Personally I adore the look of the G4 Powerbooks. I wouldn't accept much (if any) degrading of it's slimness and portability simply to stick a faster processor in there. But that's just me, maybe...
the 15" degraded already from 1" to a whopping 1.1"
Motorola have, for some time it seems, given Apple nothing but minor revisions. Why did you expect this to change now?
Last year (it is now almost 10 months from then), Motorola said they have in the works dual core G4s. However, no roadmap was publicly available at the time, nor now I think. They could also improve the performance of the system bus and FPU. It looks like nothing of these happened. The dynamically changing clock frequency is a welcome addition though.
Well, here is the fact sheet of the new Motorola chip. Compare now the Dhrystone 2.1 MIPS benchmarks of the new G4, with the ones of the new G5 in Smircle's link. Let's see:
G4: 3000 @ 1.3 GHz
G5: 7584 @ 2.0 GHz
Now, even if we assume that the G4 performance scales lineraly with frequency, the G4 would have at 2.0 GHz a score of 4615. The G5 has 64% more at that frequency. This test is not in alignment with the common belief that a G4 performs better than a G5 at the same clock speed. Of course the comparison I made includes the 90 nm G5, but the result holds even for the 130 nm G5 (this G5 has 40% higher score than the G4 at the same frequency). Finally, the absolute difference with the available frequencies (G4 @ 1.3GHz, G5 @ 2.0GHz), is 153% . Not very encouraging for the new G4.
I'm talking about on chip cache which is a more power efficient approach than off chip. Much of the extra heat production would be balanced by slowing the FSB.
Thanks
Dave
Quote:
Originally posted by Geddoe
Are you talking about a larger cache on the CPU which would mean more transistors, more power consumption, and more heat? or are you talking about an external off chip cache?
I'm still under the impression that the power disapation of the G5, even at 90nm, is very high when maximum performance is called for. I suspect that the 970 would not run long in a Powerbook at 2GHz before it throttled back. 50 watts is a lot of power to deal with, if you can't get rid of it than your only choice is to throttle the processor.
So what I'm trying to say is that maximum performance ratings of a processor mean nothing if the processor can not sustain that performance in its enclosure. Frankly we don't know exactly how IBM, or motorola for that matter, come up with their typical power usage numbers. We do know that Apple advertises that their XServes have processors using 50 watts of power. So I suspect that average usage in a powerbook at 2GHz, will be someplace between the 24.5 watts IBM says is typical and the 50 watts Apple advertises.
If you have a work load that is atypical then you are going to have problems. I suspect that Apple has already ran tests with both porcessors in knows in detail what the performance, in portable situations, is with both of these processors.
Even after all of that testing I don't believe it will make a bit of differrence which one performs better. Apple want to take leadership in the 64 bit market, they will go with the 970 as soon as they can. If we did see this G4(a) in the next rev of laptops it will only be because Apple still has issues getting the 970 to work on the platform. Skipping 64 bits in the next PowerBook rev is not something they want to do, it could be that they will have to skip for at least one revision. It is a matter of having all the ducks in a row, the new G4 looks like a drop in processor, thus allowing a cheap refresh while working on those ducks.
Dave
Quote:
Originally posted by PB
Well, here is the fact sheet of the new Motorola chip. Compare now the Dhrystone 2.1 MIPS benchmarks of the new G4, with the ones of the new G5 in Smircle's link. Let's see:
G4: 3000 @ 1.3 GHz
G5: 7584 @ 2.0 GHz
Now, even if we assume that the G4 performance scales lineraly with frequency, the G4 would have at 2.0 GHz a score of 4615. The G5 has 64% more at that frequency. This test is not in alignment with the common belief that a G4 performs better than a G5 at the same clock speed. Of course the comparison I made includes the 90 nm G5, but the result holds even for the 130 nm G5 (this G5 has 40% higher score than the G4 at the same frequency). Finally, the absolute difference with the available frequencies (G4 @ 1.3GHz, G5 @ 2.0GHz), is 153% . Not very encouraging for the new G4.
It is a matter of having all the ducks in a row, the new G4 looks like a drop in processor, thus allowing a cheap refresh while working on those ducks.
Dave
Absolutely. By the way, what became the dual core G4s Motorola talked about last year?
Yup --- I think we'll see a different PowerBook design. Some may say that's absurd (I mean, we just got the 15" G4 aluminum design), but I think it's going to have lose some of its sexiness to fit the new technology. And going back to the Pismo-level specs will not be an atrocity by any means at all. If it keeps the components going longer, then it should happen.
7457-RM was a G4 with an onboard memory controller that talked to the rest of the board via RapidIO.
The dual core processor is something else that we won't see until the 90nm node, which Motorola (and AMD, who are using Mot's process tech for 90nm) are hoping to roll out this summer. Unless it's dual-core with an onboard memory controller or Elastic Bus, I don't think Apple will be interested.
The 7447A has "contractual obligation" written all over it. Given that it's the '47 rather than the '57 (no provision for L3 cache) it's probably an iBook/eMac CPU - and maybe, if Apple gets a lock on supply before they're publicly available (which they have before) they can go into one last PowerBook refresh. But I can only see this if Apple is paying a lot less than the insane $250/unit price Mot is asking.
No, I mean the dual core G4. Motorola talked about that last year for the first time, with estimated production somewhere this year. This chip has yet to appear in any roadmap. The Register had an article about that.
No, I mean the dual core G4. Motorola talked about that last year for the first time, with estimated production somewhere this year. This chip has yet to appear in any roadmap.
It is a given that Moto has to integrate a fast RAM-controller into its CPUs before they can go dual-core, because else, the cores will be sitting around waiting for the 167Mhz bus to deliver data.
So: 7447A, then a true DDR-interface, then dual-core - if Apple is still buying chips from them, that is.
Moto is spinning off its CPU division - usually not the condition where a company reaches peak performance. OTOH, implementing a faster interface cannot be that hard. I guess, we are talking about 3 month till 7447, 9 month till RapidIO or similar interface and 12 - 18 month till dual-core.
First, before completely pooh-poohing a dual-core G4 with MaxBus, remember that PowerMacs shipped with a fairly similar arrangement (two G4s on one MaxBus) for years, and they actually did SMP pretty well.
A dual-core CPU could have much lower cache coherency penalties, since the cores wouldn't have to snoop over MaxBus the way the separate G4s did.
For a laptop it would be a pretty sweet chip. Of course, it'd be sweeter with an onboard memory controller and a higher-speed bus to RAM, but that's not necessary to see a real increase in performance.
As I've been saying for a while now, if Mot would move the memory controller on die and add an FP unit, the G4 would undergo a fairly radical transformation. Not anything to make the G5 hide in shame, but a solid performer even under load - certainly, good enough to kick the Pentium M around. Two of those in a single die and you'd have something that would own the laptop space (if not the battery-life-be-damned "portable desktop" space).
First, before completely pooh-poohing a dual-core G4 with MaxBus, remember that PowerMacs shipped with a fairly similar arrangement (two G4s on one MaxBus) for years, and they actually did SMP pretty well.
I don't think it's the dual-core/MaxBus combination that gets people all riled up; it's the fact that Motorola put it on a roadmap and started talking about it. Still if Freescale ends up successfully shipping it, we may have something along the lines of a face off, with this new dual-core, built-in memory controller über-G4, potentially going up against IBM's "G4".
Between those two chips and the new 90-nm 970 that everyone's talking about, it looks like the next few years are going to be very good indeed to Apple's low end.
First, before completely pooh-poohing a dual-core G4 with MaxBus, remember that PowerMacs shipped with a fairly similar arrangement (two G4s on one MaxBus) for years, and they actually did SMP pretty well.
I must confess this has slipped my mind completely. I was convinced that Apple had specifically designed an ASIC (UMA 2.5? UniNorth 2? - I don't really remember) with two MaxBus interfaces (for the CPUs) and one faster DDR interface (facing the rest of the system) to at least ease the bottleneck on duals.
I definitely remember Moto downgrading the SMP capabilities from MERSI to MESI with the 7450 translation - a tacit acknowledgement that with the growing divergence between CPU clockspeed and MPX bus capabilities, more than two CPUs would be a complete waste?
But of course - a dual-core chip G4 would be really neat in a Powerbook, even if it would be somewhat RAM-starved. However, this is unlikely to happen before they move their design to 90nm - it would be much too hot with the larger structures.
If Moto was further able to slap a second FP unit and a decent (RapidIO) RAM interface onto this chip, give it a deeper pipeline to boost the clock speed, it would give intel a run for their money.
Macminute has kind of a follow-up quoting Glaskowsky:
Quote:
He speculates that the new laptops could be introduced basically at any time. "I would not have been surprised to see the new PowerBook announced last month, and I won't be surprised if it doesn't come out until summer," Glaskowasky concluded.
I like the idea of "last month", but I don't believe so-called analysts have access to any information we here don't have (read: he is speculating as much as we do).
First, before completely pooh-poohing a dual-core G4 with MaxBus,
.....if Mot would move the memory controller on die and add an FP unit, the G4 would undergo a fairly radical transformation. ......
How many manufacturers currently offer dual core processors? I don't have anything against Motorola, but I would expect IBM to manufacture a dual core cpu for Apple before Motorola.
Seems like a mighty big if.
Hopefully Motorola's new alliances with SMT Microelectronics and Phillips will bear fruit, unless they've had a falling out?? It would only help Apple to have such a cpu design.
Well, as one who has been out of the portable Mac market since the debut of the PB 3400, I'd hope that my next PB would be a step up in technology from what's currently available.
Be it a unit powered by a new Motorola process, or a brand new IBM driven G5....I just want a great laptop that I can build web pages with, play games and run Office.
That said...if I had to choose, my decision would be based solely on who can provide the faster computing experience. Battery life is important, but most of my portable computing will be done near an outlet.
Comments
Originally posted by KANE
Personally I adore the look of the G4 Powerbooks. I wouldn't accept much (if any) degrading of it's slimness and portability simply to stick a faster processor in there. But that's just me, maybe...
the 15" degraded already from 1" to a whopping 1.1"
this is a bad omen.
Originally posted by KANE
Motorola have, for some time it seems, given Apple nothing but minor revisions. Why did you expect this to change now?
Last year (it is now almost 10 months from then), Motorola said they have in the works dual core G4s. However, no roadmap was publicly available at the time, nor now I think. They could also improve the performance of the system bus and FPU. It looks like nothing of these happened. The dynamically changing clock frequency is a welcome addition though.
G4: 3000 @ 1.3 GHz
G5: 7584 @ 2.0 GHz
Now, even if we assume that the G4 performance scales lineraly with frequency, the G4 would have at 2.0 GHz a score of 4615. The G5 has 64% more at that frequency. This test is not in alignment with the common belief that a G4 performs better than a G5 at the same clock speed. Of course the comparison I made includes the 90 nm G5, but the result holds even for the 130 nm G5 (this G5 has 40% higher score than the G4 at the same frequency). Finally, the absolute difference with the available frequencies (G4 @ 1.3GHz, G5 @ 2.0GHz), is 153%
Thanks
Dave
Originally posted by Geddoe
Are you talking about a larger cache on the CPU which would mean more transistors, more power consumption, and more heat? or are you talking about an external off chip cache?
So what I'm trying to say is that maximum performance ratings of a processor mean nothing if the processor can not sustain that performance in its enclosure. Frankly we don't know exactly how IBM, or motorola for that matter, come up with their typical power usage numbers. We do know that Apple advertises that their XServes have processors using 50 watts of power. So I suspect that average usage in a powerbook at 2GHz, will be someplace between the 24.5 watts IBM says is typical and the 50 watts Apple advertises.
If you have a work load that is atypical then you are going to have problems. I suspect that Apple has already ran tests with both porcessors in knows in detail what the performance, in portable situations, is with both of these processors.
Even after all of that testing I don't believe it will make a bit of differrence which one performs better. Apple want to take leadership in the 64 bit market, they will go with the 970 as soon as they can. If we did see this G4(a) in the next rev of laptops it will only be because Apple still has issues getting the 970 to work on the platform. Skipping 64 bits in the next PowerBook rev is not something they want to do, it could be that they will have to skip for at least one revision. It is a matter of having all the ducks in a row, the new G4 looks like a drop in processor, thus allowing a cheap refresh while working on those ducks.
Dave
Originally posted by PB
Well, here is the fact sheet of the new Motorola chip. Compare now the Dhrystone 2.1 MIPS benchmarks of the new G4, with the ones of the new G5 in Smircle's link. Let's see:
G4: 3000 @ 1.3 GHz
G5: 7584 @ 2.0 GHz
Now, even if we assume that the G4 performance scales lineraly with frequency, the G4 would have at 2.0 GHz a score of 4615. The G5 has 64% more at that frequency. This test is not in alignment with the common belief that a G4 performs better than a G5 at the same clock speed. Of course the comparison I made includes the 90 nm G5, but the result holds even for the 130 nm G5 (this G5 has 40% higher score than the G4 at the same frequency). Finally, the absolute difference with the available frequencies (G4 @ 1.3GHz, G5 @ 2.0GHz), is 153%
Originally posted by wizard69
It is a matter of having all the ducks in a row, the new G4 looks like a drop in processor, thus allowing a cheap refresh while working on those ducks.
Dave
Absolutely. By the way, what became the dual core G4s Motorola talked about last year?
Originally posted by PB
Absolutely. By the way, what became the dual core G4s Motorola talked about last year?
if you mean the 7457RM
moto killed the project.
The dual core processor is something else that we won't see until the 90nm node, which Motorola (and AMD, who are using Mot's process tech for 90nm) are hoping to roll out this summer. Unless it's dual-core with an onboard memory controller or Elastic Bus, I don't think Apple will be interested.
The 7447A has "contractual obligation" written all over it. Given that it's the '47 rather than the '57 (no provision for L3 cache) it's probably an iBook/eMac CPU - and maybe, if Apple gets a lock on supply before they're publicly available (which they have before) they can go into one last PowerBook refresh. But I can only see this if Apple is paying a lot less than the insane $250/unit price Mot is asking.
Originally posted by gar
if you mean the 7457RM
moto killed the project.
No, I mean the dual core G4. Motorola talked about that last year for the first time, with estimated production somewhere this year. This chip has yet to appear in any roadmap. The Register had an article about that.
Originally posted by PB
No, I mean the dual core G4. Motorola talked about that last year for the first time, with estimated production somewhere this year. This chip has yet to appear in any roadmap.
It is a given that Moto has to integrate a fast RAM-controller into its CPUs before they can go dual-core, because else, the cores will be sitting around waiting for the 167Mhz bus to deliver data.
So: 7447A, then a true DDR-interface, then dual-core - if Apple is still buying chips from them, that is.
Originally posted by Smircle
So: 7447A, then a true DDR-interface, then dual-core - if Apple is still buying chips from them, that is.
We talk about a long time here, no?
Originally posted by PB
We talk about a long time here, no?
Moto is spinning off its CPU division - usually not the condition where a company reaches peak performance. OTOH, implementing a faster interface cannot be that hard. I guess, we are talking about 3 month till 7447, 9 month till RapidIO or similar interface and 12 - 18 month till dual-core.
http://www.macdailynews.com/comments.php?id=P2216_0_1_0
I could not find it linked in any active thread.
A dual-core CPU could have much lower cache coherency penalties, since the cores wouldn't have to snoop over MaxBus the way the separate G4s did.
For a laptop it would be a pretty sweet chip. Of course, it'd be sweeter with an onboard memory controller and a higher-speed bus to RAM, but that's not necessary to see a real increase in performance.
As I've been saying for a while now, if Mot would move the memory controller on die and add an FP unit, the G4 would undergo a fairly radical transformation. Not anything to make the G5 hide in shame, but a solid performer even under load - certainly, good enough to kick the Pentium M around. Two of those in a single die and you'd have something that would own the laptop space (if not the battery-life-be-damned "portable desktop" space).
Originally posted by Amorph
First, before completely pooh-poohing a dual-core G4 with MaxBus, remember that PowerMacs shipped with a fairly similar arrangement (two G4s on one MaxBus) for years, and they actually did SMP pretty well.
I don't think it's the dual-core/MaxBus combination that gets people all riled up; it's the fact that Motorola put it on a roadmap and started talking about it. Still if Freescale ends up successfully shipping it, we may have something along the lines of a face off, with this new dual-core, built-in memory controller über-G4, potentially going up against IBM's "G4".
Between those two chips and the new 90-nm 970 that everyone's talking about, it looks like the next few years are going to be very good indeed to Apple's low end.
Originally posted by Amorph
First, before completely pooh-poohing a dual-core G4 with MaxBus, remember that PowerMacs shipped with a fairly similar arrangement (two G4s on one MaxBus) for years, and they actually did SMP pretty well.
I must confess this has slipped my mind completely. I was convinced that Apple had specifically designed an ASIC (UMA 2.5? UniNorth 2? - I don't really remember) with two MaxBus interfaces (for the CPUs) and one faster DDR interface (facing the rest of the system) to at least ease the bottleneck on duals.
I definitely remember Moto downgrading the SMP capabilities from MERSI to MESI with the 7450 translation - a tacit acknowledgement that with the growing divergence between CPU clockspeed and MPX bus capabilities, more than two CPUs would be a complete waste?
But of course - a dual-core chip G4 would be really neat in a Powerbook, even if it would be somewhat RAM-starved. However, this is unlikely to happen before they move their design to 90nm - it would be much too hot with the larger structures.
If Moto was further able to slap a second FP unit and a decent (RapidIO) RAM interface onto this chip, give it a deeper pipeline to boost the clock speed, it would give intel a run for their money.
Originally posted by oldmacfan
Did everyone catch this.
http://www.macdailynews.com/comments.php?id=P2216_0_1_0
I could not find it linked in any active thread.
Macminute has kind of a follow-up quoting Glaskowsky:
He speculates that the new laptops could be introduced basically at any time. "I would not have been surprised to see the new PowerBook announced last month, and I won't be surprised if it doesn't come out until summer," Glaskowasky concluded.
I like the idea of "last month", but I don't believe so-called analysts have access to any information we here don't have (read: he is speculating as much as we do).
Originally posted by Amorph
First, before completely pooh-poohing a dual-core G4 with MaxBus,
.....if Mot would move the memory controller on die and add an FP unit, the G4 would undergo a fairly radical transformation. ......
How many manufacturers currently offer dual core processors? I don't have anything against Motorola, but I would expect IBM to manufacture a dual core cpu for Apple before Motorola.
Seems like a mighty big if.
Hopefully Motorola's new alliances with SMT Microelectronics and Phillips will bear fruit, unless they've had a falling out?? It would only help Apple to have such a cpu design.
Be it a unit powered by a new Motorola process, or a brand new IBM driven G5....I just want a great laptop that I can build web pages with, play games and run Office.
That said...if I had to choose, my decision would be based solely on who can provide the faster computing experience. Battery life is important, but most of my portable computing will be done near an outlet.