the abstract, the author is clearly trying to evoke an emotional response. Such shenanigans are typical in many web forums, but are most often associated with the sites where teenagers hang out and taunt each other.
The atmosphere here makes me feel like I am associating with some sort of religious cult. This site is often informative, and occasionally I see wording in an article that is marginally critical of Apple. But there is some sort of unwritten law that forbids unmitigated criticism of Apple when it is deserved. This release is not without warts, and anyone who dares to mention any of them on this site gets jumped on by a bunch of those people who behave as though the primary source of their sense of self-worth comes by way of their affinity to Apple. Its the same syndrome as with car branding.
It seems to me that the public interest would be far better served if sites such as this provided a forum for people to register and discuss the problems that they are having with a new product such as this release. There is a little of that here, but it is implicitly discouraged, and there are the apologists who jump all over anyone who dares to criticize Apple even when the reason is undeniable.
The video player was screwed up due to it not being possible to position the control panel anywhere except on top of the video, and due to the slider being much too short to be fully functional with respect to its manifest purpose. But the thing that annoys me is that rather than see an improvement in the way that text is rendered graphically on the screen, this release exhibits a step backwards in that department, owing to the loss of the ability to disable the text dithering, a.k.a. "font smudging". There is obviously an error in the implementation, as evidenced by what you see in the System Preferences>Appearance screen. Previously, this screen did not imply that you have the ability to disable dithering entirely, but you were able to disable dithering for font sizes 12 and smaller. As of this release, the design of this screen now implies that you have the ability to disable dithering unconditionally. This would be fantastic if it worked as advertised, but the problem is that aside from limited effect in some applications, the capability to disable dithering no longer works at all, not even for font sizes 12 and smaller.
This sort of obvious, in-your-face error suggests to me an army of undisciplined, junior programmers working without adequate supervision. I would find it much more tolerable if there were some way to formally log an error. But Apple does not see fit to allow the at-large user community to formally log errors. The reason, presumably, is that it would be a huge task to sort through and figure out what is real. But when there is a real, legitimate bug, there isn't any way for an ordinary user to convey the existence of the bug to Apple and get any sort of formal acknowledgement that they are aware of the problem and have someone looking into it. I can't stand font smudging, and there just isn't a reason in the world why anyone should want to force this on me. From the day that Apple introduced this, whenever it was exactly, they have failed to consider the preferences of people who just don't like it. The mentality that is evidenced here is the same as what we see with many of their product decisions, which seem to come from a very high level within company. Someone decides what is best for the world, and if you do not agree, then you are just out of luck. I abhor this sort of control-freak mentality more than I could possibly convey. I have seem programmers stuck in this mentality more times than I could count. The logic required to prevent the user from disabling font smudging for fonts sizes larger than 12 is inherently more complex than the logic sans that restriction.
Takes 4 seconds for any page except yahoo to load up
whats up with that?
There must be issues with your particular machine or its connection to the web, as Safari is super fast on my MBP. With SL my webpages seem to load almost instantaneously. Are you accessing the web on a low-bandwidth connection?
I had already looked there, but could not find that phrase: 64-bit Kernel and Extensions
There are references to 64 bit all over the place, but I'm concerned that they are not referring to which kernel is being used.
Is there are more definitive test for this?
After you click 'About This Mac' a System Profiler window will appear. On the left side of that window, in the 'Content' column, click directly on 'Software'. In the column to the right you will see 'System Software Overview'. There you will see '64-bit Kernel and Extensions:'. If 'Yes', then your Mac OS X is running in it's 64-bit kernel. This line is definitive.
I forgot to add 'click "More Info…' to get to expanded System Profiler window
I can't provide an easy-to-understand answer. I too have an 2007 24" iMac, with an Intel Core 2 Duo CPU, and have upgraded to 4GB of memory. Yes, the CPU is 64-bit. The EFI is 64-bit. Nevertheless, my iMac is model 7,1. Apple decided that only models 8,1 and 9,1 are enabled to run K64 (by booting with the '6' and '4' key pressed down). I've tried it – the System Profile says K32. If I go to the Activity Monitor, I can see that most of my bundled apps like Finder, Mail, Safari and more are running as 64-bit binaries. Yes, I do benefit the use of these 64-bit app binaries instead of the 32-bit binaries. My Samsung laser printer is 32-bit and the 'Smart Panel' software supporting that printer runs as a 32-bit binary.
I rarely use more than 1 GB of memory, even with muliple apps open. I do benefit from SL; I experience better responsiveness across the OS and it's bundled apps. Kudos to Grand Central Dispatch! So, I hope to postpone buying a new iMac until 2012 (five years past 2007).
As a previous post mentioned, Apple limited the booting of K64 to certain machines to avoid compatibility problems. It make good business sense. It is possible that a future point release of SL will add drivers to support the earlier machines (before 2008) with Intel Core 2 Duo CPUs. I'm not privy to Apple's product plans.
It's possible that Apple compiled the 64-bit kernel with SSE4.1. That would explain why only Macs with Penryn or newer processors support the 64-bit kernel making 64-bit EFI not the only requirement. I'm just speculating though. Unless someone knows how to check if software is compiled to require a certain SSE set. I believe normally OS X software is compiled against SSE2, while parts of OS X have required SSE3 like Rosetta.
EDIT: For reference 65nm Yonah Core Duos have SSE3, 65nm Merom Core 2 Duos have SSSE3, 45nm Penryn Core 2 Duos have SSE4.1, and 45nm Nehalem has SSE4.2. I believe the minimum SSE requirement was set at SSE2 because the AppleTV uses an older Dothan Pentium M which only has SSE2. All 64-bit Intel processors used in Macs have at least SSSE3 so that really should have been the minimum compile target for 64-bit apps.
It's possible that Apple compiled the 64-bit kernel with SSE4.1. That would explain why only Macs with Penryn or newer processors support the 64-bit kernel making 64-bit EFI not the only requirement. I'm just speculating though. Unless someone knows how to check if software is compiled to require a certain SSE set. I believe normally OS X software is compiled against SSE2, while parts of OS X have required SSE3 like Rosetta.
EDIT: For reference 65nm Yonah Core Duos have SSE3, 65nm Merom Core 2 Duos have SSSE3, 45nm Penryn Core 2 Duos have SSE4.1, and 45nm Nehalem has SSE4.2. I believe the minimum SSE requirement was set at SSE2 because the AppleTV uses an older Dothan Pentium M which only has SSE2. All 64-bit Intel processors used in Macs have at least SSSE3 so that really should have been the minimum compile target for 64-bit apps.
Well, I've got the Merom. I'll survive. By the way, I saw a comment that the ATI Radeon HD 2600 does not support double-precision floating point, and thus cannot support OpenCL. I don't know enough to confirm that comment.
After you click 'About This Mac' a System Profiler window will appear. On the left side of that window, in the 'Content' column, click directly on 'Software'. In the column to the right you will see 'System Software Overview'. There you will see '64-bit Kernel and Extensions:'. If 'Yes', then your Mac OS X is running in it's 64-bit kernel. This line is definitive.
I forgot to add 'click "More Info?' to get to expanded System Profiler window
Many thanks ? found it. I was looking everywhere BUT in the overview section (in extensions and frameworks, etc).
Well, I've got the Merom. I'll survive. By the way, I saw a comment that the ATI Radeon HD 2600 does not support double-precision floating point, and thus cannot support OpenCL. I don't know enough to confirm that comment.
Double precision floating point should have nothing to do with supporting OpenCL. In fact, the nVidia 8000 and 9000 series don't support DP floats either. Only the ATI HD3000, HD4000, and nVidia GTX200 series do. I'm fairly certain that DP floats are not part of OpenCL 1.0 spec anyways and will come in subsequent versions most likely because of the small installed base right now.
I have asked AMD's Stream Computing team and they've told me that OpenCL support won't be coming to the HD2000 and HD3000 series because of a hardware limitation. They didn't confirm what it was, but I'm thinking it has to do with their memory hierarchy. nVidia GPUs have a local memory store in their Streaming Multiprocessors (SMs) shared between 8 Stream Processors (SPs). The HD2000 and HD3000 series seem to have their shared memory store at a higher level, between all SPs. With OpenCL being closer to CUDA, it wouldn't surprise me if lack of local memory store is a problem. The Folding@home team has identified lack of local memory store in the HD2000 and HD3000 series as a major performance impediment, even though Folding@home is using ATI's native Brook+ language for the ATI client. ATI's HD4000 series adopts a memory model closer to nVidia's with local memory stores so that could certainly be a differentiating factor in OpenCL compatibility. It's not like I'm an expert though. It's just my guess from what I remember from GPU architecture reviews.
Double precision floating point should have nothing to do with supporting OpenCL. In fact, the nVidia 8000 and 9000 series don't support DP floats either. Only the ATI HD3000, HD4000, and nVidia GTX200 series do. I'm fairly certain that DP floats are not part of OpenCL 1.0 spec anyways and will come in subsequent versions most likely because of the small installed base right now.
I have asked AMD's Stream Computing team and they've told me that OpenCL support won't be coming to the HD2000 and HD3000 series because of a hardware limitation. They didn't confirm what it was, but I'm thinking it has to do with their memory hierarchy. nVidia GPUs have a local memory store in their Streaming Multiprocessors (SMs) shared between 8 Stream Processors (SPs). The HD2000 and HD3000 series seem to have their shared memory store at a higher level, between all SPs. With OpenCL being closer to CUDA, it wouldn't surprise me if lack of local memory store is a problem. The Folding@home team has identified lack of local memory store in the HD2000 and HD3000 series as a major performance impediment, even though Folding@home is using ATI's native Brook+ language for the ATI client. ATI's HD4000 series adopts a memory model closer to nVidia's with local memory stores so that could certainly be a differentiating factor in OpenCL compatibility. It's not like I'm an expert though. It's just my guess from what I remember from GPU architecture reviews.
I don't have the knowledge to discuss what you said, except I'm aware that OpenCL is derived from CUDA. John Siracusa at Ars Technica, in his review of SL, had some good
info in his review regarding OpenCL. Check it out if you have not read it.
I don't have the knowledge to discuss what you said, except I'm aware that OpenCL is derived from CUDA. John Siracusa at Ars Technica, in his review of SL, had some good
info in his review regarding OpenCL. Check it out if you have not read it.
I actually asked John whether he thought memory hierarchy might be the problem with HD2000 and HD3000 support. He thought it was possible although he also speculated that a good driver implementation might also be able to work around it. Even if that were the case, seeing the already less than spectacular performance of the HD4000 series, I doubt it'd be worthwhile to support older ATI GPUs since emulating the memory model would make the end performance not worthwhile.
I actually asked John whether he thought memory hierarchy might be the problem with HD2000 and HD3000 support. He thought it was possible although he also speculated that a good driver implementation might also be able to work around it. Even if that were the case, seeing the already less than spectacular performance of the HD4000 series, I doubt it'd be worthwhile to support older ATI GPUs since emulating the memory model would make the end performance not worthwhile.
If I remember correctly, the last G5 powermacs were sold alongside the first core duo Intel Macs. The g5s just got cut off. Would it really surprise you if 10.7 didn't support 32 bit machines.
My core 2 duo MBP has 32 bit EFI firmware and I'm not holding my breath that 10.7 will support my machine.
If we assume 18 months for the nexzt overhaul it?ll be over 4 years since Macs went both 64-bit CPUs -AND- 64-bit chipsets (remember the first C2Ds only had 32-bit RAM addressing), so that seems likely. The real question is whether enough 3rd-party drivers have gone 64-bit -AND- there isn?t enough old HW out there that won?t get a 64-bit driver but is too costly to replace or just inconvenient for the user. I?d say that 64-bit will at likely be the default for 10.7, but I?m not sold yet on the 32-bit kernel being nixed when it?s not hard to keep, when I assume that 10.7 will mainly be a UI overhaul more than all these core changes we?ve seen in 10.6.
Quote:
Originally Posted by franksargent
As to my current laptop 8GB, 2.5GHz Core 2 Duo, 500GB 7200RPM HD, Vista 64-bit Ultimate SP2, total cost ~$1,800. Price a 15" MacBook Pro with these specs for me if you don't mind.
How can price the same HW when you didn?t give any decent specifics that affect the price. You?ve only posted the surface specs that people with basic technical understanding of computers look at. Don?t get me wrong, that is fine. I can find a 17? notebook for half the price of a 13? MacBook, but that doesn?t mean it?s better simply because the display is bigger and with higher resolution.
Furthermore, you can find Dells that has the same basic specs as a 15? MBP but cost anywhere from a fraction of the price to about the same price. Is Dell screwing customers with those more expensive machines or do those machines with higher L2, lower TDP, and extra thin chassises that are also durable also cost money to make? I think you are missing the point about what is good enough for your needs and what is an actual bona fide comparable machine.
A final thought. Vista x64 is great. Windows 7 x64 hasn?t really shown to speed much up, only that they have thinned out the overly hostile processes so it performs better on slower/older HW and that without these processes there is more RAM and more cycles to be used more efficiently. Since SP1 Vista has many of the initial kinks worked out, but I ask you, what would you do if you were in a situation where you had to use a printer, say at a hotel, and the printer driver was only 32-bit? You can?t restart into 32-bit mood and reinstalling with the 32-bit kernel means no more than 4GB of memory addressing and no 64-bit apps. Apple?s stepped solution for 64-bit is ideal for the average consumer.
I was very likely doing 64-bit programming on a Cray supercomputer before you were even born.
25 years ago, to be exact.
As to my current laptop 8GB, 2.5GHz Core 2 Duo, 500GB 7200RPM HD, Vista 64-bit Ultimate SP2, total cost ~$1,800. Price a 15" MacBook Pro with these specs for me if you don't mind.
Well it's hard to do an informative comparison without full specs for your laptop... in June I did a price comparison on a dual quad core Mac Pro and a Dell T7500:
Dell T7500: $6062
Dual Quad Core 2.66 GHz Xeon X5550
4 GB 1066 MHz memory (they didn't have an option for 6GB 1006 MHz mem)
NVIDIA Quadro FX 580, 512 MB VRAM ($175 retail)
1 TB Hard Drive, SATA 3Gb/s, 7200 RPM, 16 MB cache
The Dell is more scalable (1 more drive bay, 1 more PCI express slot, and 4 more memory slots) and it comes with a support plan (I think Apple Care is ~$300) but I'll go with the superior Apple build quality and save some $$$. By chance, I checked again earlier today, and the price difference is still ~$1k (although this time Dell has a 6GB option)... also checked an HP box (can't remember which) with Linux installed and the price was closer, but Apple was still cheaper... I think by ~$200-$300.
I do have to agree with other posters that the 64 bit transition is inevitable, and not conceptually any different than the 16-32 bit transition... I actually do find 64 bit useful for my line of work... processing of very large images.
And I did my first 60 bit coding on a CDC 7600 in 1980, so I'm no spring chicken...
Well it's hard to do an informative comparison without full specs for your laptop... in June I did a price comparison on a dual core Mac Pro and a Dell T7500:
Dell T7500: $6062
Dual Quad Core 2.66 GHz Xeon X5550
4 GB 1066 MHz memory (they didn't have an option for 6GB 1006 MHz mem)
NVIDIA Quadro FX 580, 512 MB VRAM ($175 retail)
1 TB Hard Drive, SATA 3Gb/s, 7200 RPM, 16 MB cache
The Dell is more scalable (1 more drive bay, 1 more PCI express slot, and 4 more memory slots) and it comes with a support plan (I think Apple Care is ~$300) but I'll go with the superior Apple build quality and save some $$$. By chance, I checked again earlier today, and the price difference is still ~$1k (although this time Dell has a 6GB option)... also checked an HP box (can't remember which) with Linux installed and the price was closer, but Apple was still cheaper... I think by ~$200-$300.
I do have to agree with other posters that the 64 bit transition is inevitable, and not conceptually any different than the 16-32 bit transition... I actually do find 64 bit useful for my line of work... processing of very large images.
And I did my first 60 bit coding on a CDC 7600 in 1980, so I'm no spring chicken...
I also checked Alienware for a 15" laptop today and the specs/price are as follows:
Core2duo 2.8Ghz 800 Mhz bus (6 MB cache)
Mobility Radeon 3870
4 GB Ram (667 Mhz DDR2)
500 GB 7200 RPM HD
1920*1200 Resolution (lacking on Apple's 15" line atm)
Dual Layer 8x DVD burner
2299$
Apple 15" Mbp
Core2duo 2.8Ghz 1066 Mhz bus (3 MB cache)
9600GT M + 9400M
4 GB Ram (1066 Mhz DDR3)
500 GB 7200 RPM HD
1440*900 Resolution
Superdrive
2349$
The GPU's are almost similar in performance but Alienware has higher res screen (although I can't imagine using a 1080p screen at 15", even on my 17" Mbp, things are hard to read)
Mac has faster bus + RAM, Alienware has 3 MB more cache
+ On top of all that, the design is not even comparable. Mac is much much much thinner and looks way better. Alienware has one of the worst laptop designs I saw on PC's. Sony VAIO's look much nicer.
Edit: Alienware is more configurable in terms of GPU, if you pay 350$ more you get a 9800GT M, but Apple is more configurable in terms of CPU, you can get a 3.06 GHz.
I made an interesting discovery today. I had booted into the 64 bit kernel, but nevertheless found that I could print to an older HP 4350 that was on a network. Good news, as I had thought from the reviews that it was not going to be possible.
I also checked Alienware for a 15" laptop today and the specs/price are as follows:
Core2duo 2.8Ghz 800 Mhz bus (6 MB cache)
Mobility Radeon 3870
4 GB Ram (667 Mhz DDR2)
500 GB 7200 RPM HD
1920*1200 Resolution (lacking on Apple's 15" line atm)
Dual Layer 8x DVD burner
2299$
Apple 15" Mbp
Core2duo 2.8Ghz 1066 Mhz bus (3 MB cache)
9600GT M + 9400M
4 GB Ram (1066 Mhz DDR3)
500 GB 7200 RPM HD
1440*900 Resolution
Superdrive
2349$
The GPU's are almost similar in performance but Alienware has higher res screen (although I can't imagine using a 1080p screen at 15", even on my 17" Mbp, things are hard to read)
Mac has faster bus + RAM, Alienware has 3 MB more cache
+ On top of all that, the design is not even comparable. Mac is much much much thinner and looks way better. Alienware has one of the worst laptop designs I saw on PC's. Sony VAIO's look much nicer.
Edit: Alienware is more configurable in terms of GPU, if you pay 350$ more you get a 9800GT M, but Apple is more configurable in terms of CPU, you can get a 3.06 GHz.
If we assume 18 months for the nexzt overhaul it’ll be over 4 years since Macs went both 64-bit CPUs -AND- 64-bit chipsets (remember the first C2Ds only had 32-bit RAM addressing), so that seems likely. The real question is whether enough 3rd-party drivers have gone 64-bit -AND- there isn’t enough old HW out there that won’t get a 64-bit driver but is too costly to replace or just inconvenient for the user. I’d say that 64-bit will at likely be the default for 10.7, but I’m not sold yet on the 32-bit kernel being nixed when it’s not hard to keep, when I assume that 10.7 will mainly be a UI overhaul more than all these core changes we’ve seen in 10.6.
I also don't think the 32-bit kernel is going to disappear anytime soon. There will still be quite a few niche devices that will be slow to release 64-bit drivers, perhaps mixers or input devices in the audio/visual realm. It's highly likely though that 64-bit Intel processors will be the requirement in 10.7 and the 64-bit kernel made the default, while the 32-bit kernel is still available just like it is on the XServes now. The 32-bit kernel may disappear in 10.8, but that is still a long ways off. Even when the 32-bit kernel disappears, 32-bit frameworks to run 32-bit apps in a 64-bit OS will remain since some apps really don't need to move to 64-bit anyways. Hopefully Rosetta will remain an optional install for a while otherwise what would we do without older classics like Starcraft?
Comments
The atmosphere here makes me feel like I am associating with some sort of religious cult. This site is often informative, and occasionally I see wording in an article that is marginally critical of Apple. But there is some sort of unwritten law that forbids unmitigated criticism of Apple when it is deserved. This release is not without warts, and anyone who dares to mention any of them on this site gets jumped on by a bunch of those people who behave as though the primary source of their sense of self-worth comes by way of their affinity to Apple. Its the same syndrome as with car branding.
It seems to me that the public interest would be far better served if sites such as this provided a forum for people to register and discuss the problems that they are having with a new product such as this release. There is a little of that here, but it is implicitly discouraged, and there are the apologists who jump all over anyone who dares to criticize Apple even when the reason is undeniable.
The video player was screwed up due to it not being possible to position the control panel anywhere except on top of the video, and due to the slider being much too short to be fully functional with respect to its manifest purpose. But the thing that annoys me is that rather than see an improvement in the way that text is rendered graphically on the screen, this release exhibits a step backwards in that department, owing to the loss of the ability to disable the text dithering, a.k.a. "font smudging". There is obviously an error in the implementation, as evidenced by what you see in the System Preferences>Appearance screen. Previously, this screen did not imply that you have the ability to disable dithering entirely, but you were able to disable dithering for font sizes 12 and smaller. As of this release, the design of this screen now implies that you have the ability to disable dithering unconditionally. This would be fantastic if it worked as advertised, but the problem is that aside from limited effect in some applications, the capability to disable dithering no longer works at all, not even for font sizes 12 and smaller.
This sort of obvious, in-your-face error suggests to me an army of undisciplined, junior programmers working without adequate supervision. I would find it much more tolerable if there were some way to formally log an error. But Apple does not see fit to allow the at-large user community to formally log errors. The reason, presumably, is that it would be a huge task to sort through and figure out what is real. But when there is a real, legitimate bug, there isn't any way for an ordinary user to convey the existence of the bug to Apple and get any sort of formal acknowledgement that they are aware of the problem and have someone looking into it. I can't stand font smudging, and there just isn't a reason in the world why anyone should want to force this on me. From the day that Apple introduced this, whenever it was exactly, they have failed to consider the preferences of people who just don't like it. The mentality that is evidenced here is the same as what we see with many of their product decisions, which seem to come from a very high level within company. Someone decides what is best for the world, and if you do not agree, then you are just out of luck. I abhor this sort of control-freak mentality more than I could possibly convey. I have seem programmers stuck in this mentality more times than I could count. The logic required to prevent the user from disabling font smudging for fonts sizes larger than 12 is inherently more complex than the logic sans that restriction.
having used the "holding down the 6 and 4 keys" trick, can someone tell me how I confirm that I have properly booted into the 64 kernel, please?
> About This Mac > More Info? > Software, and look for '64-bit Kernel and Extensions:'.
64 bit???? Please
Safari is super slow
Takes 4 seconds for any page except yahoo to load up
whats up with that?
There must be issues with your particular machine or its connection to the web, as Safari is super fast on my MBP. With SL my webpages seem to load almost instantaneously. Are you accessing the web on a low-bandwidth connection?
> About This Mac > More Info? > Software, and look for '64-bit Kernel and Extensions:'.
I had already looked there, but could not find that phrase: 64-bit Kernel and Extensions
There are references to 64 bit all over the place, but I'm concerned that they are not referring to which kernel is being used.
Is there are more definitive test for this?
I had already looked there, but could not find that phrase: 64-bit Kernel and Extensions
There are references to 64 bit all over the place, but I'm concerned that they are not referring to which kernel is being used.
Is there are more definitive test for this?
After you click 'About This Mac' a System Profiler window will appear. On the left side of that window, in the 'Content' column, click directly on 'Software'. In the column to the right you will see 'System Software Overview'. There you will see '64-bit Kernel and Extensions:'. If 'Yes', then your Mac OS X is running in it's 64-bit kernel. This line is definitive.
I forgot to add 'click "More Info…' to get to expanded System Profiler window
I can't provide an easy-to-understand answer. I too have an 2007 24" iMac, with an Intel Core 2 Duo CPU, and have upgraded to 4GB of memory. Yes, the CPU is 64-bit. The EFI is 64-bit. Nevertheless, my iMac is model 7,1. Apple decided that only models 8,1 and 9,1 are enabled to run K64 (by booting with the '6' and '4' key pressed down). I've tried it – the System Profile says K32. If I go to the Activity Monitor, I can see that most of my bundled apps like Finder, Mail, Safari and more are running as 64-bit binaries. Yes, I do benefit the use of these 64-bit app binaries instead of the 32-bit binaries. My Samsung laser printer is 32-bit and the 'Smart Panel' software supporting that printer runs as a 32-bit binary.
I rarely use more than 1 GB of memory, even with muliple apps open. I do benefit from SL; I experience better responsiveness across the OS and it's bundled apps. Kudos to Grand Central Dispatch! So, I hope to postpone buying a new iMac until 2012 (five years past 2007).
As a previous post mentioned, Apple limited the booting of K64 to certain machines to avoid compatibility problems. It make good business sense. It is possible that a future point release of SL will add drivers to support the earlier machines (before 2008) with Intel Core 2 Duo CPUs. I'm not privy to Apple's product plans.
It's possible that Apple compiled the 64-bit kernel with SSE4.1. That would explain why only Macs with Penryn or newer processors support the 64-bit kernel making 64-bit EFI not the only requirement. I'm just speculating though. Unless someone knows how to check if software is compiled to require a certain SSE set. I believe normally OS X software is compiled against SSE2, while parts of OS X have required SSE3 like Rosetta.
EDIT: For reference 65nm Yonah Core Duos have SSE3, 65nm Merom Core 2 Duos have SSSE3, 45nm Penryn Core 2 Duos have SSE4.1, and 45nm Nehalem has SSE4.2. I believe the minimum SSE requirement was set at SSE2 because the AppleTV uses an older Dothan Pentium M which only has SSE2. All 64-bit Intel processors used in Macs have at least SSSE3 so that really should have been the minimum compile target for 64-bit apps.
It's possible that Apple compiled the 64-bit kernel with SSE4.1. That would explain why only Macs with Penryn or newer processors support the 64-bit kernel making 64-bit EFI not the only requirement. I'm just speculating though. Unless someone knows how to check if software is compiled to require a certain SSE set. I believe normally OS X software is compiled against SSE2, while parts of OS X have required SSE3 like Rosetta.
EDIT: For reference 65nm Yonah Core Duos have SSE3, 65nm Merom Core 2 Duos have SSSE3, 45nm Penryn Core 2 Duos have SSE4.1, and 45nm Nehalem has SSE4.2. I believe the minimum SSE requirement was set at SSE2 because the AppleTV uses an older Dothan Pentium M which only has SSE2. All 64-bit Intel processors used in Macs have at least SSSE3 so that really should have been the minimum compile target for 64-bit apps.
Well, I've got the Merom. I'll survive. By the way, I saw a comment that the ATI Radeon HD 2600 does not support double-precision floating point, and thus cannot support OpenCL. I don't know enough to confirm that comment.
After you click 'About This Mac' a System Profiler window will appear. On the left side of that window, in the 'Content' column, click directly on 'Software'. In the column to the right you will see 'System Software Overview'. There you will see '64-bit Kernel and Extensions:'. If 'Yes', then your Mac OS X is running in it's 64-bit kernel. This line is definitive.
I forgot to add 'click "More Info?' to get to expanded System Profiler window
Many thanks ? found it. I was looking everywhere BUT in the overview section (in extensions and frameworks, etc).
cheers
Well, I've got the Merom. I'll survive. By the way, I saw a comment that the ATI Radeon HD 2600 does not support double-precision floating point, and thus cannot support OpenCL. I don't know enough to confirm that comment.
Double precision floating point should have nothing to do with supporting OpenCL. In fact, the nVidia 8000 and 9000 series don't support DP floats either. Only the ATI HD3000, HD4000, and nVidia GTX200 series do. I'm fairly certain that DP floats are not part of OpenCL 1.0 spec anyways and will come in subsequent versions most likely because of the small installed base right now.
I have asked AMD's Stream Computing team and they've told me that OpenCL support won't be coming to the HD2000 and HD3000 series because of a hardware limitation. They didn't confirm what it was, but I'm thinking it has to do with their memory hierarchy. nVidia GPUs have a local memory store in their Streaming Multiprocessors (SMs) shared between 8 Stream Processors (SPs). The HD2000 and HD3000 series seem to have their shared memory store at a higher level, between all SPs. With OpenCL being closer to CUDA, it wouldn't surprise me if lack of local memory store is a problem. The Folding@home team has identified lack of local memory store in the HD2000 and HD3000 series as a major performance impediment, even though Folding@home is using ATI's native Brook+ language for the ATI client. ATI's HD4000 series adopts a memory model closer to nVidia's with local memory stores so that could certainly be a differentiating factor in OpenCL compatibility. It's not like I'm an expert though. It's just my guess from what I remember from GPU architecture reviews.
Double precision floating point should have nothing to do with supporting OpenCL. In fact, the nVidia 8000 and 9000 series don't support DP floats either. Only the ATI HD3000, HD4000, and nVidia GTX200 series do. I'm fairly certain that DP floats are not part of OpenCL 1.0 spec anyways and will come in subsequent versions most likely because of the small installed base right now.
I have asked AMD's Stream Computing team and they've told me that OpenCL support won't be coming to the HD2000 and HD3000 series because of a hardware limitation. They didn't confirm what it was, but I'm thinking it has to do with their memory hierarchy. nVidia GPUs have a local memory store in their Streaming Multiprocessors (SMs) shared between 8 Stream Processors (SPs). The HD2000 and HD3000 series seem to have their shared memory store at a higher level, between all SPs. With OpenCL being closer to CUDA, it wouldn't surprise me if lack of local memory store is a problem. The Folding@home team has identified lack of local memory store in the HD2000 and HD3000 series as a major performance impediment, even though Folding@home is using ATI's native Brook+ language for the ATI client. ATI's HD4000 series adopts a memory model closer to nVidia's with local memory stores so that could certainly be a differentiating factor in OpenCL compatibility. It's not like I'm an expert though. It's just my guess from what I remember from GPU architecture reviews.
I don't have the knowledge to discuss what you said, except I'm aware that OpenCL is derived from CUDA. John Siracusa at Ars Technica, in his review of SL, had some good
info in his review regarding OpenCL. Check it out if you have not read it.
I don't have the knowledge to discuss what you said, except I'm aware that OpenCL is derived from CUDA. John Siracusa at Ars Technica, in his review of SL, had some good
info in his review regarding OpenCL. Check it out if you have not read it.
I actually asked John whether he thought memory hierarchy might be the problem with HD2000 and HD3000 support. He thought it was possible although he also speculated that a good driver implementation might also be able to work around it. Even if that were the case, seeing the already less than spectacular performance of the HD4000 series, I doubt it'd be worthwhile to support older ATI GPUs since emulating the memory model would make the end performance not worthwhile.
I actually asked John whether he thought memory hierarchy might be the problem with HD2000 and HD3000 support. He thought it was possible although he also speculated that a good driver implementation might also be able to work around it. Even if that were the case, seeing the already less than spectacular performance of the HD4000 series, I doubt it'd be worthwhile to support older ATI GPUs since emulating the memory model would make the end performance not worthwhile.
OK. I enjoyed this discussion.
If I remember correctly, the last G5 powermacs were sold alongside the first core duo Intel Macs. The g5s just got cut off. Would it really surprise you if 10.7 didn't support 32 bit machines.
My core 2 duo MBP has 32 bit EFI firmware and I'm not holding my breath that 10.7 will support my machine.
If we assume 18 months for the nexzt overhaul it?ll be over 4 years since Macs went both 64-bit CPUs -AND- 64-bit chipsets (remember the first C2Ds only had 32-bit RAM addressing), so that seems likely. The real question is whether enough 3rd-party drivers have gone 64-bit -AND- there isn?t enough old HW out there that won?t get a 64-bit driver but is too costly to replace or just inconvenient for the user. I?d say that 64-bit will at likely be the default for 10.7, but I?m not sold yet on the 32-bit kernel being nixed when it?s not hard to keep, when I assume that 10.7 will mainly be a UI overhaul more than all these core changes we?ve seen in 10.6.
As to my current laptop 8GB, 2.5GHz Core 2 Duo, 500GB 7200RPM HD, Vista 64-bit Ultimate SP2, total cost ~$1,800. Price a 15" MacBook Pro with these specs for me if you don't mind.
How can price the same HW when you didn?t give any decent specifics that affect the price. You?ve only posted the surface specs that people with basic technical understanding of computers look at. Don?t get me wrong, that is fine. I can find a 17? notebook for half the price of a 13? MacBook, but that doesn?t mean it?s better simply because the display is bigger and with higher resolution.
Furthermore, you can find Dells that has the same basic specs as a 15? MBP but cost anywhere from a fraction of the price to about the same price. Is Dell screwing customers with those more expensive machines or do those machines with higher L2, lower TDP, and extra thin chassises that are also durable also cost money to make? I think you are missing the point about what is good enough for your needs and what is an actual bona fide comparable machine.
A final thought. Vista x64 is great. Windows 7 x64 hasn?t really shown to speed much up, only that they have thinned out the overly hostile processes so it performs better on slower/older HW and that without these processes there is more RAM and more cycles to be used more efficiently. Since SP1 Vista has many of the initial kinks worked out, but I ask you, what would you do if you were in a situation where you had to use a printer, say at a hotel, and the printer driver was only 32-bit? You can?t restart into 32-bit mood and reinstalling with the 32-bit kernel means no more than 4GB of memory addressing and no 64-bit apps. Apple?s stepped solution for 64-bit is ideal for the average consumer.
How old are you?
I was very likely doing 64-bit programming on a Cray supercomputer before you were even born.
25 years ago, to be exact.
As to my current laptop 8GB, 2.5GHz Core 2 Duo, 500GB 7200RPM HD, Vista 64-bit Ultimate SP2, total cost ~$1,800. Price a 15" MacBook Pro with these specs for me if you don't mind.
Well it's hard to do an informative comparison without full specs for your laptop... in June I did a price comparison on a dual quad core Mac Pro and a Dell T7500:
Dell T7500: $6062
Dual Quad Core 2.66 GHz Xeon X5550
4 GB 1066 MHz memory (they didn't have an option for 6GB 1006 MHz mem)
NVIDIA Quadro FX 580, 512 MB VRAM ($175 retail)
1 TB Hard Drive, SATA 3Gb/s, 7200 RPM, 16 MB cache
16x DVD+/-RW Drive
Apple MacPro: $4,999
Dual Quad Core 2.66 GHz Xeon X5550
6 GB 1066 MHz memory
NVIDIA GeForce GT 120 512 MB ($150 retail)
1TB Hard Drive, SATA 3Gb/s, 7200 RPM, 32 MB cache
18x double-layer SuperDrive (DVD±R DL/DVD±RW/CD-RW)
The Dell is more scalable (1 more drive bay, 1 more PCI express slot, and 4 more memory slots) and it comes with a support plan (I think Apple Care is ~$300) but I'll go with the superior Apple build quality and save some $$$. By chance, I checked again earlier today, and the price difference is still ~$1k (although this time Dell has a 6GB option)... also checked an HP box (can't remember which) with Linux installed and the price was closer, but Apple was still cheaper... I think by ~$200-$300.
I do have to agree with other posters that the 64 bit transition is inevitable, and not conceptually any different than the 16-32 bit transition... I actually do find 64 bit useful for my line of work... processing of very large images.
And I did my first 60 bit coding on a CDC 7600 in 1980, so I'm no spring chicken...
Well it's hard to do an informative comparison without full specs for your laptop... in June I did a price comparison on a dual core Mac Pro and a Dell T7500:
Dell T7500: $6062
Dual Quad Core 2.66 GHz Xeon X5550
4 GB 1066 MHz memory (they didn't have an option for 6GB 1006 MHz mem)
NVIDIA Quadro FX 580, 512 MB VRAM ($175 retail)
1 TB Hard Drive, SATA 3Gb/s, 7200 RPM, 16 MB cache
16x DVD+/-RW Drive
Apple MacPro: $4,999
Dual Quad Core 2.66 GHz Xeon X5550
6 GB 1066 MHz memory
NVIDIA GeForce GT 120 512 MB ($150 retail)
1TB Hard Drive, SATA 3Gb/s, 7200 RPM, 32 MB cache
18x double-layer SuperDrive (DVD±R DL/DVD±RW/CD-RW)
The Dell is more scalable (1 more drive bay, 1 more PCI express slot, and 4 more memory slots) and it comes with a support plan (I think Apple Care is ~$300) but I'll go with the superior Apple build quality and save some $$$. By chance, I checked again earlier today, and the price difference is still ~$1k (although this time Dell has a 6GB option)... also checked an HP box (can't remember which) with Linux installed and the price was closer, but Apple was still cheaper... I think by ~$200-$300.
I do have to agree with other posters that the 64 bit transition is inevitable, and not conceptually any different than the 16-32 bit transition... I actually do find 64 bit useful for my line of work... processing of very large images.
And I did my first 60 bit coding on a CDC 7600 in 1980, so I'm no spring chicken...
I also checked Alienware for a 15" laptop today and the specs/price are as follows:
Core2duo 2.8Ghz 800 Mhz bus (6 MB cache)
Mobility Radeon 3870
4 GB Ram (667 Mhz DDR2)
500 GB 7200 RPM HD
1920*1200 Resolution (lacking on Apple's 15" line atm)
Dual Layer 8x DVD burner
2299$
Apple 15" Mbp
Core2duo 2.8Ghz 1066 Mhz bus (3 MB cache)
9600GT M + 9400M
4 GB Ram (1066 Mhz DDR3)
500 GB 7200 RPM HD
1440*900 Resolution
Superdrive
2349$
The GPU's are almost similar in performance but Alienware has higher res screen (although I can't imagine using a 1080p screen at 15", even on my 17" Mbp, things are hard to read)
Mac has faster bus + RAM, Alienware has 3 MB more cache
+ On top of all that, the design is not even comparable. Mac is much much much thinner and looks way better. Alienware has one of the worst laptop designs I saw on PC's. Sony VAIO's look much nicer.
Edit: Alienware is more configurable in terms of GPU, if you pay 350$ more you get a 9800GT M, but Apple is more configurable in terms of CPU, you can get a 3.06 GHz.
And the price difference is only 50$.
I also checked Alienware for a 15" laptop today and the specs/price are as follows:
Core2duo 2.8Ghz 800 Mhz bus (6 MB cache)
Mobility Radeon 3870
4 GB Ram (667 Mhz DDR2)
500 GB 7200 RPM HD
1920*1200 Resolution (lacking on Apple's 15" line atm)
Dual Layer 8x DVD burner
2299$
Apple 15" Mbp
Core2duo 2.8Ghz 1066 Mhz bus (3 MB cache)
9600GT M + 9400M
4 GB Ram (1066 Mhz DDR3)
500 GB 7200 RPM HD
1440*900 Resolution
Superdrive
2349$
The GPU's are almost similar in performance but Alienware has higher res screen (although I can't imagine using a 1080p screen at 15", even on my 17" Mbp, things are hard to read)
Mac has faster bus + RAM, Alienware has 3 MB more cache
+ On top of all that, the design is not even comparable. Mac is much much much thinner and looks way better. Alienware has one of the worst laptop designs I saw on PC's. Sony VAIO's look much nicer.
Edit: Alienware is more configurable in terms of GPU, if you pay 350$ more you get a 9800GT M, but Apple is more configurable in terms of CPU, you can get a 3.06 GHz.
And the price difference is only 50$.
Asus Core 2 Duo Laptops (26 models to choose from, $999 < 13 Models < $1299)
My favorite? The N51 series.
ASUS N51 Series N51Vn-X1A P8700(2.53GHz) 15.6" 4GB Memory 320GB HDD 7200rpm NVIDIA GeForce GT 240M ($1049, no 7% MS sales tax either)
1 GB GPU (48 stream processors), HDMI, eSATA
ASUS N51 Series N51VN-A1 T9600(2.80GHz) 15.6" 4GB Memory 320GB HDD 7200rpm NVIDIA GeForce GT 240M ($1249, no 7% MS sales tax either)
1 GB GPU (48 stream processors), HDMI, eSATA
Same model as above at ZipZoomfly for $1149 (in stock, no tax/shipping charges, e. g. final price)
I just wonder what drivers are available for the MacBook that aren't available? My iMac is an 8,1.
As you've discovered, there's apparently a little more to it than even 64-bit capable CPU and EFI.
Amit Singh (Author of Max OS X Internals) has written a piece, about this, titled "Is Your Machine Good Enough for Snow Leopard"
If we assume 18 months for the nexzt overhaul it’ll be over 4 years since Macs went both 64-bit CPUs -AND- 64-bit chipsets (remember the first C2Ds only had 32-bit RAM addressing), so that seems likely. The real question is whether enough 3rd-party drivers have gone 64-bit -AND- there isn’t enough old HW out there that won’t get a 64-bit driver but is too costly to replace or just inconvenient for the user. I’d say that 64-bit will at likely be the default for 10.7, but I’m not sold yet on the 32-bit kernel being nixed when it’s not hard to keep, when I assume that 10.7 will mainly be a UI overhaul more than all these core changes we’ve seen in 10.6.
I also don't think the 32-bit kernel is going to disappear anytime soon. There will still be quite a few niche devices that will be slow to release 64-bit drivers, perhaps mixers or input devices in the audio/visual realm. It's highly likely though that 64-bit Intel processors will be the requirement in 10.7 and the 64-bit kernel made the default, while the 32-bit kernel is still available just like it is on the XServes now. The 32-bit kernel may disappear in 10.8, but that is still a long ways off. Even when the 32-bit kernel disappears, 32-bit frameworks to run 32-bit apps in a 64-bit OS will remain since some apps really don't need to move to 64-bit anyways. Hopefully Rosetta will remain an optional install for a while otherwise what would we do without older classics like Starcraft?