I have not said f***g otherwise. I, out of my head, said that for non-Altivec tasks the Pentium M is 50% faster than a G4. That number is naturally quite arbitrary but it certainly fits your description that the 'Pentium M beats a 7447a with no problem'.
But nobody is talking about that comparison, since there won't be any Pentium M native version of Photoshop running on a Mac for about a year. You see, I don't care how fast Photoshop runs under Windows because using Windows is simply not an option for me.
I am talking about how Photoshop runs on G4s compared to what it would run under Rosetta on a Pentium M.
That's what I'm talking about as well. Intel's new chips will be even faster. the point being that the extra power in these chips combined with the fact that they will be dual core, will allow PS even without Altivec support to have a roughly comparable performance. remember that Apple will NOT be using the P M chip, but its successors.
Even if Rosetta only works with one core (which we don't know as yet), the OS will still speed it up over a single G4 core design.
That's what I'm talking about as well. Intel's new chips will be even faster. the point being that the extra power in these chips combined with the fact that they will be dual core, will allow PS even without Altivec support to have a roughly comparable performance. remember that Apple will NOT be using the P M chip, but its successors.
Even if Rosetta only works with one core (which we don't know as yet), the OS will still speed it up over a single G4 core design.
If Photoshop via Rosetta on dual-core Yonahs will be as fast as on current Powerbooks, then this would require IMHO a quite dramatic speed up of dual-core Yonahs in respect to current Pentium Ms. Going dual-core might provide this dramatic speed increase, but then Windows users should be really rejoicing, since their version of Photoshop will see dramatic speed increases when they switch to dual-cores. Then again, maybe dual processor support in Windows might be so poor that they won't see this dramatic increase.
SSE doesn't equal Altivec. Even SSE3 which so far, I believe, only AMD supports isn't the same thing.
SSE3 can do much of what Altivec does. Altivec still has some advantages, but for a large percentage of the Altivec codebase, SSE3 will suffice.
The trouble is, relevant to this discussion, Rosetta doesn't support either, so stuff that depends on Altivec for good performance is going to run like a dog under Rosetta. Some Photoshop functions fall into this category.
Therefore, the native Intel compile will be much faster because it will use SSE3. If Rosetta is upgraded by next year to do Altivec->SSE3 translation, we're good to go. But until then, well, that's why G5 towers are getting converted over last.
Carbon is the exact opposite and heralds from a time when writing for a specific platform was seen as neccessary to eek out more performance.
You're thinking of the Mac Toolbox, much of which got left behind in the change to Carbon.
Carbon began life as the cross-platform port Mac Toolbox to Windows for Quicktime for Windows.
Now, you can do things under Carbon like use WaitNextEvent() instead of Carbon Events. That kind of stuff has been discouraged for years but several big apps still do it. A good, clean Carbon app is likely to have less other cruft like inline assembly, so it should be easier to port, but that's as much good development practices as it is the application framework. People who have been lazy and 'getting by' on the backs of Mac users who suffer their legacy code are in for the most rude awakenings. Good.
I'm definitely not thinking of the mac toolbox. There are huge differences in the underpinnings of cocoa and carbon, one of which being how hardware abstraction figured into their design.
Despite your denials, this discussion was started with a comparison between PPC Powerbooks and Mactel Powerbooks (or whatever they will be called).
It was between G4's and whatever Apple would be using instead. Rosetta performance being based on that. From the last line of the report:
"It has been reported that Apple holds one final PowerPC-based PowerBook G4 update that should debut this fall."
That's because there are 4 stories merged into the one report. The dwindling supply situation of 17" Powerbooks has nothing to do with the very first story - Adobe's Bruce Chizen saying that porting isn't easy.
The discussion got mixed but I was only ever talking about the performance of non-native software and when would be optimal to shift to Intel.
Quote:
Originally posted by melgross
We aren't talking about G5's, except in the unlikely chance that Apple will use them for this last PB update. We are talking about Powerbook performance.
You may not want to talk about it. That's fine. Then limit your discussion to something else. As I said, if you want to talk about workstations, fine . But don't mix the chipsets, and compare the M Dothan to the Dual G5's. No one else is doing so.
I never brought them up. Only comparison I did was Pentium M v G5 now running native software on each OS. I'm not impressed with the Pentium M. Sorry.
Quote:
Originally posted by melgross
Also, you may not keep with things, but more work is being done on PB's all the time. I know quite a few pro's from my own dealings through my company who do most of their PS work on a portable.
Yup, because they are fast "enough" as they are now for many people. People put up with a lot less performance on a laptop just to get portability. These days it's not so much a big deal slumming it on a laptop as there's more than enough for most people.
Quote:
Originally posted by melgross
If you have such performance differences, I suggest that you have problems of your own with your equipment.
It was just an off the shelf comparison. I can't see why the Pentium M laptop was so slow bearing in mind the hype they get. Just wasn't impressed as I was expecting more. My iMac toasted it.
Quote:
Originally posted by melgross
Again, it's a poor comparison to compare a laptop with a tower. Compare your 1.8GHz G5 with the same price PC tower and come back here. You will be rather disappointed.
I won't because for me it's not about ultimate speed. I'd have to run Windows for one so I'd get an awful lot less done and some things I couldn't do. But I don't see why I should upgrade to Intel if it's slower than I currently have at running the software I use now. That's all.
Quote:
Originally posted by melgross
There is no point in discussing this any further until you understand what the discussion is about, and can make the proper comparisons.
You seem to be making lots of hardware comparisons but it's about software.
Quote:
Originally posted by melgross
Laptop use by pro's is a VERY common thing. Both for PS and video editing. Maybe you're not a full time pro and so don't have to work under the same conditions. But I can tell you that many pro's work on location with laptops. More all the time. You're dismissing that is something I can't understand.
Most Pros I know have both. But it's totally irrelevant to the discussion of when Adobe will get their software native and therefore when it will be viable to upgrade for most pros.
Quote:
Originally posted by melgross
With digital having almost completely taken over the professional market in catalog, fashion, and studio product photography, laptops are becoming ever more important.
If I was a jobbing photographer then yes, if I'm sat in the same office every day churning out print or web designs then no, give me a desktop any day. Bigger screens, faster hard drives, better keyboards.
The point you're consistently missing here is that I don't care what they stuff in a box to run my software, just that it doesn't run slower than it does now. If you think a dual Yonah running a G3 emulation will run faster than today's G4 laptops then great. But the meat and potatoes for me is when they've got computers (laptop or desktop) that run software faster than the G5s be that an iMac or a PowerMac. I don't see that happening till Adobe gets it's software native AND there's significantly faster CPUs than currently in the Intel roadmap.
So my bet is it'll be worthwhile upgrading to an Intel PowerMac or Intel PowerBook only when the software is native and since Bruce is saying it'll be a while, that might dovetail in with Apple having some kit that may run faster than a PowerMac G5. May as well wait for OSX 10.5 too as 10.4 is going to be a lash up at best.
It's the Quark effect all over. Nobody in print design upgraded to OSX till Quark was there or they otherwise got fed up and went to InDesign or Windows. Even now, most shops I know still keep OS9 machines around to run Quark4.11. I can see it happening again but this time it's more serious as people won't be buying hardware till it runs Adobe, Macromedia and Apple Pro apps faster than their old PPC hardware. What would be the point in upgrading otherwise? That's more dangerous for Apple than the OS9->OSX switch.
That's because there are 4 stories merged into the one report. The dwindling supply situation of 17" Powerbooks has nothing to do with the very first story - Adobe's Bruce Chizen saying that porting isn't easy.
The discussion got mixed but I was only ever talking about the performance of non-native software and when would be optimal to shift to Intel.
I never brought them up. Only comparison I did was Pentium M v G5 now running native software on each OS. I'm not impressed with the Pentium M. Sorry.
Yup, because they are fast "enough" as they are now for many people. People put up with a lot less performance on a laptop just to get portability. These days it's not so much a big deal slumming it on a laptop as there's more than enough for most people.
It was just an off the shelf comparison. I can't see why the Pentium M laptop was so slow bearing in mind the hype they get. Just wasn't impressed as I was expecting more. My iMac toasted it.
I won't because for me it's not about ultimate speed. I'd have to run Windows for one so I'd get an awful lot less done and some things I couldn't do. But I don't see why I should upgrade to Intel if it's slower than I currently have at running the software I use now. That's all.
You seem to be making lots of hardware comparisons but it's about software.
Most Pros I know have both. But it's totally irrelevant to the discussion of when Adobe will get their software native and therefore when it will be viable to upgrade for most pros.
If I was a jobbing photographer then yes, if I'm sat in the same office every day churning out print or web designs then no, give me a desktop any day. Bigger screens, faster hard drives, better keyboards.
The point you're consistently missing here is that I don't care what they stuff in a box to run my software, just that it doesn't run slower than it does now. If you think a dual Yonah running a G3 emulation will run faster than today's G4 laptops then great. But the meat and potatoes for me is when they've got computers (laptop or desktop) that run software faster than the G5s be that an iMac or a PowerMac. I don't see that happening till Adobe gets it's software native AND there's significantly faster CPUs than currently in the Intel roadmap.
So my bet is it'll be worthwhile upgrading to an Intel PowerMac or Intel PowerBook only when the software is native and since Bruce is saying it'll be a while, that might dovetail in with Apple having some kit that may run faster than a PowerMac G5. May as well wait for OSX 10.5 too as 10.4 is going to be a lash up at best.
It's the Quark effect all over. Nobody in print design upgraded to OSX till Quark was there or they otherwise got fed up and went to InDesign or Windows. Even now, most shops I know still keep OS9 machines around to run Quark4.11. I can see it happening again but this time it's more serious as people won't be buying hardware till it runs Adobe, Macromedia and Apple Pro apps faster than their old PPC hardware. What would be the point in upgrading otherwise? That's more dangerous for Apple than the OS9->OSX switch.
Aegis, you're the one making the G5 to mobile cpu comparisons, and they're invalid. Compare the iMac to a Xenon or other machine in its class. The G5 loses. Not by too much, but it loses.
IBM has shown us nothing past what they have just announced. Intels offerings over the next year, trounce that. Over the next two years they bury it. The same for the mobile lines.
When running an app in Rosetta, that has to be accounted for. Right now, the comparisons are being made to chips that exist now. By the time Apple begins the switch, Intels chips will be more powerful than they are now, and dual designs as well.
The simple point that I've been trying to make is that as these phases ripple through the Mac line several things will happen.
First, when Powerbooks go Intel, they will have much more powerful dual chips in them than they will have had with the PPC chip line just before. As the speed comparisons are being made with the current P M chip, they come out to about 65-75% of the performance of the current Macs's. The 7448 will add 10% improvement to that. So the final performance can be between 60-70% of the best that a PPC Powerbook will be able to to. But the Yonah will be about 25% more powerful than the current P M. It will also have dual cores. This combination should cause the Rosetta performance to increase to close to the best PPC PB times. The second half of 2006, Merom will be out with 10-15% greater performance and less power requirements.
This is good news.
For desktops the story is the same, but takes place over a greater time period. Don't forget that the PPC Powermac will be out until end of 2007, well after Adobe says that their software will be ported over. Even the IMac G5 will be around until the end of 2006, if not longer. Adobe will be finished by then as well.
I don't see what the problem is then. You'll be able to run CS2 or 3 as long as you need with a PPC machine. When Apple is ready, so will be Adobe. The only area in which Adobe will be late is for the PB. But As I've shown, that won't be a problem. Perfomance certainly won't improve under Rosetta - that's not its purpose, but by the time it's needed fot pro apps, things should be ok.
And you DON'T have to get a Macintel until you want to.
Aegis, you're the one making the G5 to mobile cpu comparisons, and they're invalid. Compare the iMac to a Xenon or other machine in its class. The G5 loses. Not by too much, but it loses.
Not in this thread I wasn't. Not until YOU brought up laptops! We had previously discussed this in another thread and there it made sense.
And what has a Xenon got to do with it. I'm not going to be running Photoshop on an XBox! Photoshop also has enough problems on multiprocessor boxes without having to give it the burden of working on a chopped PPC 3 core CPU that only does in-order code.
Or did you mean a Xeon? which is a server class chip and costs as much for the chip as a whole iMac and usually comes in pairs and still gets beaten by the PowerMac in benchmarks.
Personally I'd take Dual Opterons anyway - and I do. I've 3 Dual Opteron servers and 2 Dual Xeons. The AMD boxes win hands down. And were quite a bit cheaper too. But that's servers, not Photoshop boxes.
Quote:
Originally posted by melgross
IBM has shown us nothing past what they have just announced. Intels offerings over the next year, trounce that. Over the next two years they bury it. The same for the mobile lines.
Sorry. You're partly wrong.
The 970mp will still trounce the Intel desktop chips for the next year at least and all but the most expensive workstation/server chips which typically cost $1000 each when Intel release them. Intel have so far limited their dual-core chips to allow only 1 CPU in a desktop class box and the 970mp isn't. You couldn't run a dual-dual Intel Conroe, Yonah or Merom with their current policy whereas you could a Dual-dual 970mp.
Intel's mobile lineup is impressive, particularly for power requirements. A dual-core processor will go most of the way to mitigate for what we're losing in PPC and what will be taken away if we have no native software. That's provided Apple use the dual-cores of course.
Quote:
Originally posted by melgross
When running an app in Rosetta, that has to be accounted for. Right now, the comparisons are being made to chips that exist now. By the time Apple begins the switch, Intels chips will be more powerful than they are now, and dual designs as well.
I agree. But so far Rosetta is showing speeds of 25-50% at integer and 10% floating point of the P4 3.6Ghz Intel dev boxes and doesn't do AltiVec at all. This partly confirms reports that Apple's design goal was 800Mhz G3 speeds for Rosetta for the dev boxes. It feels snappier of course because most of the time your app is sitting in native core graphics calls doing nothing but you're going to hit the slow stuff as soon as you try a photoshop filter or use an airbrush.
My point, way back at the start of this thread was that switching to Intel would be pointless until the software is native and there were Intel machines significantly faster than today or even the next year.
I reckon it's then secondly pointless switching in the first round of Intel Macs as there WILL be problems.
And then we've got a 64bit transition to do as well. It amazes me that Apple went with 32bit Intel cpus. They could have waited 6 months and missed out the 32bit problems.
Quote:
Originally posted by melgross
The simple point that I've been trying to make is that as these phases ripple through the Mac line several things will happen.
First, when Powerbooks go Intel, they will have much more powerful dual chips in them than they will have had with the PPC chip line just before. As the speed comparisons are being made with the current P M chip, they come out to about 65-75% of the performance of the current Macs's. The 7448 will add 10% improvement to that. So the final performance can be between 60-70% of the best that a PPC Powerbook will be able to to. But the Yonah will be about 25% more powerful than the current P M. It will also have dual cores. This combination should cause the Rosetta performance to increase to close to the best PPC PB times. The second half of 2006, Merom will be out with 10-15% greater performance and less power requirements.
That is just pure speculation and your figures only make sense with integer performance. You don't know how Rosetta works. I don't. Lots of people are guessing currently based on what little info we know from Transitive and it's even doubtful if Apple actually are using Transitive's method as they've been claiming it's an all Apple technology. We don't even know if it works on dual processors. Does it allow you to use both cores at once or will one be relegated to doing translation entirely? The translation is PPC -> VM -> X86 and supposedly works in place so chunks of code can be translated and left in order. You'll need a lot of RAM too. 4GB laptops here we come.
Plus that's laptops. A 2.1Ghz Yonah isn't going to emulate a 2.7Ghz Dual CPU 970fx at any comparable speed never mind a 970mp based PowerMac. 800Mhz to 1.2Ghz G3 seems much more likely which is frankly useless for anything but running Word these days.
Quote:
Originally posted by melgross
For desktops the story is the same, but takes place over a greater time period. Don't forget that the PPC Powermac will be out until end of 2007, well after Adobe says that their software will be ported over. Even the IMac G5 will be around until the end of 2006, if not longer. Adobe will be finished by then as well.
I don't see what the problem is then. You'll be able to run CS2 or 3 as long as you need with a PPC machine. When Apple is ready, so will be Adobe. The only area in which Adobe will be late is for the PB. But As I've shown, that won't be a problem. Perfomance certainly won't improve under Rosetta - that's not its purpose, but by the time it's needed fot pro apps, things should be ok.
And you DON'T have to get a Macintel until you want to.
Which puts us back to my original posit, that it may be pointless upgrading until there's native software AND more powerful CPUs than Intel can do in the next year. Rosetta and the initial Intel Macs won't be enough.
So 2007/08 is it at the earliest. Add 6-9 months for Apple/Adobe to sort out the problems in the transition and I reckon it's still an eminently sensible idea to keep buying PowerMacs for Pro work right up until the day you can't any more.
And then Apple will transition to 64bit a couple of years behind everyone else. ;-)
Not in this thread I wasn't. Not until YOU brought up laptops! We had previously discussed this in another thread and there it made sense.
And what has a Xenon got to do with it. I'm not going to be running Photoshop on an XBox! Photoshop also has enough problems on multiprocessor boxes without having to give it the burden of working on a chopped PPC 3 core CPU that only does in-order code.
Or did you mean a Xeon? which is a server class chip and costs as much for the chip as a whole iMac and usually comes in pairs and still gets beaten by the PowerMac in benchmarks.
Personally I'd take Dual Opterons anyway - and I do. I've 3 Dual Opteron servers and 2 Dual Xeons. The AMD boxes win hands down. And were quite a bit cheaper too. But that's servers, not Photoshop boxes.
Sorry. You're partly wrong.
The 970mp will still trounce the Intel desktop chips for the next year at least and all but the most expensive workstation/server chips which typically cost $1000 each when Intel release them. Intel have so far limited their dual-core chips to allow only 1 CPU in a desktop class box and the 970mp isn't. You couldn't run a dual-dual Intel Conroe, Yonah or Merom with their current policy whereas you could a Dual-dual 970mp.
Intel's mobile lineup is impressive, particularly for power requirements. A dual-core processor will go most of the way to mitigate for what we're losing in PPC and what will be taken away if we have no native software. That's provided Apple use the dual-cores of course.
I agree. But so far Rosetta is showing speeds of 25-50% at integer and 10% floating point of the P4 3.6Ghz Intel dev boxes and doesn't do AltiVec at all. This partly confirms reports that Apple's design goal was 800Mhz G3 speeds for Rosetta for the dev boxes. It feels snappier of course because most of the time your app is sitting in native core graphics calls doing nothing but you're going to hit the slow stuff as soon as you try a photoshop filter or use an airbrush.
My point, way back at the start of this thread was that switching to Intel would be pointless until the software is native and there were Intel machines significantly faster than today or even the next year.
I reckon it's then secondly pointless switching in the first round of Intel Macs as there WILL be problems.
And then we've got a 64bit transition to do as well. It amazes me that Apple went with 32bit Intel cpus. They could have waited 6 months and missed out the 32bit problems.
That is just pure speculation and your figures only make sense with integer performance. You don't know how Rosetta works. I don't. Lots of people are guessing currently based on what little info we know from Transitive and it's even doubtful if Apple actually are using Transitive's method as they've been claiming it's an all Apple technology. We don't even know if it works on dual processors. Does it allow you to use both cores at once or will one be relegated to doing translation entirely? The translation is PPC -> VM -> X86 and supposedly works in place so chunks of code can be translated and left in order. You'll need a lot of RAM too. 4GB laptops here we come.
Plus that's laptops. A 2.1Ghz Yonah isn't going to emulate a 2.7Ghz Dual CPU 970fx at any comparable speed never mind a 970mp based PowerMac. 800Mhz to 1.2Ghz G3 seems much more likely which is frankly useless for anything but running Word these days.
Which puts us back to my original posit, that it may be pointless upgrading until there's native software AND more powerful CPUs than Intel can do in the next year. Rosetta and the initial Intel Macs won't be enough.
So 2007/08 is it at the earliest. Add 6-9 months for Apple/Adobe to sort out the problems in the transition and I reckon it's still an eminently sensible idea to keep buying PowerMacs for Pro work right up until the day you can't any more.
And then Apple will transition to 64bit a couple of years behind everyone else. ;-)
You're wrong. The very second post was about laptops and G4 vs. P M dual core. You're the one that changed the discussion to G5 vs. P M, which is an invalid comparison.
Dual Xenon machines cost about what dual 2.7G5 Powermacs do. Give or take two or three hundred. PS works just fine on dual chip machines.
And following your way of this discussion. I don't care about how an iMac G5 performs. I would never use one for PS anyway if I'm using a desktop.
I'm not mentioning Opterons because they're irrelevant. Apple won't be using them, so who cares?
Don't be so sure about the 970MP. It will perform about the same as a dual 2.7 G5. And we have yet to see if Apple will use two of them in a box. I'm hoping that they put it in an Express box. I would like that to be my last PPC Powermac. But who knows?
I'm seeing much better performance for Rosetta. But don't forget that it, as well as the OS itself is in beta. These are not final products. They will most likely run faster once Apple gets down to the wire because that's when rewrites for speed are done. There are a lot of questions about Rosetta at this time. This 800 G3 thing is just speculation. I've seen estimates all over the place. Again, with everything in beta, it's not going to perform at final speed by any means.
The other thing is that I've NEVER said that people should switch before everything is ready, and the time dependent apps are native. I also tell people not to buy into a new tech right away. I always wait 'till a bit later. to give them time to get the bugs out. It's not pointless though.
Apple came out with a 32 bit box first for a very simple and obvious reason. They were working with Intel chips for at least five years. There were no 64 bit chips available. When Apple made this decision they had no time to rework the OS to 64 bits. The assumption that is being made is that Leopard will be available for Macintel, and that it will be 32/64 bit. Remember that we have about one year four months until it's out. Until then we will just have the 32 bit Mini eMac, and laptops to contend with which are not 64 bits. No loss there. None of these were going 64 bits anytime soon anyway.
Apple couldn't have waited any longer. As you point out, Adobe and others will take long enough as it is.
I'll let your fractured description of how Rosetta works go by, but we won't need 4GB RAM for it. The estimate is perhaps 50% more RAM for performance.
Again, stop comparing laptops with desktops. you might as well compare a PB to a desktop P4 3.8GHz. It's not relevant. When I edit FCP on a PB I don't expect it to perform as quickly as a dual G5 2.7. No one does. But we get our work done. It serves the purpose.
And again, I don't disagree with you about waiting, if you need the performance. My whole point is that WHEN they come out, they will outperform the PPC Macs because IBM is not putting the development into the chips that Apple needs. Now that Apple is leaving, even less resources will be devoted to that as Apple was(is) 75% of that part of IBM's PPC business. IBM was trying to drum up more support for the G5 with its "Power Everywhere" initiative. Without Apple that might be dead. In 2007 we might be very disappointed by what IBM puts out. That would make the difference even greater than it is today. Ditto for Freescale.
That's why Rosetta in its final form might be better than you think. That's all I've been saying, but you keep twisting it.
Adobe said that they would be ready by late 2006, early 2007. Plenty of time for new Mactel iMacs and Powermacs. Even a second generatoon.
I've only mentioned laptop chips because every sodding post you mention them. Apple won't be using laptop chips in pro desktops. Then you mention chips from an Xbox? WTF?
The G3 800Mhz quote came from the Apple engineer responsible for Rosetta at WWDC.
The 25-50% and 10% figures are from actual CPU bound benchmarks.
Xbench (yeah I know) scores come in slightly higher than those figures.
If I was a pro wanting top performance out of my computer I'd buy a desktop. Until there's native software, the fastest performing desktop will be a PowerMac G5 or an Intel/AMD running Windows. Rosetta isn't a viable option there. Anything less, such as slow laptop CPUs is a compromise. End of story.
Perhaps you should read some of the discussions on Ars about Rosetta. Nobody is seeing the 80% speeds Transitive were claiming and that was of just for integer performance on two similar ISAs. You don't get more dissimilar than IA32 and PPC.
"Apple came out with a 32 bit box first for a very simple and obvious reason. They were working with Intel chips for at least five years. There were no 64 bit chips available. When Apple made this decision they had no time to rework the OS to 64 bits."
btw.
They've had a 64bit CPU for quite a few years - it's called the G5.
I'm already running 64bit CentOS 4.1 and have been for months.
Windows 64bit has been out a year at least. Longer in beta. Longer still on Itanium.
They run on chip simulators before a chip is released anyway. I did it back writing native code generators for compilers on CPUs that we'd only seen the specs for not the actual silicon. I'm sure a company the size of Apple does too.
I don't think the Intel Dev kits are really about 32 or 64bit. They are more about getting people to write portable code with Cocoa and XCode. Once they've persuaded Microsoft, Adobe, Macromedia etc to switch to Cocoa development, it's much easier for them to switch the entire OS to 64bit or anything else for that matter since the application developers are now writing for a much more portable framework than carbon.
I've only mentioned laptop chips because every sodding post you mention them. Apple won't be using laptop chips in pro desktops. Then you mention chips from an Xbox? WTF?
The G3 800Mhz quote came from the Apple engineer responsible for Rosetta at WWDC.
The 25-50% and 10% figures are from actual CPU bound benchmarks.
Xbench (yeah I know) scores come in slightly higher than those figures.
If I was a pro wanting top performance out of my computer I'd buy a desktop. Until there's native software, the fastest performing desktop will be a PowerMac G5 or an Intel/AMD running Windows. Rosetta isn't a viable option there. Anything less, such as slow laptop CPUs is a compromise. End of story.
Perhaps you should read some of the discussions on Ars about Rosetta. Nobody is seeing the 80% speeds Transitive were claiming and that was of just for integer performance on two similar ISAs. You don't get more dissimilar than IA32 and PPC.
Because the discussion started out about PD's and you hijacked it. I suppose I should have ignored your post like everyone else did and we wouldn't be here now.
But as we are...
Yes. I stated as much about the desktop as well. Except that many of us use PB's because we are at a live venue where Powermacs can't be used i.e.: on location, an increasingly more important area of need with constrained budgets and time. You may not have seen the figures, but portable sales have crossed ahead of desktop sales throughout the industry including pro usage.
I read (and contribute to) almost all of the threads on ARs. I also have friends in Apple software engineering, and they are more optimistic than some of what I've been reading, which is being conservative at this early beta stage.
"Apple came out with a 32 bit box first for a very simple and obvious reason. They were working with Intel chips for at least five years. There were no 64 bit chips available. When Apple made this decision they had no time to rework the OS to 64 bits."
btw.
They've had a 64bit CPU for quite a few years - it's called the G5.
I'm already running 64bit CentOS 4.1 and have been for months.
Windows 64bit has been out a year at least. Longer in beta. Longer still on Itanium.
They run on chip simulators before a chip is released anyway. I did it back writing native code generators for compilers on CPUs that we'd only seen the specs for not the actual silicon. I'm sure a company the size of Apple does too.
I don't think the Intel Dev kits are really about 32 or 64bit. They are more about getting people to write portable code with Cocoa and XCode. Once they've persuaded Microsoft, Adobe, Macromedia etc to switch to Cocoa development, it's much easier for them to switch the entire OS to 64bit or anything else for that matter since the application developers are now writing for a much more portable framework than carbon.
Cripes! The G5 has nothing to do with this! The OS running on Intel was a 32 bit OS from the beginning, as were the chips.
Win 64 has been out less than a year. Itanium has also nothing to do with this.
Apple didn't feel rushed to do 64 bit here because they had no intention to go to Intel until a relatively short time ago. Once they did, they decided to go with the lower end machines first, which as I pointed out are all 32 bit anyway. This gives them plenty of time to work out the problems of going 64 bit on a very different architecture. No surprise there. If YOU read the ARs threads there, you would find that it's considered to be good that they are going this way. I agree with Hannibal.
Your last paragraph is correct. It's what I've also been saying.
The summary is really very simple.
If you need a portable for your work, use a PPC PB until the x86 models come out. Buy one them if you HAVE to. You will most likely not find too much of a performance difference between those first models and the discontinued PPC models.
If you don't need a portable, but need the power, stick with the PPC Powermac. You won't have a choice until sometime in mid to late 2007 anyway.
By the time an x86 Powermac comes out, Adobe and others will be ready with their Univ. bimaries.
Pretty simple. I don't know what the fuss is all about here.
Windows 64bit has been available since Windows 2003 Server was launched. What you're confusing it with is the XP x64 edition.
Linux has been 64bit since the Redhat 7.1 era.
Apple coming out with a 32bit OS after everyone has already come out with 64bit OSs despite Apple already having 64bit hardware is terribly slack IMHO. It's not like they didn't have enough warning or have 64bit hardware of their own on which to iron out the issues. Be it Intel or PPC it doesn't really matter.
If they get the tranistion right then it's possibly not that big a deal going from IA32 to 64 but those 8 extra registers and > 4GB addressing would have been very useful for the transition from PPC to Intel. Otherwise there's a lot of register shuffling to do. Switching to 64bit afterwards means they have to thunk 'old' 32bit intel apps creating issues they need not have created in the first place.
I'm sure they didn't just decide one morning to use the Intel backup plan because IBM/Freescale weren't delivering which makes the 32bit switch a little odd. Surely they knew 64bit was coming years ago.
I just find it peculiar and another reason to wait and see how things pan out rather than getting burned with early software and early hardware.
Windows 64bit has been available since Windows 2003 Server was launched. What you're confusing it with is the XP x64 edition.
Linux has been 64bit since the Redhat 7.1 era.
Apple coming out with a 32bit OS after everyone has already come out with 64bit OSs despite Apple already having 64bit hardware is terribly slack IMHO. It's not like they didn't have enough warning or have 64bit hardware of their own on which to iron out the issues. Be it Intel or PPC it doesn't really matter.
If they get the tranistion right then it's possibly not that big a deal going from IA32 to 64 but those 8 extra registers and > 4GB addressing would have been very useful for the transition from PPC to Intel. Otherwise there's a lot of register shuffling to do. Switching to 64bit afterwards means they have to thunk 'old' 32bit intel apps creating issues they need not have created in the first place.
I'm sure they didn't just decide one morning to use the Intel backup plan because IBM/Freescale weren't delivering which makes the 32bit switch a little odd. Surely they knew 64bit was coming years ago.
I just find it peculiar and another reason to wait and see how things pan out rather than getting burned with early software and early hardware.
Win 64 is what matters here.
Linux isn't important in this context.
Apparently they didn't take all that much time to make this decision.
Again, you're ignoring the 32 bit transition that will be first.
You seem to think that moving the test x86 OS to 64 bit x86 extentions is simple, a snap of the fingers. It took MS quite some time to do it, and they were already on x86.
The server OSs however were a proving ground and for Win64 it meant they could iron out the OS issues before having to hit the mass of users with hardware and legacy software compatibility issues. Windows had extreme legacy issues to contend with that Apple wouldn't have had.
And in user space on Linux there's some very cool 3D applications that went 64bit very early on and as servers and workstations the Opterons in 64bit toast Xeons.
Quote:
Originally posted by melgross
Apparently they didn't take all that much time to make this decision.
That much is becoming more obvious. ;-)
I personally think they were a little early to move to Intel but what do you do if the alternative is more Freescale chips for your laptops or late, hot IBM chips for your desktops. It's perfectly understandable on that level though I'm sure the engineers would rather have waited architecturally.
Quote:
Originally posted by melgross
Again, you're ignoring the 32 bit transition that will be first.
Nope. I was explicitly making a point of pointing out that a 32bit transition and then a 64bit transition is odd from a users and developers point of view.
Quote:
Originally posted by melgross
You seem to think that moving the test x86 OS to 64 bit x86 extentions is simple, a snap of the fingers. It took MS quite some time to do it, and they were already on x86.
Nope. I'm fully aware of the issues and I never said it was easy but shoving it out into user space means they now have both OS and Application issues to contend with instead of just the OS. And users and developers will have to contend with another architecture at some point. Perhaps not a big deal when it's all Cocoa and Core but it's another architecture to support, qa, optimize for even.
What are they going to do there? Are we going to see universal binaries with three architectures - X64, IA32 and PPC?
If they'd shifted the OS to 64bit earlier before the transition then it would have been Apple's problem, not ours also and it would be easier for Apple to switch to 64bit without having to support 32bit Intel too. The switch raised a lot of questions and I'm still waiting for the other foot to fall.
The server OSs however were a proving ground and for Win64 it meant they could iron out the OS issues before having to hit the mass of users with hardware and legacy software compatibility issues. Windows had extreme legacy issues to contend with that Apple wouldn't have had.
And in user space on Linux there's some very cool 3D applications that went 64bit very early on and as servers and workstations the Opterons in 64bit toast Xeons.
That much is becoming more obvious. ;-)
I personally think they were a little early to move to Intel but what do you do if the alternative is more Freescale chips for your laptops or late, hot IBM chips for your desktops. It's perfectly understandable on that level though I'm sure the engineers would rather have waited architecturally.
Nope. I was explicitly making a point of pointing out that a 32bit transition and then a 64bit transition is odd from a users and developers point of view.
Nope. I'm fully aware of the issues and I never said it was easy but shoving it out into user space means they now have both OS and Application issues to contend with instead of just the OS. And users and developers will have to contend with another architecture at some point. Perhaps not a big deal when it's all Cocoa and Core but it's another architecture to support, qa, optimize for even.
What are they going to do there? Are we going to see universal binaries with three architectures - X64, IA32 and PPC?
If they'd shifted the OS to 64bit earlier before the transition then it would have been Apple's problem, not ours also and it would be easier for Apple to switch to 64bit without having to support 32bit Intel too. The switch raised a lot of questions and I'm still waiting for the other foot to fall.
I'll just take the last point here.
When AMD first announced the 64 bit extentions it was far from clear that they would be accepted. Ms said that they would have Win 64 ready when AMD's chips came out. They didn't. A year and a half later, after Intel first came out with their version, Win 64 first came out.
This is a very limited distribution. You can only get it with a new machine if you request it. you can't downgrade if you find it won't work for you as there are MANY problems with firmware and drivers. It's NOT an upgrade for current users of Windows, etc.
There are very few programs in 64 bit as yet. There is very little hardware that will work with it either.
Even MS says that until Vista comes out, 64 bits is not a viable OS for most users. Apple has that window as well.
Comments
Originally posted by noirdesir
I have not said f***g otherwise. I, out of my head, said that for non-Altivec tasks the Pentium M is 50% faster than a G4. That number is naturally quite arbitrary but it certainly fits your description that the 'Pentium M beats a 7447a with no problem'.
But nobody is talking about that comparison, since there won't be any Pentium M native version of Photoshop running on a Mac for about a year. You see, I don't care how fast Photoshop runs under Windows because using Windows is simply not an option for me.
I am talking about how Photoshop runs on G4s compared to what it would run under Rosetta on a Pentium M.
That's what I'm talking about as well. Intel's new chips will be even faster. the point being that the extra power in these chips combined with the fact that they will be dual core, will allow PS even without Altivec support to have a roughly comparable performance. remember that Apple will NOT be using the P M chip, but its successors.
Even if Rosetta only works with one core (which we don't know as yet), the OS will still speed it up over a single G4 core design.
Originally posted by melgross
That's what I'm talking about as well. Intel's new chips will be even faster. the point being that the extra power in these chips combined with the fact that they will be dual core, will allow PS even without Altivec support to have a roughly comparable performance. remember that Apple will NOT be using the P M chip, but its successors.
Even if Rosetta only works with one core (which we don't know as yet), the OS will still speed it up over a single G4 core design.
If Photoshop via Rosetta on dual-core Yonahs will be as fast as on current Powerbooks, then this would require IMHO a quite dramatic speed up of dual-core Yonahs in respect to current Pentium Ms. Going dual-core might provide this dramatic speed increase, but then Windows users should be really rejoicing, since their version of Photoshop will see dramatic speed increases when they switch to dual-cores. Then again, maybe dual processor support in Windows might be so poor that they won't see this dramatic increase.
Originally posted by melgross
SSE doesn't equal Altivec. Even SSE3 which so far, I believe, only AMD supports isn't the same thing.
SSE3 can do much of what Altivec does. Altivec still has some advantages, but for a large percentage of the Altivec codebase, SSE3 will suffice.
The trouble is, relevant to this discussion, Rosetta doesn't support either, so stuff that depends on Altivec for good performance is going to run like a dog under Rosetta. Some Photoshop functions fall into this category.
Therefore, the native Intel compile will be much faster because it will use SSE3. If Rosetta is upgraded by next year to do Altivec->SSE3 translation, we're good to go. But until then, well, that's why G5 towers are getting converted over last.
Originally posted by dfiler
Carbon is the exact opposite and heralds from a time when writing for a specific platform was seen as neccessary to eek out more performance.
You're thinking of the Mac Toolbox, much of which got left behind in the change to Carbon.
Carbon began life as the cross-platform port Mac Toolbox to Windows for Quicktime for Windows.
Now, you can do things under Carbon like use WaitNextEvent() instead of Carbon Events. That kind of stuff has been discouraged for years but several big apps still do it. A good, clean Carbon app is likely to have less other cruft like inline assembly, so it should be easier to port, but that's as much good development practices as it is the application framework. People who have been lazy and 'getting by' on the backs of Mac users who suffer their legacy code are in for the most rude awakenings. Good.
Originally posted by melgross
Despite your denials, this discussion was started with a comparison between PPC Powerbooks and Mactel Powerbooks (or whatever they will be called).
It was between G4's and whatever Apple would be using instead. Rosetta performance being based on that. From the last line of the report:
"It has been reported that Apple holds one final PowerPC-based PowerBook G4 update that should debut this fall."
That's because there are 4 stories merged into the one report. The dwindling supply situation of 17" Powerbooks has nothing to do with the very first story - Adobe's Bruce Chizen saying that porting isn't easy.
The discussion got mixed but I was only ever talking about the performance of non-native software and when would be optimal to shift to Intel.
Originally posted by melgross
We aren't talking about G5's, except in the unlikely chance that Apple will use them for this last PB update. We are talking about Powerbook performance.
You may not want to talk about it. That's fine. Then limit your discussion to something else. As I said, if you want to talk about workstations, fine . But don't mix the chipsets, and compare the M Dothan to the Dual G5's. No one else is doing so.
I never brought them up. Only comparison I did was Pentium M v G5 now running native software on each OS. I'm not impressed with the Pentium M. Sorry.
Originally posted by melgross
Also, you may not keep with things, but more work is being done on PB's all the time. I know quite a few pro's from my own dealings through my company who do most of their PS work on a portable.
Yup, because they are fast "enough" as they are now for many people. People put up with a lot less performance on a laptop just to get portability. These days it's not so much a big deal slumming it on a laptop as there's more than enough for most people.
Originally posted by melgross
If you have such performance differences, I suggest that you have problems of your own with your equipment.
It was just an off the shelf comparison. I can't see why the Pentium M laptop was so slow bearing in mind the hype they get. Just wasn't impressed as I was expecting more. My iMac toasted it.
Originally posted by melgross
Again, it's a poor comparison to compare a laptop with a tower. Compare your 1.8GHz G5 with the same price PC tower and come back here. You will be rather disappointed.
I won't because for me it's not about ultimate speed. I'd have to run Windows for one so I'd get an awful lot less done and some things I couldn't do. But I don't see why I should upgrade to Intel if it's slower than I currently have at running the software I use now. That's all.
Originally posted by melgross
There is no point in discussing this any further until you understand what the discussion is about, and can make the proper comparisons.
You seem to be making lots of hardware comparisons but it's about software.
Originally posted by melgross
Laptop use by pro's is a VERY common thing. Both for PS and video editing. Maybe you're not a full time pro and so don't have to work under the same conditions. But I can tell you that many pro's work on location with laptops. More all the time. You're dismissing that is something I can't understand.
Most Pros I know have both. But it's totally irrelevant to the discussion of when Adobe will get their software native and therefore when it will be viable to upgrade for most pros.
Originally posted by melgross
With digital having almost completely taken over the professional market in catalog, fashion, and studio product photography, laptops are becoming ever more important.
If I was a jobbing photographer then yes, if I'm sat in the same office every day churning out print or web designs then no, give me a desktop any day. Bigger screens, faster hard drives, better keyboards.
The point you're consistently missing here is that I don't care what they stuff in a box to run my software, just that it doesn't run slower than it does now. If you think a dual Yonah running a G3 emulation will run faster than today's G4 laptops then great. But the meat and potatoes for me is when they've got computers (laptop or desktop) that run software faster than the G5s be that an iMac or a PowerMac. I don't see that happening till Adobe gets it's software native AND there's significantly faster CPUs than currently in the Intel roadmap.
So my bet is it'll be worthwhile upgrading to an Intel PowerMac or Intel PowerBook only when the software is native and since Bruce is saying it'll be a while, that might dovetail in with Apple having some kit that may run faster than a PowerMac G5. May as well wait for OSX 10.5 too as 10.4 is going to be a lash up at best.
It's the Quark effect all over. Nobody in print design upgraded to OSX till Quark was there or they otherwise got fed up and went to InDesign or Windows. Even now, most shops I know still keep OS9 machines around to run Quark4.11. I can see it happening again but this time it's more serious as people won't be buying hardware till it runs Adobe, Macromedia and Apple Pro apps faster than their old PPC hardware. What would be the point in upgrading otherwise? That's more dangerous for Apple than the OS9->OSX switch.
Originally posted by aegisdesign
That's because there are 4 stories merged into the one report. The dwindling supply situation of 17" Powerbooks has nothing to do with the very first story - Adobe's Bruce Chizen saying that porting isn't easy.
The discussion got mixed but I was only ever talking about the performance of non-native software and when would be optimal to shift to Intel.
I never brought them up. Only comparison I did was Pentium M v G5 now running native software on each OS. I'm not impressed with the Pentium M. Sorry.
Yup, because they are fast "enough" as they are now for many people. People put up with a lot less performance on a laptop just to get portability. These days it's not so much a big deal slumming it on a laptop as there's more than enough for most people.
It was just an off the shelf comparison. I can't see why the Pentium M laptop was so slow bearing in mind the hype they get. Just wasn't impressed as I was expecting more. My iMac toasted it.
I won't because for me it's not about ultimate speed. I'd have to run Windows for one so I'd get an awful lot less done and some things I couldn't do. But I don't see why I should upgrade to Intel if it's slower than I currently have at running the software I use now. That's all.
You seem to be making lots of hardware comparisons but it's about software.
Most Pros I know have both. But it's totally irrelevant to the discussion of when Adobe will get their software native and therefore when it will be viable to upgrade for most pros.
If I was a jobbing photographer then yes, if I'm sat in the same office every day churning out print or web designs then no, give me a desktop any day. Bigger screens, faster hard drives, better keyboards.
The point you're consistently missing here is that I don't care what they stuff in a box to run my software, just that it doesn't run slower than it does now. If you think a dual Yonah running a G3 emulation will run faster than today's G4 laptops then great. But the meat and potatoes for me is when they've got computers (laptop or desktop) that run software faster than the G5s be that an iMac or a PowerMac. I don't see that happening till Adobe gets it's software native AND there's significantly faster CPUs than currently in the Intel roadmap.
So my bet is it'll be worthwhile upgrading to an Intel PowerMac or Intel PowerBook only when the software is native and since Bruce is saying it'll be a while, that might dovetail in with Apple having some kit that may run faster than a PowerMac G5. May as well wait for OSX 10.5 too as 10.4 is going to be a lash up at best.
It's the Quark effect all over. Nobody in print design upgraded to OSX till Quark was there or they otherwise got fed up and went to InDesign or Windows. Even now, most shops I know still keep OS9 machines around to run Quark4.11. I can see it happening again but this time it's more serious as people won't be buying hardware till it runs Adobe, Macromedia and Apple Pro apps faster than their old PPC hardware. What would be the point in upgrading otherwise? That's more dangerous for Apple than the OS9->OSX switch.
Aegis, you're the one making the G5 to mobile cpu comparisons, and they're invalid. Compare the iMac to a Xenon or other machine in its class. The G5 loses. Not by too much, but it loses.
IBM has shown us nothing past what they have just announced. Intels offerings over the next year, trounce that. Over the next two years they bury it. The same for the mobile lines.
When running an app in Rosetta, that has to be accounted for. Right now, the comparisons are being made to chips that exist now. By the time Apple begins the switch, Intels chips will be more powerful than they are now, and dual designs as well.
The simple point that I've been trying to make is that as these phases ripple through the Mac line several things will happen.
First, when Powerbooks go Intel, they will have much more powerful dual chips in them than they will have had with the PPC chip line just before. As the speed comparisons are being made with the current P M chip, they come out to about 65-75% of the performance of the current Macs's. The 7448 will add 10% improvement to that. So the final performance can be between 60-70% of the best that a PPC Powerbook will be able to to. But the Yonah will be about 25% more powerful than the current P M. It will also have dual cores. This combination should cause the Rosetta performance to increase to close to the best PPC PB times. The second half of 2006, Merom will be out with 10-15% greater performance and less power requirements.
This is good news.
For desktops the story is the same, but takes place over a greater time period. Don't forget that the PPC Powermac will be out until end of 2007, well after Adobe says that their software will be ported over. Even the IMac G5 will be around until the end of 2006, if not longer. Adobe will be finished by then as well.
I don't see what the problem is then. You'll be able to run CS2 or 3 as long as you need with a PPC machine. When Apple is ready, so will be Adobe. The only area in which Adobe will be late is for the PB. But As I've shown, that won't be a problem. Perfomance certainly won't improve under Rosetta - that's not its purpose, but by the time it's needed fot pro apps, things should be ok.
And you DON'T have to get a Macintel until you want to.
Originally posted by melgross
Aegis, you're the one making the G5 to mobile cpu comparisons, and they're invalid. Compare the iMac to a Xenon or other machine in its class. The G5 loses. Not by too much, but it loses.
Not in this thread I wasn't. Not until YOU brought up laptops! We had previously discussed this in another thread and there it made sense.
And what has a Xenon got to do with it. I'm not going to be running Photoshop on an XBox! Photoshop also has enough problems on multiprocessor boxes without having to give it the burden of working on a chopped PPC 3 core CPU that only does in-order code.
Or did you mean a Xeon? which is a server class chip and costs as much for the chip as a whole iMac and usually comes in pairs and still gets beaten by the PowerMac in benchmarks.
Personally I'd take Dual Opterons anyway - and I do. I've 3 Dual Opteron servers and 2 Dual Xeons. The AMD boxes win hands down. And were quite a bit cheaper too. But that's servers, not Photoshop boxes.
Originally posted by melgross
IBM has shown us nothing past what they have just announced. Intels offerings over the next year, trounce that. Over the next two years they bury it. The same for the mobile lines.
Sorry. You're partly wrong.
The 970mp will still trounce the Intel desktop chips for the next year at least and all but the most expensive workstation/server chips which typically cost $1000 each when Intel release them. Intel have so far limited their dual-core chips to allow only 1 CPU in a desktop class box and the 970mp isn't. You couldn't run a dual-dual Intel Conroe, Yonah or Merom with their current policy whereas you could a Dual-dual 970mp.
Intel's mobile lineup is impressive, particularly for power requirements. A dual-core processor will go most of the way to mitigate for what we're losing in PPC and what will be taken away if we have no native software. That's provided Apple use the dual-cores of course.
Originally posted by melgross
When running an app in Rosetta, that has to be accounted for. Right now, the comparisons are being made to chips that exist now. By the time Apple begins the switch, Intels chips will be more powerful than they are now, and dual designs as well.
I agree. But so far Rosetta is showing speeds of 25-50% at integer and 10% floating point of the P4 3.6Ghz Intel dev boxes and doesn't do AltiVec at all. This partly confirms reports that Apple's design goal was 800Mhz G3 speeds for Rosetta for the dev boxes. It feels snappier of course because most of the time your app is sitting in native core graphics calls doing nothing but you're going to hit the slow stuff as soon as you try a photoshop filter or use an airbrush.
My point, way back at the start of this thread was that switching to Intel would be pointless until the software is native and there were Intel machines significantly faster than today or even the next year.
I reckon it's then secondly pointless switching in the first round of Intel Macs as there WILL be problems.
And then we've got a 64bit transition to do as well. It amazes me that Apple went with 32bit Intel cpus. They could have waited 6 months and missed out the 32bit problems.
Originally posted by melgross
The simple point that I've been trying to make is that as these phases ripple through the Mac line several things will happen.
First, when Powerbooks go Intel, they will have much more powerful dual chips in them than they will have had with the PPC chip line just before. As the speed comparisons are being made with the current P M chip, they come out to about 65-75% of the performance of the current Macs's. The 7448 will add 10% improvement to that. So the final performance can be between 60-70% of the best that a PPC Powerbook will be able to to. But the Yonah will be about 25% more powerful than the current P M. It will also have dual cores. This combination should cause the Rosetta performance to increase to close to the best PPC PB times. The second half of 2006, Merom will be out with 10-15% greater performance and less power requirements.
That is just pure speculation and your figures only make sense with integer performance. You don't know how Rosetta works. I don't. Lots of people are guessing currently based on what little info we know from Transitive and it's even doubtful if Apple actually are using Transitive's method as they've been claiming it's an all Apple technology. We don't even know if it works on dual processors. Does it allow you to use both cores at once or will one be relegated to doing translation entirely? The translation is PPC -> VM -> X86 and supposedly works in place so chunks of code can be translated and left in order. You'll need a lot of RAM too. 4GB laptops here we come.
Plus that's laptops. A 2.1Ghz Yonah isn't going to emulate a 2.7Ghz Dual CPU 970fx at any comparable speed never mind a 970mp based PowerMac. 800Mhz to 1.2Ghz G3 seems much more likely which is frankly useless for anything but running Word these days.
Originally posted by melgross
For desktops the story is the same, but takes place over a greater time period. Don't forget that the PPC Powermac will be out until end of 2007, well after Adobe says that their software will be ported over. Even the IMac G5 will be around until the end of 2006, if not longer. Adobe will be finished by then as well.
I don't see what the problem is then. You'll be able to run CS2 or 3 as long as you need with a PPC machine. When Apple is ready, so will be Adobe. The only area in which Adobe will be late is for the PB. But As I've shown, that won't be a problem. Perfomance certainly won't improve under Rosetta - that's not its purpose, but by the time it's needed fot pro apps, things should be ok.
And you DON'T have to get a Macintel until you want to.
Which puts us back to my original posit, that it may be pointless upgrading until there's native software AND more powerful CPUs than Intel can do in the next year. Rosetta and the initial Intel Macs won't be enough.
So 2007/08 is it at the earliest. Add 6-9 months for Apple/Adobe to sort out the problems in the transition and I reckon it's still an eminently sensible idea to keep buying PowerMacs for Pro work right up until the day you can't any more.
And then Apple will transition to 64bit a couple of years behind everyone else. ;-)
Originally posted by aegisdesign
Not in this thread I wasn't. Not until YOU brought up laptops! We had previously discussed this in another thread and there it made sense.
And what has a Xenon got to do with it. I'm not going to be running Photoshop on an XBox! Photoshop also has enough problems on multiprocessor boxes without having to give it the burden of working on a chopped PPC 3 core CPU that only does in-order code.
Or did you mean a Xeon? which is a server class chip and costs as much for the chip as a whole iMac and usually comes in pairs and still gets beaten by the PowerMac in benchmarks.
Personally I'd take Dual Opterons anyway - and I do. I've 3 Dual Opteron servers and 2 Dual Xeons. The AMD boxes win hands down. And were quite a bit cheaper too. But that's servers, not Photoshop boxes.
Sorry. You're partly wrong.
The 970mp will still trounce the Intel desktop chips for the next year at least and all but the most expensive workstation/server chips which typically cost $1000 each when Intel release them. Intel have so far limited their dual-core chips to allow only 1 CPU in a desktop class box and the 970mp isn't. You couldn't run a dual-dual Intel Conroe, Yonah or Merom with their current policy whereas you could a Dual-dual 970mp.
Intel's mobile lineup is impressive, particularly for power requirements. A dual-core processor will go most of the way to mitigate for what we're losing in PPC and what will be taken away if we have no native software. That's provided Apple use the dual-cores of course.
I agree. But so far Rosetta is showing speeds of 25-50% at integer and 10% floating point of the P4 3.6Ghz Intel dev boxes and doesn't do AltiVec at all. This partly confirms reports that Apple's design goal was 800Mhz G3 speeds for Rosetta for the dev boxes. It feels snappier of course because most of the time your app is sitting in native core graphics calls doing nothing but you're going to hit the slow stuff as soon as you try a photoshop filter or use an airbrush.
My point, way back at the start of this thread was that switching to Intel would be pointless until the software is native and there were Intel machines significantly faster than today or even the next year.
I reckon it's then secondly pointless switching in the first round of Intel Macs as there WILL be problems.
And then we've got a 64bit transition to do as well. It amazes me that Apple went with 32bit Intel cpus. They could have waited 6 months and missed out the 32bit problems.
That is just pure speculation and your figures only make sense with integer performance. You don't know how Rosetta works. I don't. Lots of people are guessing currently based on what little info we know from Transitive and it's even doubtful if Apple actually are using Transitive's method as they've been claiming it's an all Apple technology. We don't even know if it works on dual processors. Does it allow you to use both cores at once or will one be relegated to doing translation entirely? The translation is PPC -> VM -> X86 and supposedly works in place so chunks of code can be translated and left in order. You'll need a lot of RAM too. 4GB laptops here we come.
Plus that's laptops. A 2.1Ghz Yonah isn't going to emulate a 2.7Ghz Dual CPU 970fx at any comparable speed never mind a 970mp based PowerMac. 800Mhz to 1.2Ghz G3 seems much more likely which is frankly useless for anything but running Word these days.
Which puts us back to my original posit, that it may be pointless upgrading until there's native software AND more powerful CPUs than Intel can do in the next year. Rosetta and the initial Intel Macs won't be enough.
So 2007/08 is it at the earliest. Add 6-9 months for Apple/Adobe to sort out the problems in the transition and I reckon it's still an eminently sensible idea to keep buying PowerMacs for Pro work right up until the day you can't any more.
And then Apple will transition to 64bit a couple of years behind everyone else. ;-)
You're wrong. The very second post was about laptops and G4 vs. P M dual core. You're the one that changed the discussion to G5 vs. P M, which is an invalid comparison.
Dual Xenon machines cost about what dual 2.7G5 Powermacs do. Give or take two or three hundred. PS works just fine on dual chip machines.
And following your way of this discussion. I don't care about how an iMac G5 performs. I would never use one for PS anyway if I'm using a desktop.
I'm not mentioning Opterons because they're irrelevant. Apple won't be using them, so who cares?
Don't be so sure about the 970MP. It will perform about the same as a dual 2.7 G5. And we have yet to see if Apple will use two of them in a box. I'm hoping that they put it in an Express box. I would like that to be my last PPC Powermac. But who knows?
I'm seeing much better performance for Rosetta. But don't forget that it, as well as the OS itself is in beta. These are not final products. They will most likely run faster once Apple gets down to the wire because that's when rewrites for speed are done. There are a lot of questions about Rosetta at this time. This 800 G3 thing is just speculation. I've seen estimates all over the place. Again, with everything in beta, it's not going to perform at final speed by any means.
The other thing is that I've NEVER said that people should switch before everything is ready, and the time dependent apps are native. I also tell people not to buy into a new tech right away. I always wait 'till a bit later. to give them time to get the bugs out. It's not pointless though.
Apple came out with a 32 bit box first for a very simple and obvious reason. They were working with Intel chips for at least five years. There were no 64 bit chips available. When Apple made this decision they had no time to rework the OS to 64 bits. The assumption that is being made is that Leopard will be available for Macintel, and that it will be 32/64 bit. Remember that we have about one year four months until it's out. Until then we will just have the 32 bit Mini eMac, and laptops to contend with which are not 64 bits. No loss there. None of these were going 64 bits anytime soon anyway.
Apple couldn't have waited any longer. As you point out, Adobe and others will take long enough as it is.
I'll let your fractured description of how Rosetta works go by, but we won't need 4GB RAM for it. The estimate is perhaps 50% more RAM for performance.
Again, stop comparing laptops with desktops. you might as well compare a PB to a desktop P4 3.8GHz. It's not relevant. When I edit FCP on a PB I don't expect it to perform as quickly as a dual G5 2.7. No one does. But we get our work done. It serves the purpose.
And again, I don't disagree with you about waiting, if you need the performance. My whole point is that WHEN they come out, they will outperform the PPC Macs because IBM is not putting the development into the chips that Apple needs. Now that Apple is leaving, even less resources will be devoted to that as Apple was(is) 75% of that part of IBM's PPC business. IBM was trying to drum up more support for the G5 with its "Power Everywhere" initiative. Without Apple that might be dead. In 2007 we might be very disappointed by what IBM puts out. That would make the difference even greater than it is today. Ditto for Freescale.
That's why Rosetta in its final form might be better than you think. That's all I've been saying, but you keep twisting it.
Adobe said that they would be ready by late 2006, early 2007. Plenty of time for new Mactel iMacs and Powermacs. Even a second generatoon.
I've only mentioned laptop chips because every sodding post you mention them. Apple won't be using laptop chips in pro desktops. Then you mention chips from an Xbox? WTF?
The G3 800Mhz quote came from the Apple engineer responsible for Rosetta at WWDC.
The 25-50% and 10% figures are from actual CPU bound benchmarks.
Xbench (yeah I know) scores come in slightly higher than those figures.
If I was a pro wanting top performance out of my computer I'd buy a desktop. Until there's native software, the fastest performing desktop will be a PowerMac G5 or an Intel/AMD running Windows. Rosetta isn't a viable option there. Anything less, such as slow laptop CPUs is a compromise. End of story.
Perhaps you should read some of the discussions on Ars about Rosetta. Nobody is seeing the 80% speeds Transitive were claiming and that was of just for integer performance on two similar ISAs. You don't get more dissimilar than IA32 and PPC.
btw.
They've had a 64bit CPU for quite a few years - it's called the G5.
I'm already running 64bit CentOS 4.1 and have been for months.
Windows 64bit has been out a year at least. Longer in beta. Longer still on Itanium.
They run on chip simulators before a chip is released anyway. I did it back writing native code generators for compilers on CPUs that we'd only seen the specs for not the actual silicon. I'm sure a company the size of Apple does too.
I don't think the Intel Dev kits are really about 32 or 64bit. They are more about getting people to write portable code with Cocoa and XCode. Once they've persuaded Microsoft, Adobe, Macromedia etc to switch to Cocoa development, it's much easier for them to switch the entire OS to 64bit or anything else for that matter since the application developers are now writing for a much more portable framework than carbon.
Originally posted by aegisdesign
I didn't make the second post!
I've only mentioned laptop chips because every sodding post you mention them. Apple won't be using laptop chips in pro desktops. Then you mention chips from an Xbox? WTF?
The G3 800Mhz quote came from the Apple engineer responsible for Rosetta at WWDC.
The 25-50% and 10% figures are from actual CPU bound benchmarks.
Xbench (yeah I know) scores come in slightly higher than those figures.
If I was a pro wanting top performance out of my computer I'd buy a desktop. Until there's native software, the fastest performing desktop will be a PowerMac G5 or an Intel/AMD running Windows. Rosetta isn't a viable option there. Anything less, such as slow laptop CPUs is a compromise. End of story.
Perhaps you should read some of the discussions on Ars about Rosetta. Nobody is seeing the 80% speeds Transitive were claiming and that was of just for integer performance on two similar ISAs. You don't get more dissimilar than IA32 and PPC.
Because the discussion started out about PD's and you hijacked it. I suppose I should have ignored your post like everyone else did and we wouldn't be here now.
But as we are...
Yes. I stated as much about the desktop as well. Except that many of us use PB's because we are at a live venue where Powermacs can't be used i.e.: on location, an increasingly more important area of need with constrained budgets and time. You may not have seen the figures, but portable sales have crossed ahead of desktop sales throughout the industry including pro usage.
I read (and contribute to) almost all of the threads on ARs. I also have friends in Apple software engineering, and they are more optimistic than some of what I've been reading, which is being conservative at this early beta stage.
Originally posted by aegisdesign
"Apple came out with a 32 bit box first for a very simple and obvious reason. They were working with Intel chips for at least five years. There were no 64 bit chips available. When Apple made this decision they had no time to rework the OS to 64 bits."
btw.
They've had a 64bit CPU for quite a few years - it's called the G5.
I'm already running 64bit CentOS 4.1 and have been for months.
Windows 64bit has been out a year at least. Longer in beta. Longer still on Itanium.
They run on chip simulators before a chip is released anyway. I did it back writing native code generators for compilers on CPUs that we'd only seen the specs for not the actual silicon. I'm sure a company the size of Apple does too.
I don't think the Intel Dev kits are really about 32 or 64bit. They are more about getting people to write portable code with Cocoa and XCode. Once they've persuaded Microsoft, Adobe, Macromedia etc to switch to Cocoa development, it's much easier for them to switch the entire OS to 64bit or anything else for that matter since the application developers are now writing for a much more portable framework than carbon.
Cripes! The G5 has nothing to do with this! The OS running on Intel was a 32 bit OS from the beginning, as were the chips.
Win 64 has been out less than a year. Itanium has also nothing to do with this.
Apple didn't feel rushed to do 64 bit here because they had no intention to go to Intel until a relatively short time ago. Once they did, they decided to go with the lower end machines first, which as I pointed out are all 32 bit anyway. This gives them plenty of time to work out the problems of going 64 bit on a very different architecture. No surprise there. If YOU read the ARs threads there, you would find that it's considered to be good that they are going this way. I agree with Hannibal.
Your last paragraph is correct. It's what I've also been saying.
The summary is really very simple.
If you need a portable for your work, use a PPC PB until the x86 models come out. Buy one them if you HAVE to. You will most likely not find too much of a performance difference between those first models and the discontinued PPC models.
If you don't need a portable, but need the power, stick with the PPC Powermac. You won't have a choice until sometime in mid to late 2007 anyway.
By the time an x86 Powermac comes out, Adobe and others will be ready with their Univ. bimaries.
Pretty simple. I don't know what the fuss is all about here.
Threads go in all sorts of directions though so my apologies if you didn't read all but the posts you were interested in.
Originally posted by aegisdesign
Funny, I thought it was about Adobe MacTel apps, hence the title.
Threads go in all sorts of directions though so my apologies if you didn't read all but the posts you were interested in.
Actually I did, but that's fine.
Perhaps we can end this on a high note.
Linux has been 64bit since the Redhat 7.1 era.
Apple coming out with a 32bit OS after everyone has already come out with 64bit OSs despite Apple already having 64bit hardware is terribly slack IMHO. It's not like they didn't have enough warning or have 64bit hardware of their own on which to iron out the issues. Be it Intel or PPC it doesn't really matter.
If they get the tranistion right then it's possibly not that big a deal going from IA32 to 64 but those 8 extra registers and > 4GB addressing would have been very useful for the transition from PPC to Intel. Otherwise there's a lot of register shuffling to do. Switching to 64bit afterwards means they have to thunk 'old' 32bit intel apps creating issues they need not have created in the first place.
I'm sure they didn't just decide one morning to use the Intel backup plan because IBM/Freescale weren't delivering which makes the 32bit switch a little odd. Surely they knew 64bit was coming years ago.
I just find it peculiar and another reason to wait and see how things pan out rather than getting burned with early software and early hardware.
Originally posted by aegisdesign
Windows 64bit has been available since Windows 2003 Server was launched. What you're confusing it with is the XP x64 edition.
Linux has been 64bit since the Redhat 7.1 era.
Apple coming out with a 32bit OS after everyone has already come out with 64bit OSs despite Apple already having 64bit hardware is terribly slack IMHO. It's not like they didn't have enough warning or have 64bit hardware of their own on which to iron out the issues. Be it Intel or PPC it doesn't really matter.
If they get the tranistion right then it's possibly not that big a deal going from IA32 to 64 but those 8 extra registers and > 4GB addressing would have been very useful for the transition from PPC to Intel. Otherwise there's a lot of register shuffling to do. Switching to 64bit afterwards means they have to thunk 'old' 32bit intel apps creating issues they need not have created in the first place.
I'm sure they didn't just decide one morning to use the Intel backup plan because IBM/Freescale weren't delivering which makes the 32bit switch a little odd. Surely they knew 64bit was coming years ago.
I just find it peculiar and another reason to wait and see how things pan out rather than getting burned with early software and early hardware.
Win 64 is what matters here.
Linux isn't important in this context.
Apparently they didn't take all that much time to make this decision.
Again, you're ignoring the 32 bit transition that will be first.
You seem to think that moving the test x86 OS to 64 bit x86 extentions is simple, a snap of the fingers. It took MS quite some time to do it, and they were already on x86.
Originally posted by melgross
Win 64 is what matters here.
Linux isn't important in this context.
In a user context, yes.
The server OSs however were a proving ground and for Win64 it meant they could iron out the OS issues before having to hit the mass of users with hardware and legacy software compatibility issues. Windows had extreme legacy issues to contend with that Apple wouldn't have had.
And in user space on Linux there's some very cool 3D applications that went 64bit very early on and as servers and workstations the Opterons in 64bit toast Xeons.
Originally posted by melgross
Apparently they didn't take all that much time to make this decision.
That much is becoming more obvious. ;-)
I personally think they were a little early to move to Intel but what do you do if the alternative is more Freescale chips for your laptops or late, hot IBM chips for your desktops. It's perfectly understandable on that level though I'm sure the engineers would rather have waited architecturally.
Originally posted by melgross
Again, you're ignoring the 32 bit transition that will be first.
Nope. I was explicitly making a point of pointing out that a 32bit transition and then a 64bit transition is odd from a users and developers point of view.
Originally posted by melgross
You seem to think that moving the test x86 OS to 64 bit x86 extentions is simple, a snap of the fingers. It took MS quite some time to do it, and they were already on x86.
Nope. I'm fully aware of the issues and I never said it was easy but shoving it out into user space means they now have both OS and Application issues to contend with instead of just the OS. And users and developers will have to contend with another architecture at some point. Perhaps not a big deal when it's all Cocoa and Core but it's another architecture to support, qa, optimize for even.
What are they going to do there? Are we going to see universal binaries with three architectures - X64, IA32 and PPC?
If they'd shifted the OS to 64bit earlier before the transition then it would have been Apple's problem, not ours also and it would be easier for Apple to switch to 64bit without having to support 32bit Intel too. The switch raised a lot of questions and I'm still waiting for the other foot to fall.
Originally posted by aegisdesign
In a user context, yes.
The server OSs however were a proving ground and for Win64 it meant they could iron out the OS issues before having to hit the mass of users with hardware and legacy software compatibility issues. Windows had extreme legacy issues to contend with that Apple wouldn't have had.
And in user space on Linux there's some very cool 3D applications that went 64bit very early on and as servers and workstations the Opterons in 64bit toast Xeons.
That much is becoming more obvious. ;-)
I personally think they were a little early to move to Intel but what do you do if the alternative is more Freescale chips for your laptops or late, hot IBM chips for your desktops. It's perfectly understandable on that level though I'm sure the engineers would rather have waited architecturally.
Nope. I was explicitly making a point of pointing out that a 32bit transition and then a 64bit transition is odd from a users and developers point of view.
Nope. I'm fully aware of the issues and I never said it was easy but shoving it out into user space means they now have both OS and Application issues to contend with instead of just the OS. And users and developers will have to contend with another architecture at some point. Perhaps not a big deal when it's all Cocoa and Core but it's another architecture to support, qa, optimize for even.
What are they going to do there? Are we going to see universal binaries with three architectures - X64, IA32 and PPC?
If they'd shifted the OS to 64bit earlier before the transition then it would have been Apple's problem, not ours also and it would be easier for Apple to switch to 64bit without having to support 32bit Intel too. The switch raised a lot of questions and I'm still waiting for the other foot to fall.
I'll just take the last point here.
When AMD first announced the 64 bit extentions it was far from clear that they would be accepted. Ms said that they would have Win 64 ready when AMD's chips came out. They didn't. A year and a half later, after Intel first came out with their version, Win 64 first came out.
This is a very limited distribution. You can only get it with a new machine if you request it. you can't downgrade if you find it won't work for you as there are MANY problems with firmware and drivers. It's NOT an upgrade for current users of Windows, etc.
There are very few programs in 64 bit as yet. There is very little hardware that will work with it either.
Even MS says that until Vista comes out, 64 bits is not a viable OS for most users. Apple has that window as well.