64-bit will certainly speed-up any graphics when working with 64-bit colour spaces.
I imagine when a PowerPC 970-based Mac is introduced, there will a Photoshop bake off where the Mac is 10x faster than the Pentium 4 (as opposed to the current 2x faster).
This thread should be locked. There simply is no debate to be had on this issue.
<strong>It seems that the reason to go 64 bit is not very strong compared to the 'standard' processor improvements as bandwith, clockspeed etc.
Why would you want to go 64 bit on desktop computers anyway?</strong><hr></blockquote>
2 GB of RAM is not that much anymore. I consider 512 MB to be minimum for a poweruser. 1-2 GB is certainly not outrageous if you do some heavy Java programming, Photoshopping, 3D modelling or video editing. Before long a lot of people might feel a need for 64 bit. It's important that heavy duty users have this option on the Mac platform or they might switch to the x86-64 platform. I'm sure that Apple doesn't want to lose their most profitable customers.
The Virtual Menory routines of Mac OS X should be more efficient when it gets a larger physical area to address. As it is now it probably need to do some magic to get a couple of really large apps (with a total usage of more than 4 GB of RAM) to play nicely. As it is now the kernel must translate addresses between VM and real memory.
Am I right? If so.. this should reduce some over head and speed things up somewhat across all apps.
<strong>64-bit will certainly speed-up any graphics when working with 64-bit colour spaces.
I imagine when a PowerPC 970-based Mac is introduced, there will a Photoshop bake off where the Mac is 10x faster than the Pentium 4 (as opposed to the current 2x faster).</strong><hr></blockquote>
The reason why Photoshop is faster is that it uses 128 bit Altivec (performing multiple instructions at once). Please tell me how one can improve that using 64 bits?
AltiVec isn't an integer unit (it can be, but primarily), it is a vector unit. So you do multiple instruction which are not depended on the results of the others simultaniously.
In terms of manipulating graphics in Photoshop, most is still done by the traditional int/float units.
Trying to do 64-bit colour on a 32-bit CPU is very possible, but is much slower than if the CPU is 64-bit in the first place.
float units are already 64-bit.
Mac OS X is very memory-intensive, so the ability to address more than 4GB will be an important advantage in the near future.
<strong>There is no issue with the 970 running 32-bit PowerPC code! None. Zip. Nada.</strong><hr></blockquote>
Sure there is. Joe User feeling stupid to not "take advantade of all of the 64 bits". So there - psychological problem!
Oh, btw - since memory managment is done by the CPU - is it possible to leave photoshop as it is and let the OS hand it over 2gb of ram or does photoshop have to be rewritten as well? From what I know the app simply asks for RAM and the OS decides if there is any left - or?
<strong>Agreed, and I think that comparing the 32->64 bit move to the OS9 -> X move for Apple's apps, Belle, is a bit off.</strong><hr></blockquote>
I wasn't really comparing the change to OS X with a change to 64-bit, rather pointing out that the marketing idea that der Kopf suggested - which seems an obvious move - isn't always obvious to Apple.
I just remember finding it odd that at every opportunity Apple would get the point across that developers must get OS X versions out now (in time for the 10.0 release), and yet labored in getting its own cornerstone apps out.
. . . Oh, btw - since memory managment is done by the CPU - is it possible to leave photoshop as it is and let the OS hand it over 2gb of ram or does photoshop have to be rewritten as well? From what I know the app simply asks for RAM and the OS decides if there is any left - or?
</strong><hr></blockquote>
I'd like to hear the answer to that too. Possibly a 64-bit Mac would be a great advantage using only 32-bit applications. You could have many programs running, and each one could use a couple GB of memory. Your virtual memory could be hundreds of GB without a problem.
Edit: Oops, I think Henriok already answered it. Well, at least when RAM prices drop, we can add as much RAM as we like. Looks like virtual memory already has a way to take care of more than 4 GB total.
<strong>All software are currently 32 bit and in order to take the maximum advantage of the hardware they have to be rewritten (write from scratch, I am afraid) to use the extra bandwidth.</strong><hr></blockquote>
Well, right off the bat you've shown that you don't really understand what the advantages of a 64-bit CPU are.
There is no "extra bandwidth" to take advantage of in a 64-bit CPU. The only "advantage" 64 bit paths give you is that you can address (read and write) higher amounts of memory.
[quote]<strong>prepare to spend 3 grand on hardware......</strong><hr></blockquote>
Wrong. The 970 could actually cost less per CPU than the G4, bringing total system costs down.
[quote]<strong>ALSO prepare to spend thousands of dollars on SOFTWARE</strong><hr></blockquote>
Wrong again (as others have already pointed out).
[quote]<strong>EDIT:I DO know this CPU can run 32bit code natively ]</strong><hr></blockquote>
Yes, it's just a pity you don't understand what that means.
And just to emphasize the point: Oracle is now working on an OS X Server native version of Oracle 9i. They are already shipping developers releases and should soon have final product released. Serious Oracle databases can require tremendous amounts of memory and typically run on 8-16 way servers. Do you think it is a co-incidence that Oracle began porting 9i to Mac OS X just before IBM announced the 970?
<strong>The Virtual Menory routines of Mac OS X should be more efficient when it gets a larger physical area to address. As it is now it probably need to do some magic to get a couple of really large apps (with a total usage of more than 4 GB of RAM) to play nicely. As it is now the kernel must translate addresses between VM and real memory.
Am I right? If so.. this should reduce some over head and speed things up somewhat across all apps.</strong><hr></blockquote>
One slight issue that needs to be overcome as far as vitual memory is concerned. Disk I/O access is much, much slower than RAM access. So even if the CPU was much more efficient than the previous one, the bottleneck remains the same. Just like the PowerMacs situation with BUS speeds. You can make the processor faster, but still not be able to feed it fast enough.
We've dodged around one edge of a point that's important I think.
For 95% + of the apps that _consumers_ currently use, a full compilation and full appreciation/adoption/exploitation of the 64bitness of the CPU will yield a 0% speed up. iCal won't be going faster just because of 64bitness. It'll get optimized, and it had BETTER speed up, but it won't be due to 64bitness.
And it extends to some of the bigger programs! Someone mentioned ProEngineer. ProE needs double precision floating point out the wazzu. But... we've had double precision floats since forever. I'm not positive about ProE, but I'm not sure even that will necessarily see a significant speed up. (For those that are unaware, ProE is a high end 3D CAD package where adding various options means you can design/build/test whole widgets. Piping, cabling, mech e, plastics, sheet metal etc etc. Very nice.)
By far the biggest advantage for most of us is access to more than 4GB of RAM. That doesn't generally require a full switch to a 64bit CPU though.
For the libraries/Frameworks, won't the 'Fat binary' history of NeXt help out here? Just readd the architectures to the frameworks - Quicktime/ppc32 _and_ Quicktime/ppc64. Use the proper version at runtime. A 64 bit app gets whatever benefits it expected from the 64bit library, while allowing the 32 bit app to completely ignore that other 32bits.
If 24 bit color means 16 million colors, a 5000x5000 pixel image could almost have a unique color on every pixel. (5000 times 5000 equals 25 million)<hr></blockquote>
Yes, but typical images don't use every color, just a narrow range of color. A night scene might use primarily blues and greys for example. 24 bit color only uses an 8 bit color plane for each of the primary colors, which means that at 24 bits you only have 256 different shades of blue. This can lead to banding and can obscure fine details. 64 bit color increases each color plane to 16 bits, giving 65536 shades and reducing these problems.
[quote]now we have 32 bit color as standard...Who will ever need 64 bit images?<hr></blockquote>
32 bit color still only has 16.8 million colors, the extra 8 bits are an alpha channel.
[quote] ...16.8 million colors, the extra 8 bits are an alpha channel... <hr></blockquote>
16million colors are enought for 98% of people. As I remember, they done a 64bits colors-videocard years ago. They never sell it because 64bits color is just more colors than the eyes can see. Even with 32bits like now, find people (exept experienced artist) who can see the difference between a 0,75,255 RGB blue and a 0,75,253 RGB blue ! 12bits per channel (thus make like 48bits color) would be near the limit of the human eyes.
I don't think anything being discussed in this thread could make PPC970 based macs suck, at least not any more than G4 based macs (which aren't that bad except for their prices).
What could make a 970 suck? 2 things. One, an unneccessary increase in PM prices (and any increase will be unneccessary!) Two, a huge delay in delivery, Apple, mid 2003 we hear, NOT A DAY LATER, Apple, even if you have to sacrifice the first born of IBM engineers throughout the lower 48 states, NOT A DAY LATER. No shitty excuses.
Matsu, no-one has said mid-2003 is a certainty. Not Apple, not IBM.
Mid-2003 is the best case senario. IBM said it aims for a 2nd half 2003 release date.
130nm moving to 90nm fits with the mid-2003 date, as Fishkill NY is starting at 130nm mid-2003, moving to 90nm.
The 6-month release target range is to give any space for delays.
The worst probably case would be a MacWorld SanFran 2004 introduction.
I'm probably going to sell this G4 before MacWorld New York 2003, and am prepared to wait until MacWorld SanFran for the next Power Mac. That's being realistic.
Comments
I imagine when a PowerPC 970-based Mac is introduced, there will a Photoshop bake off where the Mac is 10x faster than the Pentium 4 (as opposed to the current 2x faster).
This thread should be locked. There simply is no debate to be had on this issue.
Barto
<strong>It seems that the reason to go 64 bit is not very strong compared to the 'standard' processor improvements as bandwith, clockspeed etc.
Why would you want to go 64 bit on desktop computers anyway?</strong><hr></blockquote>
2 GB of RAM is not that much anymore. I consider 512 MB to be minimum for a poweruser. 1-2 GB is certainly not outrageous if you do some heavy Java programming, Photoshopping, 3D modelling or video editing. Before long a lot of people might feel a need for 64 bit. It's important that heavy duty users have this option on the Mac platform or they might switch to the x86-64 platform. I'm sure that Apple doesn't want to lose their most profitable customers.
Am I right? If so.. this should reduce some over head and speed things up somewhat across all apps.
<strong>64-bit will certainly speed-up any graphics when working with 64-bit colour spaces.
I imagine when a PowerPC 970-based Mac is introduced, there will a Photoshop bake off where the Mac is 10x faster than the Pentium 4 (as opposed to the current 2x faster).</strong><hr></blockquote>
The reason why Photoshop is faster is that it uses 128 bit Altivec (performing multiple instructions at once). Please tell me how one can improve that using 64 bits?
In terms of manipulating graphics in Photoshop, most is still done by the traditional int/float units.
Trying to do 64-bit colour on a 32-bit CPU is very possible, but is much slower than if the CPU is 64-bit in the first place.
float units are already 64-bit.
Mac OS X is very memory-intensive, so the ability to address more than 4GB will be an important advantage in the near future.
Barto
<strong>There is no issue with the 970 running 32-bit PowerPC code! None. Zip. Nada.</strong><hr></blockquote>
Sure there is. Joe User feeling stupid to not "take advantade of all of the 64 bits". So there - psychological problem!
Oh, btw - since memory managment is done by the CPU - is it possible to leave photoshop as it is and let the OS hand it over 2gb of ram or does photoshop have to be rewritten as well? From what I know the app simply asks for RAM and the OS decides if there is any left - or?
<strong>Agreed, and I think that comparing the 32->64 bit move to the OS9 -> X move for Apple's apps, Belle, is a bit off.</strong><hr></blockquote>
I wasn't really comparing the change to OS X with a change to 64-bit, rather pointing out that the marketing idea that der Kopf suggested - which seems an obvious move - isn't always obvious to Apple.
I just remember finding it odd that at every opportunity Apple would get the point across that developers must get OS X versions out now (in time for the 10.0 release), and yet labored in getting its own cornerstone apps out.
Anyway, I'm going off topic.
If 24 bit color means 16 million colors, a 5000x5000 pixel image could almost have a unique color on every pixel. (5000 times 5000 equals 25 million)
now we have 32 bit color as standard...Who will ever need 64 bit images?
Isn't it possible to use the 64 bit processing power in a more clever way?
ap
[ 11-29-2002: Message edited by: ap ]</p>
<strong>
. . . Oh, btw - since memory managment is done by the CPU - is it possible to leave photoshop as it is and let the OS hand it over 2gb of ram or does photoshop have to be rewritten as well? From what I know the app simply asks for RAM and the OS decides if there is any left - or?
</strong><hr></blockquote>
I'd like to hear the answer to that too. Possibly a 64-bit Mac would be a great advantage using only 32-bit applications. You could have many programs running, and each one could use a couple GB of memory. Your virtual memory could be hundreds of GB without a problem.
Edit: Oops, I think Henriok already answered it. Well, at least when RAM prices drop, we can add as much RAM as we like. Looks like virtual memory already has a way to take care of more than 4 GB total.
[ 11-29-2002: Message edited by: snoopy ]</p>
<strong>All software are currently 32 bit and in order to take the maximum advantage of the hardware they have to be rewritten (write from scratch, I am afraid) to use the extra bandwidth.</strong><hr></blockquote>
Well, right off the bat you've shown that you don't really understand what the advantages of a 64-bit CPU are.
There is no "extra bandwidth" to take advantage of in a 64-bit CPU. The only "advantage" 64 bit paths give you is that you can address (read and write) higher amounts of memory.
[quote]<strong>prepare to spend 3 grand on hardware......</strong><hr></blockquote>
Wrong. The 970 could actually cost less per CPU than the G4, bringing total system costs down.
[quote]<strong>ALSO prepare to spend thousands of dollars on SOFTWARE</strong><hr></blockquote>
Wrong again (as others have already pointed out).
[quote]<strong>EDIT:I DO know this CPU can run 32bit code natively ]</strong><hr></blockquote>
Yes, it's just a pity you don't understand what that means.
And just to emphasize the point: Oracle is now working on an OS X Server native version of Oracle 9i. They are already shipping developers releases and should soon have final product released. Serious Oracle databases can require tremendous amounts of memory and typically run on 8-16 way servers. Do you think it is a co-incidence that Oracle began porting 9i to Mac OS X just before IBM announced the 970?
I don't.
<strong>The Virtual Menory routines of Mac OS X should be more efficient when it gets a larger physical area to address. As it is now it probably need to do some magic to get a couple of really large apps (with a total usage of more than 4 GB of RAM) to play nicely. As it is now the kernel must translate addresses between VM and real memory.
Am I right? If so.. this should reduce some over head and speed things up somewhat across all apps.</strong><hr></blockquote>
One slight issue that needs to be overcome as far as vitual memory is concerned. Disk I/O access is much, much slower than RAM access. So even if the CPU was much more efficient than the previous one, the bottleneck remains the same. Just like the PowerMacs situation with BUS speeds. You can make the processor faster, but still not be able to feed it fast enough.
Edit: can't spell
[ 11-29-2002: Message edited by: MacJedai ]</p>
For 95% + of the apps that _consumers_ currently use, a full compilation and full appreciation/adoption/exploitation of the 64bitness of the CPU will yield a 0% speed up. iCal won't be going faster just because of 64bitness. It'll get optimized, and it had BETTER speed up, but it won't be due to 64bitness.
And it extends to some of the bigger programs! Someone mentioned ProEngineer. ProE needs double precision floating point out the wazzu. But... we've had double precision floats since forever. I'm not positive about ProE, but I'm not sure even that will necessarily see a significant speed up. (For those that are unaware, ProE is a high end 3D CAD package where adding various options means you can design/build/test whole widgets. Piping, cabling, mech e, plastics, sheet metal etc etc. Very nice.)
By far the biggest advantage for most of us is access to more than 4GB of RAM. That doesn't generally require a full switch to a 64bit CPU though.
For the libraries/Frameworks, won't the 'Fat binary' history of NeXt help out here? Just readd the architectures to the frameworks - Quicktime/ppc32 _and_ Quicktime/ppc64. Use the proper version at runtime. A 64 bit app gets whatever benefits it expected from the 64bit library, while allowing the 32 bit app to completely ignore that other 32bits.
<strong>Do you think it is a co-incidence that Oracle began porting 9i to Mac OS X just before IBM announced the 970?
I don't.</strong><hr></blockquote>
PPC 970 had nothing to do with that.
Hint: <a href="http://www.oracle.com/ip/index.html?rac_home.html"; target="_blank">http://www.oracle.com/ip/index.html?rac_home.html</a>;
<strong>PPC 970 had nothing to do with that.</strong><hr></blockquote>
You're saying Apple's clustering plans did?
<strong>
You're saying Apple's clustering plans did?</strong><hr></blockquote>
No, Oracle's clustering capabilities. There is no need for a big Oracle server - just several small ones (Xserve anyone?).
[ 11-29-2002: Message edited by: JLL ]</p>
If 24 bit color means 16 million colors, a 5000x5000 pixel image could almost have a unique color on every pixel. (5000 times 5000 equals 25 million)<hr></blockquote>
Yes, but typical images don't use every color, just a narrow range of color. A night scene might use primarily blues and greys for example. 24 bit color only uses an 8 bit color plane for each of the primary colors, which means that at 24 bits you only have 256 different shades of blue. This can lead to banding and can obscure fine details. 64 bit color increases each color plane to 16 bits, giving 65536 shades and reducing these problems.
[quote]now we have 32 bit color as standard...Who will ever need 64 bit images?<hr></blockquote>
32 bit color still only has 16.8 million colors, the extra 8 bits are an alpha channel.
16million colors are enought for 98% of people. As I remember, they done a 64bits colors-videocard years ago. They never sell it because 64bits color is just more colors than the eyes can see. Even with 32bits like now, find people (exept experienced artist) who can see the difference between a 0,75,255 RGB blue and a 0,75,253 RGB blue ! 12bits per channel (thus make like 48bits color) would be near the limit of the human eyes.
What could make a 970 suck? 2 things. One, an unneccessary increase in PM prices (and any increase will be unneccessary!) Two, a huge delay in delivery, Apple, mid 2003 we hear, NOT A DAY LATER, Apple, even if you have to sacrifice the first born of IBM engineers throughout the lower 48 states, NOT A DAY LATER. No shitty excuses.
<strong>
16million colors are enought for 98% of people.</strong><hr></blockquote>
Bull$hit. So was 640k.
I can tell the difference between certain colours in 24-bit. 64-bit will be much closer to the limit.
Also, 64-bit colour spaces ≠ 64-bit final image.
Working in 64-bit and rendering at 32-bit gives a much higher-quality image, as you have more precision when applying effects.
Barto
Mid-2003 is the best case senario. IBM said it aims for a 2nd half 2003 release date.
130nm moving to 90nm fits with the mid-2003 date, as Fishkill NY is starting at 130nm mid-2003, moving to 90nm.
The 6-month release target range is to give any space for delays.
The worst probably case would be a MacWorld SanFran 2004 introduction.
I'm probably going to sell this G4 before MacWorld New York 2003, and am prepared to wait until MacWorld SanFran for the next Power Mac. That's being realistic.
Barto