Motorola is back
The 7447A was announced in January with an estimated delivery date of April - June. They actually delivered in April, not in July (as per my guess).
This alone is not noteworthy, but they did really well moving the G4 to the smaller 130nm process - seemingly much smoother than the IBM transition to 90nm went.
People are saying that the current iteration of the G4 is the end of the road, but I don't buy this. After having invested in the new process and the structure shrink, it is inconceivable they are not putting this investment to use. I can only see them picking up the usual development where faster CPUs come out every 6 - 9 month.
I therefore believe we will see the next G4 rev in fall/winter with speeds ranging from 1.33 Ghz to 1.67 or 1.84 Ghz and then an architectural upgrade with faster FSB and a deeper pipeline in spring (2 Ghz+).
I sure hope Apple is going to use them in iMac/eMacs and not in Powerbooks. They are unlikely to go into the iBooks, because they should be too expensive.
This alone is not noteworthy, but they did really well moving the G4 to the smaller 130nm process - seemingly much smoother than the IBM transition to 90nm went.
People are saying that the current iteration of the G4 is the end of the road, but I don't buy this. After having invested in the new process and the structure shrink, it is inconceivable they are not putting this investment to use. I can only see them picking up the usual development where faster CPUs come out every 6 - 9 month.
I therefore believe we will see the next G4 rev in fall/winter with speeds ranging from 1.33 Ghz to 1.67 or 1.84 Ghz and then an architectural upgrade with faster FSB and a deeper pipeline in spring (2 Ghz+).
I sure hope Apple is going to use them in iMac/eMacs and not in Powerbooks. They are unlikely to go into the iBooks, because they should be too expensive.
Comments
Originally posted by Smircle
The 7447A was announced in January with an estimated delivery date of April - June. They actually delivered in April, not in July (as per my guess).
This alone is not noteworthy, but they did really well moving the G4 to the smaller 130nm process - seemingly much smoother than the IBM transition to 90nm went.
Motorola isn't really "back". As you wrote yourself: They shrinked their CPU to 130nm, while IBM's problems are related to moving to 90nm. I'm not that much of an expert but I'm sure that moving from 130 to 90nm is a much more complicated affair.
Therefore Motorolas and IBMs achievements are not comparable IMHO
Originally posted by Smircle
I therefore believe we will see the next G4 rev in fall/winter with speeds ranging from 1.33 Ghz to 1.67 or 1.84 Ghz and then an architectural upgrade with faster FSB and a deeper pipeline in spring (2 Ghz+).
Hmm... we'll see about that one, but I don't think that Moto will really make it "back" within the next 12 months...
greetings,
durandal
Originally posted by Smircle
People are saying that the current iteration of the G4 is the end of the road, but I don't buy this.
Let me repost the last official Motorola roadmap (sort of) from last year's SNDF presentation. Note that there is no mention of 7447A, only 7447, no big deal though. However, they talk about some very nice integrated dual core G4 for this year. If we judge from the more recent experience (summer 2003 to now), then those dual core G4s should be taken very seriously. It is the whole Motorola taste that worries me.
Originally posted by Smircle
I can only see them picking up the usual development where faster CPUs come out every 6 - 9 month.
I therefore believe we will see the next G4 rev in fall/winter with speeds ranging from 1.33 Ghz to 1.67 or 1.84 Ghz and then an architectural upgrade with faster FSB and a deeper pipeline in spring (2 Ghz+).
I sure hope Apple is going to use them in iMac/eMacs and not in Powerbooks. They are unlikely to go into the iBooks, because they should be too expensive.
G4/1.42 GHz dual 512/120/SuperDrive/Radeon 9000 introduced 2003.01.28.13 at $2,699
Let me get this right, Moto moves from 1.42 Ghz to 1.5 Ghz in about 14 months, a whopping 80 mhz increase and you expect them to be able to maintain a development cycle that would put out faster chips every 6-9 months. What is that a 40 mhz increase every iteration, it will take them a long time to get up to 1.67 Ghz at that rate. And these chips will be more expensive than the current line of G4's? At least more expensive than those used in the iBook based on your statement. With the price of the G5 rumored to be lower than the G4 it does not make much sense to me to continued to be hampered by the G4's slow FSB and lack of direct support for DDR memory. If these chips were signifigantly lower cost and power than the G5 and offered support for industry standard memory then I could see using them, but from all I have read they are not. The best use for these is to keep Cubes alive and kicking for those devotees out there.
Originally posted by JCG
What is that a 40 mhz increase every iteration, it will take them a long time to get up to 1.67 Ghz at that rate.
.
.
.
With the price of the G5 rumored to be lower than the G4 it does not make much sense to me to continued to be hampered by the G4's slow FSB and lack of direct support for DDR memory.
You almost answered it, the G4 has potential but it is in the wrong hands right now.
Originally posted by JCG
[B]Let me get this right, Moto moves from 1.42 Ghz to 1.5 Ghz in about 14 months, a whopping 80 mhz increase and you expect them to be able to maintain a development cycle that would put out faster chips every 6-9 months.
This is not the whole story. The important point is that they have transitioned to a smaller structure size. From this on, they should have it much easier to scale the thing to higher frequencies.
If you put it this way, IBM has made no (zero, zilch) success at upping the frequency for what? 11 month or so. Obviously this means missing the bigger picture here as well.
Originally posted by Smircle
This is not the whole story. The important point is that they have transitioned to a smaller structure size. From this on, they should have it much easier to scale the thing to higher frequencies.
If you put it this way, IBM has made no (zero, zilch) success at upping the frequency for what? 11 month or so. Obviously this means missing the bigger picture here as well.
The low end Dell has a 2.4 Ghz Celeron processor with a 400 mhz FSB and 333 mhz DDR ($499). For $999 you can get a 3ghz Pentium 4 with a 800 mhz FSB and 400 mhz dual channel DDR.
As I understand it the G4 is still limited to a 167 mhz FSB and does not support a full implementation of the DDR memory that it comes with (it is through Apple's brillient engineering that they are able to use DDR at all). How many years behind is Moto on supporting these basic and relatively old technologies? For Apple not to be able to at least match the specs that a Wintel computer that sells for $499 should be an embarresment for Apple and for Motorolla. As a consumer for the cost of an iMac I woud expect that the Apple computer would have specs that more closely match the $999 Dell than the $499 one. As a current Cube owner, I see no compelling reason to buy a new computer over a processor upgrade.
Originally posted by JCG
As I understand it the G4 is still limited to a 167 mhz FSB and does not support a full implementation of the DDR memory that it comes with
I am with you here. However, most people on Mac-forums seem to care more about backlighted keyboards than performance, so a market would exist.
What I am aiming is that Moto has shrunk the chip structures for a purpose - and this is _not_ selling only one revision. Unless they completely focus on the embedded market, we will see more revisions - possibly even the dual-cored ones PB mentioned.
Mind you, if they fix the FSB bandwidth problem, add decent amounts of cache and go dual-core, I would stop bitching about how impotent the G4 is and grab my CC....
Originally posted by Smircle
The 7447A was announced in January with an estimated delivery date of April - June. They actually delivered in April, not in July (as per my guess).
This alone is not noteworthy, but they did really well moving the G4 to the smaller 130nm process - seemingly much smoother than the IBM transition to 90nm went.
Motorola's transition to 0.13µm wasn't smooth. They were over a year up to 2 years behind AMD, IBM , Intel. The MPC 7447A wasn't their trasition to 0.13µm the MPC7447/7457 was and it was nightmarish for Mac fans to observe. The only difference between the MPC 7447 and MPC7447A is the increase in L2 cache and I believe their power tune thingy, and that took months. Big Deal.
Now we wait on their transition to 0.09µm which is supposed to occur in 2004 according to the Crolles alliance. We'll see, Motorola's track record meeting any kind of reasonable schedule is suspect. Maybe STMicroelectronics and Phillips along with Motorola will be able to do a better job than Motorola alone.
Opps forgot to add that this doesn't even relate to the FSB issue, which if I understand it, the MPX bus should be able to handle 200MHz., and no higher, and won't ever use DDR technology. So as it stands Motorola hasn't even scaled the FSB to 200MHz. arrrrrrgggggghhhhhhh. Motorola #1 in embedded chip sales. So what, Apple needs computer chips not router chips.
Originally posted by msantti
I think its impressive that moto went from 1.42 GHZ to 1.5 GHZ!
if I recall correctly the 1.42 were overclocked 1.25GHz chips to desperately fill the gap until the G5 were ready to be released.
The 1.42 models were then pulled almost immediately....
I'd be willing to bet that we'll be seeing good things coming out of the G4 and Freescale in the future. Who knows, we may even see something like the mythical 7457-RM.
First, they're free of Motorola management, which was truly horrible (it was the original inspiration behind Dilbert, and the source of much of that comic's material). If their current CEO institutes such unheard-of reforms as paying senior engineers industry scale wages, or treating them like human beings, they should be able to attract the talent they need to get back in the game.
Second, they're free of Motorola fabs, which had yield rates well in excess of 99% - for dust bunnies. The fab in Crolles is as big and shiny and high-tech as Fishkill, or close enough. Motorola has partnered with some serious companies here. They're not going to be fishing CPUs out of mud puddles the way Mot employees were in Austin (OK, so I'm exaggerating
I know that I too am guilty of beating up on Motorola for this - it's gotten just a bit too easy to pick on them, hasn't it?
MaxBus. Motorola developed and delivered a way to get rid of MaxBus. It's called RapidIO. It's cheaper to implement, it's scalable, and it's much, much faster (it's like HyperTransport in many respects). Motorola has apparently had a hard time selling their customers on this tech, or they'd be deploying it in their products. About the only company that picked up a license right quick: IBM! The point is that Motorola did make an effort to get away from MaxBus, they came up with a very pretty alternative, offered it up for licensing on good terms, and they haven't got a lot of takers yet. Architosh has speculated that the final straw that killed the Mot G5 was that Apple insisted on Mot using Elastic Bus/Apple PI rather than RapidIO, and Mot refused.
Before bashing MaxBus too much, remember that it's extremely difficult to get a frontside bus clockspeed to 200MHz. This is real clockspeed, not DDR or QDR. If you look at the real clockspeeds in x86 architectures, they're almost all 100MHz, 133MHz, 167MHz or 200MHz in real terms, with clocks doubled and quadrupled to try and keep up with the CPU. But the Pentium M's "400MHz" bus (actually 100MHz quad-pumped) is not as fast or as effective as a bus actually clocked at 400MHz. But since the clock's only running at 100MHz, it's a whole lot easier to design and build. MaxBus is hard to double up, ironically, because Mot put so much effort into making it run as efficiently as possible at its native clock speed.
The G4 is getting small enough that at 90nm Freescale should be able to do all kinds of interesting things with it. Anyone for a dual-core G4 with onboard memory controller in a PowerBook? I'll take one.
Originally posted by Amorph
Anyone for a dual-core G4 with onboard memory controller in a PowerBook? I'll take one.
Me too
Originally posted by Amorph
The G4 is getting small enough that at 90nm Freescale should be able to do all kinds of interesting things with it. Anyone for a dual-core G4 with onboard memory controller in a PowerBook? I'll take one.
That seems like a nice little processor, but at what timeframe? Motorola have sort of announced that they have such a processor in the works, but when? I don't see Motorola doing 90 nm stuff this year, and they haven't done a dual core CPUs ever, and even if they seem to be doing some neat stuff with the 8500-core they have hardly delivered on their two year promise.
There are so many unaswered questions about future Frescale-offerings that we just as well be hoping for a dual core 750VX from IBM. They have 90 nm in production even if they have some problems with yield. They have delivered dual core CPUs for some time, they have Hypertransport, and they have extensive SoC experience. All they have to do is make the damn thing
You have more faith in Freescale than I do.
The killing of the alleged Motorola G5 by Apple was a rumor. The true story of the fate of Motorola's G5 was most certainly more comlex than that.
Rapid I/O was/is to be used in the MPC85XX series of processors. As far as I know they still have not shipped a single one and it is not for sale. If certain posters can be believed, the embedded market is anxiously waiting on Rapid I/O and the MPC85XX processors and it's not a market turning it down but Motorola's lack of a product. Maybe, and I stress maybe, it is available in very very limited #'s for testing and I have my doubts about that. You state Motorola has pushed Rapid I/O with no takers other than IBM. Could be because the products using it, MPC85XX are late. If I recall correctly they are now approaching 2+ years late, doesn't inspire confidence for anyone to license Rapid I/O.
After reading many posts @ Ars, my take on MPX FSB is that it is very flexible so that it can be used easily in the embedded market and it is this flexiblility/complexity that inhibits using DDR, on top of the fact that existing markets using it in the embedded space want drop in replacements and DDR would necessitate them redesigning current chips it interfaces with. That's not to say the MPX bus isn't efficient, but really a 167MHz FSB in a modern computer, no matter how efficient, is depressing.
Motorola's biggest fault in my eyes is that they didn't take a chance on Apple. They could have changed the FSB for a different series of processors for Apple. They could have increase the pipeline from 7 to 10 or more stages to ramp up clockspeed without increasing heat to unacceptable levels for a desktop computer. They didn't so that the same chip could be spread across the embedded market and Apple's computers. Although Apple bought more than half the G4's produced it became a catch 22. Apple didn't buy enough in Motorola's mind to justify the risks of a separate class of processors and Apple could bang the drums marketing a competative processor after Intel and AMD went to war.
Originally posted by Henriok
That seems like a nice little processor, but at what timeframe? Motorola have sort of announced that they have such a processor in the works, but when? I don't see Motorola doing 90 nm stuff this year, and they haven't done a dual core CPUs ever, and even if they seem to be doing some neat stuff with the 8500-core they have hardly delivered on their two year promise.
Crolles is online and running at 90nm right now. Prudently, they're not trying to make anything as complex as a CPU yet.
Freescale has said that new CPUs roll out this summer, from Crolles.
Mot hasn't done dual core, ever, but it's not rocket science. They have good on-die fabrics to pull it off, and their expertise with caches is unrivalled. They've explicitly put a dual core CPU on the roadmap - for this year, as I recall - and the G4 would be a perfect size for the job. Also, Motorola was talking about dual core G4s back when the original 7400 G4 was under development. It's been part of their plan for a long time, it's just taken a while for the technology to appear that makes it worthwhile.
There are so many unaswered questions about future Frescale-offerings that we just as well be hoping for a dual core 750VX from IBM. They have 90 nm in production even if they have some problems with yield. They have delivered dual core CPUs for some time, they have Hypertransport, and they have extensive SoC experience. All they have to do is make the damn thing
Or you can hope for both. That's the great thing about having two vendors.
Originally posted by Smircle
This alone is not noteworthy, but they did really well moving the G4 to the smaller 130nm process - seemingly much smoother than the IBM transition to 90nm went.
Motorola has been trying to go to 130 nm for several years. Don't try to rewrite history and ignore the fact that they are way behind. Everyone else did 130 nm just fine, but Motorola fell on their collective face. Everyone is having troubles with 90 nm, so IBM isn't alone (and claims to have it working now, a mere 3-4 months behind schedule).