But that's not enough -- Apple has to communicate to the customer that the result of your equation is double the performance of the 3.6 GHz x86. It is too bad that "flops" is such a bad acronym.
With the 970 it should be more reasonable for Apple to start using industry standard benchmarks like SPECmarks. The G4 was always hamstrung in those benchmarks but the 970 will apparently do much better. The use of things like MP3 and MPEG compression speeds will also be useful. Even game framerates. It is all going to be much happier with 970...
and everything that did better with a g4 (blast, some compressions) should do _even better with the new bus on the 970... correct?
Yes.
My favorite example (can you tell I'm an engineer) is SpecFP, a set of pretty heavily used science/engineering code. It's a benchmark, but it's a pretty darn useful benchmark for actual computational code.
Mot G4 1.0 GHz = 157
IBM 970 1.8 GHz = 1051
That's a _lot_ faster, even scaled to exactly the same GHz.
"Proving" that a ppc970 can beat the living daylights out of a G4 isn't going to take a team of demo-designers or marketing types a year to do. Comparing it to x86 isn't neccesarily important at first - there's probably an upper limit to how many Apple can make. Only after the Mac-types-who-are-waiting (easy sells) are satisfied would Apple really push for PC types. If it does live up to the numbers of the CPU, that might take awhile.
The really frightening thing is: Just look at what Apple's squeezing out of the G4! I walked through the web pages introducing the latest FCP and DVDSP and Shake, and you'd honestly think they were talking about something that didn't run on a single G4 on a laptop (well, Shake excepted). Apple has a hardcore mathematics group now, and they've done some really impressive work squeezing performance out of that chip.
In other words, I think it'll actually help the 970 to have been a bit late to the party. Apple's had to get really creative to get the level of performance they get out of their current machines, and they've succeeded. Once the bottlenecks come out, we'll see some real speed.
Meanwhile, MS makes every version of Windows slower and less efficient so that you'll feel compelled to buy the same machine over again, but with bigger numbers in the same places. That's not going to work in their favor, no matter what the potential performance of the hardware underneath happens to be.
"The 2 mobo of the PPC 970 will have AGP 8x and hypertransport. At the beginning of 2004, with the first update, PCI Express will make its appearance.
There will be no Powerbook PPC 970 before the second quarter 2004, when the 0,9 microns will appear. The current version @ 13 microns develop too much heat. Moreover the combination hypertransport/PPC 970 would give an insufficient autonomy."
Update: 970 PowerBook Production: Accelerated 970 Production
Further discussions with Apple OEM assembly partners have given us up to date information on motherboard production for both PowerBooks and PowerMacs, both using the IBM PPC 970 processor.
First, we have been told, and we have second-sourced the claim that 15.4-inch PowerBook motherboards are now in production and are shipping regularly to Apple's contract PowerBook assembler. Next, we have been told, and we have also second-sourced a claim that bid requests for a fully-designed 970-based board for the 17-inch PowerBook were received by two assembly plants this past Friday, with a submission deadline for replies of April 30th.
Finally, a reliable source in engineering management at the ODM supplier providing the upcoming PowerMac motherboards has informed us that those boards went into volume production this past Friday, and that first shipments to Apple's final assembly partner for the new PowerMacs is to take place "about April 15th."
We will add that our sources seem consistently taken aback by what they all characterize as the unexpected and very unusual hurry involved in all work on these new desktop and portable Apple products. Every step in each process is being scheduled far tighter than is normal for a new production run."
Let me renew my "minty pessimism": hurried assembly = bad.
It'd be wonderful if PowerMac 970s are introduced June and are ready to ship the minute after the announcement, but this rush to assembly gets me worried. The rollout for this new architecture needs to go flawlessly.
The rollout for this new architecture needs to go flawlessly.
Don't worry.
I think that the manufacturers know what they are doing. I think the PPC970 mobo will look - a lot - like the nForce2 mobo : FireWire, USB2, HyperTransport, Dual DDR400, ATA, SATA, Audio 5.1, AGP 8X
One may note the proximity of the dates for IBM's introduction and WWDC. Knowing Apple, they have probably negotiated for an exclusive rollout a few days early.
Just thinking about this, what about the possibility of a full-scale PowerMac 970 announcement at WWDC, with machines shipping shortly thereafter?
1) MWNY has effectively pulled an eWeek, ceasing to be an appropriate venue for Apple to make big announcements.
2) Apple is pushing this WWDC as a big event, and not just to developers. (home page splash for weeks?)
3) This would be a great opportunity for Apple to introduce developers to the machines and all the yummy goodness they need to know about.
4) The new machines would ship with 10.2.6, which would be entirely 32-bit (970 can run 32-bit code unmodified, right?), and Panther would be the 64-bit OS. OR, the machines would be announced, but not ship until 10.3 is released. If the 10.3 schedule has GM sometime before beginning of August, this may be more likely.
4) The new machines would ship with 10.2.6, which would be entirely 32-bit (970 can run 32-bit code unmodified, right?), and Panther would be the 64-bit OS. OR, the machines would be announced, but not ship until 10.3 is released. If the 10.3 schedule has GM sometime before beginning of August, this may be more likely.
They might well ship with 10.2.6, but that would require "necessary modifications [that] have been designed to be minimal" (quote David Wang) to the kernel. The 970 can't run 32-bit OS's without modification, but 32-bit apps and code.
I wouldn't mind a quick rollout of 970's shipping with 10.2.6, but I wouldn't mind if I have to wait until August/September for 64-bit (or just 64-bit or 970-enabled) Panther. As long as I'm getting this baby this year, I'll wait patiently
Apple can put the 970's out with 10.2.6 with a fix for the OS (allowing 32-bit OS to run on PPC970) and then be able to charge big for 10.3 (full 64-bit) in September.
" . . . There will be no Powerbook PPC 970 before the second quarter 2004, when the 0,9 microns will appear. The current version @ 13 microns develop too much heat. Moreover the combination hypertransport/PPC 970 would give an insufficient autonomy."
Just an observation that the IBM 970 can operate at lower voltage, clocking 1.2 GHz and dissipating 19 Watt. The same people who say no 970 in the 17 inch PowerBook are saying it will have a dual G4. At 10 Watts each, the dual G4s would dissipate 20 Watts. Their logic escapes me. I think there is a good chance for a "G5" 17 inch PowerBook.
Just an observation that the IBM 970 can operate at lower voltage, clocking 1.2 GHz and dissipating 19 Watt. The same people who say no 970 in the 17 inch PowerBook are saying it will have a dual G4. At 10 Watts each, the dual G4s would dissipate 20 Watts. Their logic escapes me. I think there is a good chance for a "G5" 17 inch PowerBook.
How much additional power does a PPC 970 MB (faster FSB, DDR memory, memory controller, etc.) require over a dual G4 MB?
How much additional power does a PPC 970 MB (faster FSB, DDR memory, memory controller, etc.) require over a dual G4 MB?
I don't know how much information anyone has about that. DDR SDRAM is actually cooler than SDR SDRAM, since it's made on a better process, so I wouldn't expect it to contribute much to the problem. Also, the 'book can shut down unused banks of RAM to conserve power. If Apple uses RapidIO rather than HyperTransport, they'll have an especially power-efficient means to connect the various pieces of their board together. (I don't think HT is all that bad either, actually, but I can't say for sure.)
The things you really worry about in a laptop are the CPU, the GPU, the HDD, the optical drive (especially if it's writable) and the LCD backlight. Those are your major power consumers.
I seem to recall that Apple doesn't use especially dense batteries in the current PowerBooks. They could always switch to denser (and more expensive) battery tech. to make up for any difference
The chipset in a 970 based machine will run at a higher clock rate than a G4 based machine's chipset runs. That will probably increase power/heat dissapation but no hard info is available on what that really works out to be in practice. Apple could probably build a 970-based PowerBook @ 1-1.2 GHz, but I would be surprised if arrived at the same time as the PowerMacs. Of course I wouldn't complain. Personally I think its more likely than a MP PowerBook, but that still doesn't mean I think its likely.
The rushing to production jives with the news that things are going really well at IBM's production facility. Manufacturers compress production schedules all the time, and it isn't necessarily a bad thing from a reliability point of view. The reliability issues are usually the result of design mistakes, and we don't know that Apple compressed the design/testing phases. The compression of the production schedule may also be a result of a longer design/testing cycle and since the 970 is coming out better than expected this has given them the opportunity to hurry things along.
Comments
Originally posted by Aphelion
2 x 1.8 = 3.6
But that's not enough -- Apple has to communicate to the customer that the result of your equation is double the performance of the 3.6 GHz x86. It is too bad that "flops" is such a bad acronym.
With the 970 it should be more reasonable for Apple to start using industry standard benchmarks like SPECmarks. The G4 was always hamstrung in those benchmarks but the 970 will apparently do much better. The use of things like MP3 and MPEG compression speeds will also be useful. Even game framerates. It is all going to be much happier with 970...
Originally posted by keyboardf12
and everything that did better with a g4 (blast, some compressions) should do _even better with the new bus on the 970... correct?
Yes.
My favorite example (can you tell I'm an engineer) is SpecFP, a set of pretty heavily used science/engineering code. It's a benchmark, but it's a pretty darn useful benchmark for actual computational code.
Mot G4 1.0 GHz = 157
IBM 970 1.8 GHz = 1051
That's a _lot_ faster, even scaled to exactly the same GHz.
"Proving" that a ppc970 can beat the living daylights out of a G4 isn't going to take a team of demo-designers or marketing types a year to do. Comparing it to x86 isn't neccesarily important at first - there's probably an upper limit to how many Apple can make. Only after the Mac-types-who-are-waiting (easy sells) are satisfied would Apple really push for PC types. If it does live up to the numbers of the CPU, that might take awhile.
In other words, I think it'll actually help the 970 to have been a bit late to the party. Apple's had to get really creative to get the level of performance they get out of their current machines, and they've succeeded. Once the bottlenecks come out, we'll see some real speed.
Meanwhile, MS makes every version of Windows slower and less efficient so that you'll feel compelled to buy the same machine over again, but with bigger numbers in the same places. That's not going to work in their favor, no matter what the potential performance of the hardware underneath happens to be.
"The 2 mobo of the PPC 970 will have AGP 8x and hypertransport. At the beginning of 2004, with the first update, PCI Express will make its appearance.
There will be no Powerbook PPC 970 before the second quarter 2004, when the 0,9 microns will appear. The current version @ 13 microns develop too much heat. Moreover the combination hypertransport/PPC 970 would give an insufficient autonomy."
Screed
*Cheese Eating Surrender Monkeys.
970 based motherboard production
"April 10, 2003
Update: 970 PowerBook Production: Accelerated 970 Production
Further discussions with Apple OEM assembly partners have given us up to date information on motherboard production for both PowerBooks and PowerMacs, both using the IBM PPC 970 processor.
First, we have been told, and we have second-sourced the claim that 15.4-inch PowerBook motherboards are now in production and are shipping regularly to Apple's contract PowerBook assembler. Next, we have been told, and we have also second-sourced a claim that bid requests for a fully-designed 970-based board for the 17-inch PowerBook were received by two assembly plants this past Friday, with a submission deadline for replies of April 30th.
Finally, a reliable source in engineering management at the ODM supplier providing the upcoming PowerMac motherboards has informed us that those boards went into volume production this past Friday, and that first shipments to Apple's final assembly partner for the new PowerMacs is to take place "about April 15th."
We will add that our sources seem consistently taken aback by what they all characterize as the unexpected and very unusual hurry involved in all work on these new desktop and portable Apple products. Every step in each process is being scheduled far tighter than is normal for a new production run."
It'd be wonderful if PowerMac 970s are introduced June and are ready to ship the minute after the announcement, but this rush to assembly gets me worried. The rollout for this new architecture needs to go flawlessly.
Screed
Originally posted by sCreeD
The rollout for this new architecture needs to go flawlessly.
Don't worry.
I think that the manufacturers know what they are doing. I think the PPC970 mobo will look - a lot - like the nForce2 mobo : FireWire, USB2, HyperTransport, Dual DDR400, ATA, SATA, Audio 5.1, AGP 8X
Except if Apple go with Apple PI
Originally posted by Mac The Fork
One may note the proximity of the dates for IBM's introduction and WWDC. Knowing Apple, they have probably negotiated for an exclusive rollout a few days early.
Just thinking about this, what about the possibility of a full-scale PowerMac 970 announcement at WWDC, with machines shipping shortly thereafter?
1) MWNY has effectively pulled an eWeek, ceasing to be an appropriate venue for Apple to make big announcements.
2) Apple is pushing this WWDC as a big event, and not just to developers. (home page splash for weeks?)
3) This would be a great opportunity for Apple to introduce developers to the machines and all the yummy goodness they need to know about.
4) The new machines would ship with 10.2.6, which would be entirely 32-bit (970 can run 32-bit code unmodified, right?), and Panther would be the 64-bit OS. OR, the machines would be announced, but not ship until 10.3 is released. If the 10.3 schedule has GM sometime before beginning of August, this may be more likely.
Originally posted by Alexander
4) The new machines would ship with 10.2.6, which would be entirely 32-bit (970 can run 32-bit code unmodified, right?), and Panther would be the 64-bit OS. OR, the machines would be announced, but not ship until 10.3 is released. If the 10.3 schedule has GM sometime before beginning of August, this may be more likely.
They might well ship with 10.2.6, but that would require "necessary modifications [that] have been designed to be minimal" (quote David Wang) to the kernel. The 970 can't run 32-bit OS's without modification, but 32-bit apps and code.
I wouldn't mind a quick rollout of 970's shipping with 10.2.6, but I wouldn't mind if I have to wait until August/September for 64-bit (or just 64-bit or 970-enabled) Panther. As long as I'm getting this baby this year, I'll wait patiently
They wouldn't do that would they?
Originally posted by jeromba
On Mac Bidouille today ...
" . . . There will be no Powerbook PPC 970 before the second quarter 2004, when the 0,9 microns will appear. The current version @ 13 microns develop too much heat. Moreover the combination hypertransport/PPC 970 would give an insufficient autonomy."
Just an observation that the IBM 970 can operate at lower voltage, clocking 1.2 GHz and dissipating 19 Watt. The same people who say no 970 in the 17 inch PowerBook are saying it will have a dual G4. At 10 Watts each, the dual G4s would dissipate 20 Watts. Their logic escapes me. I think there is a good chance for a "G5" 17 inch PowerBook.
Originally posted by snoopy
Their logic escapes me. I think there is a good chance for a "G5" 17 inch PowerBook.
I'm an addict of "Pro-Go-Whoa" !
Hoping it will happen a second time with the 970
Originally posted by snoopy
Just an observation that the IBM 970 can operate at lower voltage, clocking 1.2 GHz and dissipating 19 Watt. The same people who say no 970 in the 17 inch PowerBook are saying it will have a dual G4. At 10 Watts each, the dual G4s would dissipate 20 Watts. Their logic escapes me. I think there is a good chance for a "G5" 17 inch PowerBook.
How much additional power does a PPC 970 MB (faster FSB, DDR memory, memory controller, etc.) require over a dual G4 MB?
Originally posted by Bigc
How much additional power does a PPC 970 MB (faster FSB, DDR memory, memory controller, etc.) require over a dual G4 MB?
I don't know how much information anyone has about that. DDR SDRAM is actually cooler than SDR SDRAM, since it's made on a better process, so I wouldn't expect it to contribute much to the problem. Also, the 'book can shut down unused banks of RAM to conserve power. If Apple uses RapidIO rather than HyperTransport, they'll have an especially power-efficient means to connect the various pieces of their board together. (I don't think HT is all that bad either, actually, but I can't say for sure.)
The things you really worry about in a laptop are the CPU, the GPU, the HDD, the optical drive (especially if it's writable) and the LCD backlight. Those are your major power consumers.
I seem to recall that Apple doesn't use especially dense batteries in the current PowerBooks. They could always switch to denser (and more expensive) battery tech. to make up for any difference
The rushing to production jives with the news that things are going really well at IBM's production facility. Manufacturers compress production schedules all the time, and it isn't necessarily a bad thing from a reliability point of view. The reliability issues are usually the result of design mistakes, and we don't know that Apple compressed the design/testing phases. The compression of the production schedule may also be a result of a longer design/testing cycle and since the 970 is coming out better than expected this has given them the opportunity to hurry things along.