Stated by Steve Jobs at its annual share holders meeting when asked about clock speed and it's low performing chips..
"there will be a time" when Apple would speak about its relationship with Motorola, and that the particular shareholder who asked the question would be invited, if he desired.
That shareholder will be allowed a few swings of the baseball bat, once Steve, Phil and Avi have softened the guy up a bit of course.
FiringSquad: Can you describe the process involved in migrating to AMD's 64-bit architecture? Has the transition been a difficult one?
Since our code is pure C++ and already ran on 32-bit Windows and Linux, the only work required was to make the code 64-bit safe. No Hammer-specific work was necessary to get the port up and running; what we did for Hammer is the same thing that would be needed to run on 64-bit PowerPC or 64-bit Itanium.
In the case of the Unreal codebase, about 99.9% of the code was already 64-bit safe and didn't need touching. Of course, with a million-line codebase, the remaining 0.1% left a hundred or so places in the code that needed updating because of assumptions we made years ago before we'd thought about 64-bit. It was a relatively straightforward process, and took Ryan Gordon about 10 days of hard work.
This bodes well if any 64-bit PowerPC comes our way
I just want to amplify on rickag's response. Everyone seems to take this rumor as fact, but it really doesn't make much sense on its own. Apple isn't the only buyer of G4s, and while it would hurt if they don't buy, Moto can get other buyers. As richag points out, it is possible that Moto looked at the 85xx series and choose to develop it into a replacement - only time will tell. I think there is a lot of unknowns to this. The 7457-RM (on paper) is an excellent design, it only helps Apple to have as many processor options as possible.
Now, we're not saying that 64-bit Photoshop et al running natively on AMD64 CPUs is going to happen anytime soon, or that if it does, pro Mac users are going to switch over in droves.
Yea, ok, right. I guess the writer is a windows guy. When will pc idiots get that by now they should realize most use mac for the OS and photosop isn't animation or high end video.
I'm sorry to hear that the 7457 RM cancelled your Motorola. Although I only have an old 7400, it has never cancelled my Motorola, so I think this bug is inexcusable and should be reported.
Oh -- About what Jobs said. Aside from a legal lawsuit, could Apple sever all ties with Motorola? I mean, can IBM cover all the CPUs Apple is currently using (i.e G3s for iBooks, standard G4s for the eMac, "notebook friendly" G4s for the Pro portables)?
In Other Other Words, say only the PowerMacs get a major revision with the 970. Could Apple switch over to IBM in toto and declare a new "partnership" with IBM. (Motorola gets Steved).
Oh -- About what Jobs said. Aside from a legal lawsuit, could Apple sever all ties with Motorola? I mean, can IBM cover all the CPUs Apple is currently using (i.e G3s for iBooks, standard G4s for the eMac, "notebook friendly" G4s for the Pro portables)?
In Other Other Words, say only the PowerMacs get a major revision with the 970. Could Apple switch over to IBM in toto and declare a new "partnership" with IBM. (Motorola gets Steved).
I'd also like to reference this thread, where Inkhead calls the device a "prototype 1 of the webpad appliance that apple has already shipped off" even though we later found it has a Palm "hotsync" logo on the original photo.
"But by far the most interesting revelation, Brian Valentine at Microsoft, was that it and AMD planned the introduction of the 64-bit processor family right from the start of the project.
"Valentine said that his company will offer Windows Server 2003 with built in support for AMD. AMD and Microsoft had been in together on the development of a 64-bit processor.
"Microsoft liked the idea of 64-bit computing being backward compatible with 32 bit applications, at a low cost, he claimed, and closely collaborated on its development.
"...Server 2003, launched on the 24th, will support 64-bit Opterons in 32-bit mode, and in 64-bit mode later this year, he said"
Win64 is guaranteed for AMD64 and will likely be here in the Sept-Dec time frame (given the date of the Athlon64 launch in Sept and the statement of 64-bit betas mid-year in the 2nd link).
So the competition for Apple and the 970 with OSX - 64 bit will be stiff. They need to execute rapidly and flawlessly to get a jump on AMD in consumer 64 bit land. The most recent rumors re WWDC, etc, point to a possible intro before the Athlon64. This would be great for them - another first, and this time one that can significantly boost their prospects.
Pity the 7457-RM might have been cancelled - it removes a good 32 bit PPC with fast FSB (and hence less bottleneck for AltiVec, etc - probably would do much better at SPECcpu2000 than the current G4s) for use in the lower end macs. But it might be because Apple is planning a full migration to the 64 bit 970 like AMD wants PC users to buy Opterons and Athlon64s across the whole range of the market.
"But by far the most interesting revelation, Brian Valentine at Microsoft, was that it and AMD planned the introduction of the 64-bit processor family right from the start of the project.
"Valentine said that his company will offer Windows Server 2003 with built in support for AMD. AMD and Microsoft had been in together on the development of a 64-bit processor.
"Microsoft liked the idea of 64-bit computing being backward compatible with 32 bit applications, at a low cost, he claimed, and closely collaborated on its development.
"...Server 2003, launched on the 24th, will support 64-bit Opterons in 32-bit mode, and in 64-bit mode later this year, he said"
This means that windows as we know it at the moment [32bit X86] will run unaltered in the Opteron. This will not however, allow any sort of 64 bit computing to take place. They aren't altering the kernel, nor doing anything special. Win2003 runs in 32bit compatibility mode. OS X on the other hand, /will/ be re-tooled to allow the kernel to run in 64bit mode [other apps can be 32bit]. This will allow the [mach] kernel to map *far* more memory than a pure 32bit mode.
No doubt that the competition will be fierce, but I do believe that MS will delay the 64bit version of their server [note SERVER, not DESKTOP] until they see proof that the Opterons are selling well. What is the point in allocating resources if no one buys the machine? Also note that the Opteron *does* take a bit of a speed hit in 32bit mode. The Unreal Tourn. 2003 benchmarks that were floating around were for a specially compiled 64 bit version.
I think the 970 as a platform will do quite well, and perhaps better than the AMD offering [as far as quick adoption rate is concerned]
This means that windows as we know it at the moment [32bit X86] will run unaltered in the Opteron. This will not however, allow any sort of 64 bit computing to take place. They aren't altering the kernel, nor doing anything special. Win2003 runs in 32bit compatibility mode. OS X on the other hand, /will/ be re-tooled to allow the kernel to run in 64bit mode [other apps can be 32bit]. This will allow the [mach] kernel to map *far* more memory than a pure 32bit mode.
True that the initial release of Windows server for Opteron will be 32 bit. But note that they are planning the 64-bit release this year as mentioned in the articles I referenced. The 64-bit version will obviously have a lot of re-written stuff. Not being an OS programmer I can't say what but they will obviously be supporting the Opteron's 48-bit addressing so I imagine a lot of re-tooling will be going into the whole OS. (256 TB)
Quote:
No doubt that the competition will be fierce, but I do believe that MS will delay the 64bit version of their server [note SERVER, not DESKTOP] until they see proof that the Opterons are selling well. What is the point in allocating resources if no one buys the machine? Also note that the Opteron *does* take a bit of a speed hit in 32bit mode. The Unreal Tourn. 2003 benchmarks that were floating around were for a specially compiled 64 bit version.
They are not going to do the work over a period of years and then not release the product as they 'wait-and-see'. They have had the resources allocated since the beginning. They will release it when it's ready (ie, when they have added sufficient bugs ).
Opteron doesn't take a hit in 32-bit mode at all. If you have proof, please provide it but I refer you to the links below for comparisons with various Athlons, P4s and Xeons:
The 2nd link has a lot of bench tests under WinXP (32-bit) and the single Opteron 1.8GHz bests even a 2.1 Ghz AthlonXP in most of them.
Quote:
I think the 970 as a platform will do quite well, and perhaps better than the AMD offering [as far as quick adoption rate is concerned]
I mostly agree with the first bit - I think in percentage of the market Apple can only increase. I think the AMD64 chips (Opteron and Athlon64) will give very good competition and we would do well not to underestimate the challenge. There is a lot of pent up demand in both camps for the new 64 bit hardware.
Opteron doesn't take a hit in 32-bit mode at all. If you have proof, please provide it but I refer you to the links below for comparisons with various Athlons, P4s and Xeons:
The 2nd link has a lot of bench tests under WinXP (32-bit) and the single Opteron 1.8GHz bests even a 2.1 Ghz AthlonXP in most of them.
MM
I read the benchmarks, and they didn't do what they should do when talking about "taking a hit"... compare a 32bit version of an app with the 64bit version running on the same system [machine/OS]. No doubt, the AMD chip rocks, but you can't say that it doesn't take a hit in performance when you only show it in comparison to other machines/chips. You need to compare a single hardware platform, and two versions of the same app [32bit and 64bit] and see if they come out the same [in theory, they should, assuming the test doesn't try to "fake" 64bits in the 32bit domain, as the real 64bit version would SPANK it]
This page clearly shows that the 64bit implementation is faster than the 32bit [with the exception of the second row]
Look at the "Floating Point Performance with Flops" table.
Do I think it is a huge difference? Well, no. As you pointed out, che chip is very fast in either 32 or 64 bit mode, fast enough to beat quite a few of its competition, even with the slight performance hit. Mind you, this is from the 64bit version being able to use the additional registers of the Opteron chip. The 32bit compatibility mode essentially ignores all the spiffy new features on the chip.
People are forgetting that both Intel and Microsoft have been playing down 64-bit for the mainstream desktop for quite a while recently. Any news as of late that states otherwise seems to be in stark contrast to the "end of the decade" time frame estimated. I and others have sort of already gone over the reasons why it isn't likely that we'll see any 64-bit Windows computing coming to the mainstream desktop any time soon. To recap ...
- Operating System ... Let's face it, what would you *expect* Micro$oft to say with the approaching 970 and 64-bit OS X? Remember, Microsoft is in bed with Intel before AMD. Try not to forget that. One of the problems the Itanium faced was arriving to market with no OS from Microsoft ready to go. It was stated in one of the articles I posted in another thread. Then there is Microsoft support of "yet another version of Windows". How many forks will this be? 25? 26? That's a lot of different versions of Windows Microsoft has to worry about. Managing that can't be easy.
- Developer Support. Will developers code for another hardware platform that essentially runs the *same* Operating system? Compatibility across multiple versions? Drivers? You get the picture... Oh, and let's not forget about all the *legacy* systems and applications floating around. Many of which were sold just today. What will become of them? Will developers allot the resources to maintain multiple builds for legacy, current and future builds? Will this bring about "forks" in applications as well? I wonder what Programmer and Amorph would have to say about forking at this stage in the Windows platform. How long before the apps start rolling in? How much more confusion will this bring to the masses?
- Intel. Will Microsoft stab Intel in the back and effectively force them to kill Itanic, a chip that Intel is looking to have succeed at whatever cost? This will effectively strap them to x86 for another God-knows-how-long period of time. Since Intel has been looking to move forward away from x86, forcing them to compete against AMD in 64-bit x86 space would likely cause them a great deal of harm and a significant lag in regard to time to market with a viable competing solution. Then there is that developer support question again. This time add Intel hardware into the mix. Hardware which would no doubt be vastly different from AMD's solution, which would pose even more questions.
So, in all honesty, it looks like it's your basic dog-and-pony show being put on by Microsoft and AMD. They are simply announcing their future plans in such a way as to make it sound that all this new and nifty stuff is right around the corner and you better wait or you'll "miss out" if you switch to another platform. This is typical Wintelon.... Hell, there isn't even a beta-release of a 64-bit *desktop* version of the Windows OS, but again, you better wait, you don't want to miss this.. It's the next greatest thing. yadda yadda yadda.... Anyway, it certainly doesn't look clear to me that there will be a 64-bit Windows/AMD-x86-64 desktop solution with all the nifty apps "right around the corner". Have I missed anything? This was all discussed at length before in other threads. It should be clear that AMD and Microsoft are just making a lot of noise. No doubt in light of the 970. Microsoft wants to be able to say "Me too! Me too!" Do you really think they want people to start considering alternatives?
- Operating System ... Let's face it, what would you *expect* Micro$oft to say with the approaching 970 and 64-bit OS X? Remember, Microsoft is in bed with Intel before AMD. Try not to forget that. One of the problems the Itanium faced was arriving to market with no OS from Microsoft ready to go.
Microsoft is out to make money. They do this [among other things] by running on the X86 platform. In order for the Opteron to survive, Windows must run on the platform. The volume isn't there in the FSF universe [Linux/BSD]. Microsoft will ensure that their product runs, as they wish to maintain their monopoly [and increase revenue] in the X86 universe. When Intel comes out with their 64bit version that runs in X86 land, MS will be there in a nanosecond. Again, this is to keep the monopoly.
Quote:
- Developer Support. Will developers code for another hardware platform that essentially runs the *same* Operating system? Compatibility across multiple versions? Drivers? You get the picture... Oh, and let's not forget about all the *legacy* systems and applications floating around. Many of which were sold just today. What will become of them? Will developers allot the resources to maintain multiple builds for legacy, current and future builds? Will this bring about "forks" in applications as well? I wonder what Programmer and Amorph would have to say about forking at this stage in the Windows platform. How long before the apps start rolling in? How much more confusion will this bring to the masses?
- Intel. Will Microsoft stab Intel in the back and effectively force them to kill Itanic, a chip that Intel is looking to have succeed at whatever cost?
The "legacy" software will run just fine on the Opteron. Check out the Ars Technica article on how the Opteron works here. New drivers may not even need to be rewritten, but I wouldn't take that as gospel. Essentially, the Opteron is a normal 32bit X86 chip with some additional registers, and the ability to play with 64bit addressing if needed. It is due to this that allows for the backwards compatibility [with a very minor hit in comparison to a 64bit version of the app [see my previous post].
And as for MS stabbing Intel in the back, I don't think that'll happen either. Both companies are out to make money. They will come up with a solution that will allow both to have maximum return. Never doubt that.
Quote:
So, in all honesty, it looks like it's your basic dog-and-pony show being put on by Microsoft and AMD. They are simply announcing their future plans in such a way as to make it sound that all this new and nifty stuff is right around the corner and you better wait or you'll "miss out" if you switch to another platform.
I agree this is a bit of "oh look here, we're cool too" in that regard. This will come into light when the Opterons are shipping in quantity. If MS quickly comes out with a real 64 bit version of the OS that allows for 64 bit binaries, then I will change my belief. MS just wants to make sure that whichever platform comes out king, their OS is the OS running on it. Plain and simple.
Apple sure need some new legs. The low end tower is alredy on sale here in Sweden 200 dollars less than the original price of only a couple of weeks ago.
Not that I am surprised, compared to the midrange it offers 40% of the CPU performance at 75% of the price...No wonder that they have to dump the price allmost right form the start. My guess is that until the 970 we will have the weakest sales of pro/semipro computers in Apples history. (that is the Mac II, Quadra/Centris, Nubus PPC, PCI PPC, G3 and so on)
An other indication is IMGs test of the ATI 9000 pro card. In the PC world this is a budget card costing less than 100 dollars. In a Mac running at 1.35 GHz the CPU limitiation hinders the card from being stressed with the 1600x1200 score about 85% about that of the 640x480. So while the late arrival of the 9700 card to Mac might irritate some it seems that the cheapest budget card is good enough for now
Apple do not need a G4plus but to substract the G4 alltogher! The lastest rumor at http://macbidouille.com/ indicate that Apple want to have 500 M from Motorola, such requests hardly improve the rellation between Apple and Motorola. "The lawyers of APPLE gave an opinion favorable to the continuations against Motorola. They estimate that the prospect of winning is very good. It is not less than 500 million Dollars which will be claimed." (a slightly brushed up babel fish translation).
If they start using the IBM 970 and sue Motorola for at least half a billion will they still buy G4s from them? Suing Moto for making underperforming G4 on one hand and on the other hand selling computers with G4 at the same time One way would be to sue over the lack of G5 and claiming to be happy with the G4, riiiiiight!
It might turn into something that made the parting between Ford and Firestone to have been friendly\
People are forgetting that both Intel and Microsoft have been playing down 64-bit for the mainstream desktop for quite a while recently.
Can you provide any links where Microsoft has said these kinds of things in the past? (At the risk of looking like an MS appologist) I believe it was always Intel doing the talking on this front.
Quote:
- Operating System ... Let's face it, what would you *expect* Micro$oft to say with the approaching 970 and 64-bit OS X? Remember, Microsoft is in bed with Intel before AMD. Try not to forget that. One of the problems the Itanium faced was arriving to market with no OS from Microsoft ready to go.
Please refer to my 2nd link in reply to Matsu above. (9th April MS press release) Regardless of what they may have said in the past they have actually given a timetable for a 64 bit version of WinXP (ie consumer Windows OS) by saying the beta will be release mid year. Final version probably by Christmas (probably 2004 given the past record of late releases though )
I read the benchmarks, and they didn't do what they should do when talking about "taking a hit"... compare a 32bit version of an app with the 64bit version running on the same system [machine/OS]. No doubt, the AMD chip rocks, but you can't say that it doesn't take a hit in performance when you only show it in comparison to other machines/chips. You need to compare a single hardware platform, and two versions of the same app [32bit and 64bit] and see if they come out the same [in theory, they should, assuming the test doesn't try to "fake" 64bits in the 32bit domain, as the real 64bit version would SPANK it]
This page clearly shows that the 64bit implementation is faster than the 32bit [with the exception of the second row]
Look at the "Floating Point Performance with Flops" table.
Do I think it is a huge difference? Well, no. As you pointed out, che chip is very fast in either 32 or 64 bit mode, fast enough to beat quite a few of its competition, even with the slight performance hit. Mind you, this is from the 64bit version being able to use the additional registers of the Opteron chip. The 32bit compatibility mode essentially ignores all the spiffy new features on the chip.
I initialy thought you were talking about the Opteron in 32 bit vs the Athlon in 32 bit, particularly when you brought up UT2003 (which anadtech.com shows running on Win XP rocks against allcomers). Sorry if I was mistaken.
Anyway, if we go to the first table on the Ace's page you link to we see that the Opteron in for a 32-bit (IA-32) binary operating on 32-bit integers and a 64-bit binary (AMD64) using 64-bit integers the performance of the core is exactly the same. When you have a 64-bit binary using 32-bit integers the Opteron takes a perfomance un-hit (glass half full or half empty?) "probably due to greater loop-unrolling". So it is not getting hit because of the mode it is running in at all but getting a benefit in 64-bit mode working on 32-bit integers because of architectural improvements.
As far as the 970 goes, it too in 32-bit mode on 32 bit data it should be the same speed as 64-bit mode on 64-bit data. In 64-bit mode working on 32-bit data it may not have any performance advantage. Can anyone fill in the balnks on this?
Comments
Originally posted by Producer
Stated by Steve Jobs at its annual share holders meeting when asked about clock speed and it's low performing chips..
"there will be a time" when Apple would speak about its relationship with Motorola, and that the particular shareholder who asked the question would be invited, if he desired.
That shareholder will be allowed a few swings of the baseball bat, once Steve, Phil and Avi have softened the guy up a bit of course.
</untouchables>
FiringSquad: Can you describe the process involved in migrating to AMD's 64-bit architecture? Has the transition been a difficult one?
Since our code is pure C++ and already ran on 32-bit Windows and Linux, the only work required was to make the code 64-bit safe. No Hammer-specific work was necessary to get the port up and running; what we did for Hammer is the same thing that would be needed to run on 64-bit PowerPC or 64-bit Itanium.
In the case of the Unreal codebase, about 99.9% of the code was already 64-bit safe and didn't need touching. Of course, with a million-line codebase, the remaining 0.1% left a hundred or so places in the code that needed updating because of assumptions we made years ago before we'd thought about 64-bit. It was a relatively straightforward process, and took Ryan Gordon about 10 days of hard work.
This bodes well if any 64-bit PowerPC comes our way
Originally posted by Nonsuch
That shareholder will be allowed a few swings of the baseball bat, once Steve, Phil and Avi have softened the guy up a bit of course.
</untouchables>
LOL
Originally posted by inkhead
I know this won't prove anything but it might help some of you see that I do know what I'm talking about :
http://www.theregister.co.uk/content/39/30372.html
From that article-
Now, we're not saying that 64-bit Photoshop et al running natively on AMD64 CPUs is going to happen anytime soon, or that if it does, pro Mac users are going to switch over in droves.
Yea, ok, right. I guess the writer is a windows guy. When will pc idiots get that by now they should realize most use mac for the OS and photosop isn't animation or high end video.
there is a lot of justified pro user dissatisfaction with the mac, something like that would be the last straw for many.
however, considering 64 windows is nowhere to be seen (for AMD) yet, 64 bit AMD PS seems further off than a 64 bit PS8 for OSX.
What were we...?
Oh -- About what Jobs said. Aside from a legal lawsuit, could Apple sever all ties with Motorola? I mean, can IBM cover all the CPUs Apple is currently using (i.e G3s for iBooks, standard G4s for the eMac, "notebook friendly" G4s for the Pro portables)?
In Other Other Words, say only the PowerMacs get a major revision with the 970. Could Apple switch over to IBM in toto and declare a new "partnership" with IBM. (Motorola gets Steved).
Screed
Originally posted by sCreeD
Oh dear, glad I bought a Sony-Ericsson instead.
What were we...?
Oh -- About what Jobs said. Aside from a legal lawsuit, could Apple sever all ties with Motorola? I mean, can IBM cover all the CPUs Apple is currently using (i.e G3s for iBooks, standard G4s for the eMac, "notebook friendly" G4s for the Pro portables)?
In Other Other Words, say only the PowerMacs get a major revision with the 970. Could Apple switch over to IBM in toto and declare a new "partnership" with IBM. (Motorola gets Steved).
Screed
I think Steve got Motted a long time ago.
Originally posted by CharlesS
Given this, I'd guess the former...
I'd also like to reference this thread, where Inkhead calls the device a "prototype 1 of the webpad appliance that apple has already shipped off" even though we later found it has a Palm "hotsync" logo on the original photo.
Why do we keep believing this guy?
Originally posted by Matsu
regardless, a 64 bit PS7 on AMD would get a lot of mac pros to switch over if it beat a 64 bit mac shop by a significant margin.
there is a lot of justified pro user dissatisfaction with the mac, something like that would be the last straw for many.
however, considering 64 windows is nowhere to be seen (for AMD) yet, 64 bit AMD PS seems further off than a 64 bit PS8 for OSX.
Sorry to spoil your party Matsu but check out this:
http://www.theinquirer.net/?article=9092
"But by far the most interesting revelation, Brian Valentine at Microsoft, was that it and AMD planned the introduction of the 64-bit processor family right from the start of the project.
"Valentine said that his company will offer Windows Server 2003 with built in support for AMD. AMD and Microsoft had been in together on the development of a 64-bit processor.
"Microsoft liked the idea of 64-bit computing being backward compatible with 32 bit applications, at a low cost, he claimed, and closely collaborated on its development.
"...Server 2003, launched on the 24th, will support 64-bit Opterons in 32-bit mode, and in 64-bit mode later this year, he said"
Here's another older (9th April '03) ref:
http://www.microsoft.com/presspass/p...nOpteronPR.asp
Win64 is guaranteed for AMD64 and will likely be here in the Sept-Dec time frame (given the date of the Athlon64 launch in Sept and the statement of 64-bit betas mid-year in the 2nd link).
So the competition for Apple and the 970 with OSX - 64 bit will be stiff. They need to execute rapidly and flawlessly to get a jump on AMD in consumer 64 bit land. The most recent rumors re WWDC, etc, point to a possible intro before the Athlon64. This would be great for them - another first, and this time one that can significantly boost their prospects.
Pity the 7457-RM might have been cancelled - it removes a good 32 bit PPC with fast FSB (and hence less bottleneck for AltiVec, etc - probably would do much better at SPECcpu2000 than the current G4s) for use in the lower end macs. But it might be because Apple is planning a full migration to the 64 bit 970 like AMD wants PC users to buy Opterons and Athlon64s across the whole range of the market.
MM
[edit - added 2nd link]
Originally posted by MartianMatt
[B]Sorry to spoil your party Matsu but check out this:
http://www.theinquirer.net/?article=9092
"But by far the most interesting revelation, Brian Valentine at Microsoft, was that it and AMD planned the introduction of the 64-bit processor family right from the start of the project.
"Valentine said that his company will offer Windows Server 2003 with built in support for AMD. AMD and Microsoft had been in together on the development of a 64-bit processor.
"Microsoft liked the idea of 64-bit computing being backward compatible with 32 bit applications, at a low cost, he claimed, and closely collaborated on its development.
"...Server 2003, launched on the 24th, will support 64-bit Opterons in 32-bit mode, and in 64-bit mode later this year, he said"
This means that windows as we know it at the moment [32bit X86] will run unaltered in the Opteron. This will not however, allow any sort of 64 bit computing to take place. They aren't altering the kernel, nor doing anything special. Win2003 runs in 32bit compatibility mode. OS X on the other hand, /will/ be re-tooled to allow the kernel to run in 64bit mode [other apps can be 32bit]. This will allow the [mach] kernel to map *far* more memory than a pure 32bit mode.
No doubt that the competition will be fierce, but I do believe that MS will delay the 64bit version of their server [note SERVER, not DESKTOP] until they see proof that the Opterons are selling well. What is the point in allocating resources if no one buys the machine? Also note that the Opteron *does* take a bit of a speed hit in 32bit mode. The Unreal Tourn. 2003 benchmarks that were floating around were for a specially compiled 64 bit version.
I think the 970 as a platform will do quite well, and perhaps better than the AMD offering [as far as quick adoption rate is concerned]
Originally posted by visigothe
This means that windows as we know it at the moment [32bit X86] will run unaltered in the Opteron. This will not however, allow any sort of 64 bit computing to take place. They aren't altering the kernel, nor doing anything special. Win2003 runs in 32bit compatibility mode. OS X on the other hand, /will/ be re-tooled to allow the kernel to run in 64bit mode [other apps can be 32bit]. This will allow the [mach] kernel to map *far* more memory than a pure 32bit mode.
True that the initial release of Windows server for Opteron will be 32 bit. But note that they are planning the 64-bit release this year as mentioned in the articles I referenced. The 64-bit version will obviously have a lot of re-written stuff. Not being an OS programmer I can't say what but they will obviously be supporting the Opteron's 48-bit addressing so I imagine a lot of re-tooling will be going into the whole OS. (256 TB)
No doubt that the competition will be fierce, but I do believe that MS will delay the 64bit version of their server [note SERVER, not DESKTOP] until they see proof that the Opterons are selling well. What is the point in allocating resources if no one buys the machine? Also note that the Opteron *does* take a bit of a speed hit in 32bit mode. The Unreal Tourn. 2003 benchmarks that were floating around were for a specially compiled 64 bit version.
They are not going to do the work over a period of years and then not release the product as they 'wait-and-see'. They have had the resources allocated since the beginning. They will release it when it's ready (ie, when they have added sufficient bugs
Opteron doesn't take a hit in 32-bit mode at all. If you have proof, please provide it but I refer you to the links below for comparisons with various Athlons, P4s and Xeons:
http://www.aceshardware.com/read.jsp?id=55000251
http://www.tomshardware.com/cpu/2003...pteron-23.html (direct to the WinXP tests)
The 2nd link has a lot of bench tests under WinXP (32-bit) and the single Opteron 1.8GHz bests even a 2.1 Ghz AthlonXP in most of them.
I think the 970 as a platform will do quite well, and perhaps better than the AMD offering [as far as quick adoption rate is concerned]
I mostly agree with the first bit - I think in percentage of the market Apple can only increase. I think the AMD64 chips (Opteron and Athlon64) will give very good competition and we would do well not to underestimate the challenge. There is a lot of pent up demand in both camps for the new 64 bit hardware.
MM
Originally posted by MartianMatt
Opteron doesn't take a hit in 32-bit mode at all. If you have proof, please provide it but I refer you to the links below for comparisons with various Athlons, P4s and Xeons:
http://www.aceshardware.com/read.jsp?id=55000251
http://www.tomshardware.com/cpu/2003...pteron-23.html (direct to the WinXP tests)
The 2nd link has a lot of bench tests under WinXP (32-bit) and the single Opteron 1.8GHz bests even a 2.1 Ghz AthlonXP in most of them.
MM
I read the benchmarks, and they didn't do what they should do when talking about "taking a hit"... compare a 32bit version of an app with the 64bit version running on the same system [machine/OS]. No doubt, the AMD chip rocks, but you can't say that it doesn't take a hit in performance when you only show it in comparison to other machines/chips. You need to compare a single hardware platform, and two versions of the same app [32bit and 64bit] and see if they come out the same [in theory, they should, assuming the test doesn't try to "fake" 64bits in the 32bit domain, as the real 64bit version would SPANK it]
This page clearly shows that the 64bit implementation is faster than the 32bit [with the exception of the second row]
http://www.aceshardware.com/read.jsp?id=55000265
Look at the "Floating Point Performance with Flops" table.
Do I think it is a huge difference? Well, no. As you pointed out, che chip is very fast in either 32 or 64 bit mode, fast enough to beat quite a few of its competition, even with the slight performance hit. Mind you, this is from the 64bit version being able to use the additional registers of the Opteron chip. The 32bit compatibility mode essentially ignores all the spiffy new features on the chip.
- Operating System ... Let's face it, what would you *expect* Micro$oft to say with the approaching 970 and 64-bit OS X? Remember, Microsoft is in bed with Intel before AMD. Try not to forget that. One of the problems the Itanium faced was arriving to market with no OS from Microsoft ready to go. It was stated in one of the articles I posted in another thread. Then there is Microsoft support of "yet another version of Windows". How many forks will this be? 25? 26? That's a lot of different versions of Windows Microsoft has to worry about. Managing that can't be easy.
- Developer Support. Will developers code for another hardware platform that essentially runs the *same* Operating system? Compatibility across multiple versions? Drivers? You get the picture... Oh, and let's not forget about all the *legacy* systems and applications floating around. Many of which were sold just today. What will become of them? Will developers allot the resources to maintain multiple builds for legacy, current and future builds? Will this bring about "forks" in applications as well? I wonder what Programmer and Amorph would have to say about forking at this stage in the Windows platform. How long before the apps start rolling in? How much more confusion will this bring to the masses?
- Intel. Will Microsoft stab Intel in the back and effectively force them to kill Itanic, a chip that Intel is looking to have succeed at whatever cost? This will effectively strap them to x86 for another God-knows-how-long period of time. Since Intel has been looking to move forward away from x86, forcing them to compete against AMD in 64-bit x86 space would likely cause them a great deal of harm and a significant lag in regard to time to market with a viable competing solution. Then there is that developer support question again. This time add Intel hardware into the mix. Hardware which would no doubt be vastly different from AMD's solution, which would pose even more questions.
So, in all honesty, it looks like it's your basic dog-and-pony show being put on by Microsoft and AMD. They are simply announcing their future plans in such a way as to make it sound that all this new and nifty stuff is right around the corner and you better wait or you'll "miss out" if you switch to another platform. This is typical Wintelon.... Hell, there isn't even a beta-release of a 64-bit *desktop* version of the Windows OS, but again, you better wait, you don't want to miss this.. It's the next greatest thing. yadda yadda yadda.... Anyway, it certainly doesn't look clear to me that there will be a 64-bit Windows/AMD-x86-64 desktop solution with all the nifty apps "right around the corner". Have I missed anything? This was all discussed at length before in other threads. It should be clear that AMD and Microsoft are just making a lot of noise. No doubt in light of the 970. Microsoft wants to be able to say "Me too! Me too!" Do you really think they want people to start considering alternatives?
--
Ed
Originally posted by Ed M.
- Operating System ... Let's face it, what would you *expect* Micro$oft to say with the approaching 970 and 64-bit OS X? Remember, Microsoft is in bed with Intel before AMD. Try not to forget that. One of the problems the Itanium faced was arriving to market with no OS from Microsoft ready to go.
Microsoft is out to make money. They do this [among other things] by running on the X86 platform. In order for the Opteron to survive, Windows must run on the platform. The volume isn't there in the FSF universe [Linux/BSD]. Microsoft will ensure that their product runs, as they wish to maintain their monopoly [and increase revenue] in the X86 universe. When Intel comes out with their 64bit version that runs in X86 land, MS will be there in a nanosecond. Again, this is to keep the monopoly.
- Developer Support. Will developers code for another hardware platform that essentially runs the *same* Operating system? Compatibility across multiple versions? Drivers? You get the picture... Oh, and let's not forget about all the *legacy* systems and applications floating around. Many of which were sold just today. What will become of them? Will developers allot the resources to maintain multiple builds for legacy, current and future builds? Will this bring about "forks" in applications as well? I wonder what Programmer and Amorph would have to say about forking at this stage in the Windows platform. How long before the apps start rolling in? How much more confusion will this bring to the masses?
- Intel. Will Microsoft stab Intel in the back and effectively force them to kill Itanic, a chip that Intel is looking to have succeed at whatever cost?
The "legacy" software will run just fine on the Opteron. Check out the Ars Technica article on how the Opteron works here. New drivers may not even need to be rewritten, but I wouldn't take that as gospel. Essentially, the Opteron is a normal 32bit X86 chip with some additional registers, and the ability to play with 64bit addressing if needed. It is due to this that allows for the backwards compatibility [with a very minor hit in comparison to a 64bit version of the app [see my previous post].
And as for MS stabbing Intel in the back, I don't think that'll happen either. Both companies are out to make money. They will come up with a solution that will allow both to have maximum return. Never doubt that.
So, in all honesty, it looks like it's your basic dog-and-pony show being put on by Microsoft and AMD. They are simply announcing their future plans in such a way as to make it sound that all this new and nifty stuff is right around the corner and you better wait or you'll "miss out" if you switch to another platform.
I agree this is a bit of "oh look here, we're cool too" in that regard. This will come into light when the Opterons are shipping in quantity. If MS quickly comes out with a real 64 bit version of the OS that allows for 64 bit binaries, then I will change my belief. MS just wants to make sure that whichever platform comes out king, their OS is the OS running on it. Plain and simple.
Not that I am surprised, compared to the midrange it offers 40% of the CPU performance at 75% of the price...No wonder that they have to dump the price allmost right form the start. My guess is that until the 970 we will have the weakest sales of pro/semipro computers in Apples history. (that is the Mac II, Quadra/Centris, Nubus PPC, PCI PPC, G3 and so on)
An other indication is IMGs test of the ATI 9000 pro card. In the PC world this is a budget card costing less than 100 dollars. In a Mac running at 1.35 GHz the CPU limitiation hinders the card from being stressed with the 1600x1200 score about 85% about that of the 640x480. So while the late arrival of the 9700 card to Mac might irritate some it seems that the cheapest budget card is good enough for now
Apple do not need a G4plus but to substract the G4 alltogher! The lastest rumor at http://macbidouille.com/ indicate that Apple want to have 500 M from Motorola, such requests hardly improve the rellation between Apple and Motorola. "The lawyers of APPLE gave an opinion favorable to the continuations against Motorola. They estimate that the prospect of winning is very good. It is not less than 500 million Dollars which will be claimed." (a slightly brushed up babel fish translation).
If they start using the IBM 970 and sue Motorola for at least half a billion will they still buy G4s from them? Suing Moto for making underperforming G4 on one hand and on the other hand selling computers with G4 at the same time
It might turn into something that made the parting between Ford and Firestone to have been friendly
Originally posted by Ed M.
People are forgetting that both Intel and Microsoft have been playing down 64-bit for the mainstream desktop for quite a while recently.
Can you provide any links where Microsoft has said these kinds of things in the past? (At the risk of looking like an MS appologist) I believe it was always Intel doing the talking on this front.
- Operating System ... Let's face it, what would you *expect* Micro$oft to say with the approaching 970 and 64-bit OS X? Remember, Microsoft is in bed with Intel before AMD. Try not to forget that. One of the problems the Itanium faced was arriving to market with no OS from Microsoft ready to go.
Please refer to my 2nd link in reply to Matsu above. (9th April MS press release) Regardless of what they may have said in the past they have actually given a timetable for a 64 bit version of WinXP (ie consumer Windows OS) by saying the beta will be release mid year. Final version probably by Christmas (probably 2004 given the past record of late releases though
MM
Originally posted by visigothe
I read the benchmarks, and they didn't do what they should do when talking about "taking a hit"... compare a 32bit version of an app with the 64bit version running on the same system [machine/OS]. No doubt, the AMD chip rocks, but you can't say that it doesn't take a hit in performance when you only show it in comparison to other machines/chips. You need to compare a single hardware platform, and two versions of the same app [32bit and 64bit] and see if they come out the same [in theory, they should, assuming the test doesn't try to "fake" 64bits in the 32bit domain, as the real 64bit version would SPANK it]
This page clearly shows that the 64bit implementation is faster than the 32bit [with the exception of the second row]
http://www.aceshardware.com/read.jsp?id=55000265
Look at the "Floating Point Performance with Flops" table.
Do I think it is a huge difference? Well, no. As you pointed out, che chip is very fast in either 32 or 64 bit mode, fast enough to beat quite a few of its competition, even with the slight performance hit. Mind you, this is from the 64bit version being able to use the additional registers of the Opteron chip. The 32bit compatibility mode essentially ignores all the spiffy new features on the chip.
I initialy thought you were talking about the Opteron in 32 bit vs the Athlon in 32 bit, particularly when you brought up UT2003 (which anadtech.com shows running on Win XP rocks against allcomers). Sorry if I was mistaken.
Anyway, if we go to the first table on the Ace's page you link to we see that the Opteron in for a 32-bit (IA-32) binary operating on 32-bit integers and a 64-bit binary (AMD64) using 64-bit integers the performance of the core is exactly the same. When you have a 64-bit binary using 32-bit integers the Opteron takes a perfomance un-hit (glass half full or half empty?) "probably due to greater loop-unrolling". So it is not getting hit because of the mode it is running in at all but getting a benefit in 64-bit mode working on 32-bit integers because of architectural improvements.
As far as the 970 goes, it too in 32-bit mode on 32 bit data it should be the same speed as 64-bit mode on 64-bit data. In 64-bit mode working on 32-bit data it may not have any performance advantage. Can anyone fill in the balnks on this?
MM