Motorola developing dual-core PowerPC G4, MPC 7447A successor
Three new PowerPC microprocessors under development at Motorola could provide Apple with additional options for forthcoming revisions its PowerBook product-line, assuming that the company is unable to deliver a PowerBook G5 by year's end.
Freescale, Motorola's soon to be spun-off chip division, is expected to debut its dual-core PowerPC G4 processor at the Microprocessor Forum in San Jose, CA, in October.
According to documents received by The Register, the new G4 will contain two PowerPC cores with AltiVec and expected to feature an on-board memory controller capable of supporting DDR 2 SDRAM, a Gigabit Ethernet controller, Rapid IO bus, and eventually run at speeds in excess of 2GHz.
Freescale's PowerPC Roadmap
The new chip will reportedly be a member of Freescale's e600 series, and the company is also working on a G5-style e700 processor that combines 32-bit and 64-bit operation.
The Register speculates that the dual core chip may pave the way for a future PowerBook G4, if Apple is unable to solve its G5 PowerBook complex in a reasonable amount of time.
Additionally, the article also confirms rumors that Motorola is working on successor to the MPC 7447A--the chip used in the most current PowerBook G4 systems--which AppleInsider sources have previously referred to as the "G4 extended" chip. This chip is rumored to reach speeds of 2G GHz, but is yet unnamed.
Freescale, Motorola's soon to be spun-off chip division, is expected to debut its dual-core PowerPC G4 processor at the Microprocessor Forum in San Jose, CA, in October.
According to documents received by The Register, the new G4 will contain two PowerPC cores with AltiVec and expected to feature an on-board memory controller capable of supporting DDR 2 SDRAM, a Gigabit Ethernet controller, Rapid IO bus, and eventually run at speeds in excess of 2GHz.
Freescale's PowerPC Roadmap
The new chip will reportedly be a member of Freescale's e600 series, and the company is also working on a G5-style e700 processor that combines 32-bit and 64-bit operation.
The Register speculates that the dual core chip may pave the way for a future PowerBook G4, if Apple is unable to solve its G5 PowerBook complex in a reasonable amount of time.
Additionally, the article also confirms rumors that Motorola is working on successor to the MPC 7447A--the chip used in the most current PowerBook G4 systems--which AppleInsider sources have previously referred to as the "G4 extended" chip. This chip is rumored to reach speeds of 2G GHz, but is yet unnamed.
Comments
Good news (if he ship in time ....)
I'll believe it when I have a dual-core PowerBook sitting on my lap without causing skin to turn black and flake off...and not a second sooner.
I really hope Freescale succeeds though. Always better to have two CPU suppliers than one. But these over-joyous posts about a product that is still considered vaporware *and* coming from a company that has gathered a bad rep as far as processor development goes is really funny.
the G4 was never able to get past the 167MHz mark. the G5 tops out at 1.25GHz bus. who cares how fast it is, if it still has a bottleneck like this.
my 2 cents.
-dornball
Dual core G4s would be great for a bridge chip until the G5 can be put into a PB and would also keep the iBook and eMac moving right along for quite a while.
If it performs it will also be a stick goading IBM into working hard on their roadmap.
Personally I believe that Apple's switch to IBM for the G5 is a very strong motivation for Freescale to perform like the old Moto hasn't for years. It should be good news over the next few years.
Originally posted by dornball
what about the bus speed?
the G4 was never able to get past the 167MHz mark. the G5 tops out at 1.25GHz bus. who cares how fast it is, if it still has a bottleneck like this.
my 2 cents.
-dornball
The article clearly states that the chip will support DDR 2 RAM, with an on-board memory controller. This is very good, and better than the 1.25 GHz bus the G5 has to offer. I just wonder if it the memory controller is single or dual channel.
Originally posted by Zapchud
The article clearly states that the chip will support DDR 2 RAM, with an on-board memory controller. This is very good, and better than the 1.25 GHz bus the G5 has to offer. I just wonder if it the memory controller is single or dual channel.
Again...on lap...or on desktop.
Originally posted by Zapchud
The article clearly states that the chip will support DDR 2 RAM, with an on-board memory controller. This is very good, and better than the 1.25 GHz bus the G5 has to offer. I just wonder if it the memory controller is single or dual channel.
wow! ok, thanks for clearing that up, Zapchud.
-dornball
Consumers wouldn't mind as the Powerbooks are G4 to this day and that's not keeping them from buying them up. I like the SoC features of the Freescale chips. I'd like to see the Powerbooks go dual core and then when the dual core G5s are ready for Powerbooks you can move the e600 chips down to the iBook/eMac.
Aug 1, 2006
Powermac- G5 Dual Cores from 2.6Ghz to 3.6Ghz
iMac- G5 Dual Core 2-2.4Ghz
Powerbook- G5 Dual Core 1.8- 2.2Ghz
iBook- e600 Dual Core 2Ghz
eMac- e600 Dual Core 1.8Ghz
Originally posted by kim kap sol
Again...on lap...or on desktop.
I'm guessing that you can't have it in a laptop in dual core version, unless you want your penis fried. We'll have to see about that, but I think this chip has good potential for lower-end desktop offerings, like the eMac and the iMac.
Hopefully, it'll be cool enough for a laptop.
I have a lot of hope for this. If this is the CPU that goes into the PowerBooks, PowerBooks will own except for the cases where single-core FPU performance is critical — and then it'll only look bad relative to 14 pound "schleptop" P4s.
Originally posted by Powerdoc
I will prefer this chip (dual core G4 with on board memory controller) to a G5.
I bet Apple won't since e600 is no 64 bit processor. Apple have pretty much equated the G5 with 64 bit processing. The e700 on the other hand..
Originally posted by Henriok
I bet Apple won't since e600 is no 64 bit processor. Apple have pretty much equated the G5 with 64 bit processing. The e700 on the other hand..
I don't know a single application turning under a laptop, that need 64 bit processing.
Marketing have his own reasons, but sometimes logic and wisdom should prevail. G3 and G4 have co-existed for years, I don't see why it will not continue with G4 and G5.
It's the case in the Intel word with the Pentium mobile chip and the prescott.
Originally posted by Amorph
... Crolles 2 fixes that).
... they're using a more sophisticated 90nm process than IBM is, which should reduce leakage current and lower overall wattage.
I have a lot of hope for this. If this is the CPU that goes into the PowerBooks, PowerBooks will own except for the cases where single-core FPU performance is critical — and then it'll only look bad relative to 14 pound "schleptop" P4s.
Even as uneducated in cpu technology as I am, a dual core G4 with an on board memory controller sounds good/great to me. I do question whether Crolles is quite up to speed yet. If they were, how come the MPC8540 and MPC8560 still aren't available for purchase? I know these aren't desktop cpu's, but they were supposed to be out a long time ago, what is it up to now, over a year late?
Could you elaborate on your statement that "they're using a more sophisticated 90nm process than IBM? How is Freescale reducing leakage current and lower overal wattage compared to IBM?
If available soon enough, I would think a 2.0+GHz G4 with a dual core and on die memory controller would be acceptable in more than just laptops. The Cossbar switch and RapidI/O Motorola has for so long described reached almost mythical status during previous discussions on this forum and over at Arestechnica. As far as floating point, if I remember correctly, the G4's floating point on a per MHz/GHz basis is actually quite good, Motorola just never kept up in the GHz department and stuck with just one floating point unit( I apologize in advance if my memory is faulty).
Anyway, I guess is, better late than never.
Originally posted by hmurchison
These are interesting options. Apple could go with a e600 based platform for the Powerbooks in 2005 and wait until 65nm before moving the G5 into portables.
...
Aug 1, 2006
Powermac- G5 Dual Cores from 2.6Ghz to 3.6Ghz
iMac- G5 Dual Core 2-2.4Ghz
Powerbook- G5 Dual Core 1.8- 2.2Ghz
iBook- e600 Dual Core 2Ghz
eMac- e600 Dual Core 1.8Ghz
Oh, that's funny. IBM (nor anyone else yet, it seems) still can't get 90nm to work reliably, and you're already expecting 65nm around the corner!
Oh, and I can't believe you have us waiting another two years before we see the G5 get to 3.6GHz? This is going to cause another round of "Oh, how apple has fallen behind" stories.
Originally posted by rickag
Even as uneducated in cpu technology as I am, a dual core G4 with an on board memory controller sounds good/great to me. I do question whether Crolles is quite up to speed yet. If they were, how come the MPC8540 and MPC8560 still aren't available for purchase? I know these aren't desktop cpu's, but they were supposed to be out a long time ago, what is it up to now, over a year late?
I don't actually know. If they're 130nm CPUs, that would explain why — 130nm was Motorola's last process in their own fab, and it was by far their most star-crossed.
They could also be delayed for other reasons. The 85xx series seems to have been something of a dud.
Could you elaborate on your statement that "they're using a more sophisticated 90nm process than IBM? How is Freescale reducing leakage current and lower overal wattage compared to IBM?
Sure. They (and AMD, who've licensed Mot's process tech) are using the same list of process technologies (copper interconnects, SOI, the so-called "Black Diamond" low-k dielectric, which will go a long way to reduce leakage) that turned the 1GHz 7455 into the 1.42GHz 7455A when they applied them to the 180nm node, and which enabled the 7447 to reach 1.5GHz at much lower wattage when Mot finally succeeded at applying them to 130nm. Freescale and AMD both have said that their 90nm chips will be made with all these process technologies. Neither IBM nor Intel applies anything so sophisticated to their 90nm fabrication; instead, they tried to get out of the gate early.
If available soon enough, I would think a 2.0+GHz G4 with a dual core and on die memory controller would be acceptable in more than just laptops. The Cossbar switch and RapidI/O Motorola has for so long described reached almost mythical status during previous discussions on this forum and over at Arestechnica. As far as floating point, if I remember correctly, the G4's floating point on a per MHz/GHz basis is actually quite good, Motorola just never kept up in the GHz department and stuck with just one floating point unit( I apologize in advance if my memory is faulty).
Yes, the G4's FPU is quite nice, and its integer capability is formidable enough to easily handle the sort of integer work that tends to accompany FP code (e.g., array indexing) so that the FPU stays busy. It's just that there's only one.
It should be a solid scalar FP contender. AltiVec code will also be able to break through the "glass ceiling" imposed by the narrow MaxBus, and one core should be able to best the G5 (handily, in some applications) clock for clock.
Note that the Register says there'll also be a follow-on single core G4, basically just a better version of what's available now. That should be great in iBooks and eMacs.
Oh, that's funny. IBM (nor anyone else yet, it seems) still can't get 90nm to work reliably, and you're already expecting 65nm around the corner!
Oh, and I can't believe you have us waiting another two years before we see the G5 get to 3.6GHz? This is going to cause another round of "Oh, how apple has fallen behind" stories.
Louzer,
IBMs problems are heat and yields. The problem is likely fixed but now that they are months behind in yielding that doesn't help Apple ship iMac G5s today but people are making far too much of an issue of 90nm than they should. IBM should be fine by Q1 2005. Parallel development should be going on right regarding 65nm. As Amorph stated above there are ways to fab at 90nm that are more sophisticated than what we're getting today. The use of strained silicon, low-k dialectics and other tech(eFuse) will play a difference in fabbing future projects.
As for the 3.6Ghz. I'm being conservative but I don't expect IBM to be much higher in two years. We'll have 2 or 4 cores at that speed but it may be a while before we get good scaling back. However keep in mind I say "G5" but Apple may call this the G6 because the processor I have in mind in 2006 at 65nm is the POWER5 derivative. I expect this CPU to to be more efficient. It should be able to queue 10 intructions and dispatch 2 per cycle(just like the POWER5) so clock for clock it'll be faster than a 970fx then we toss ondie memory controllers and SMT and we have a 3.6Ghz dual core processor that just kills.
The consumer no longer needs to care about megahertz. They need to understand architecture better. As computers evolve so must their users.