If Moto would contract out their fabs to IBM, who has proven processes I might feel better about a roadmap that they put out. But Moto dosnt just have a fab problem, it seams to me that they are yet to fully recover from the "brain drain" that I remember talked about a few years ago when Intel hired a bunch of their engineers away.
Philips and STM are not newcomers to fab tech, and for that matter neither is Motorola. Mot SPS's chief problems have been malignant, metastatic Dilbertoma, lots of excess baggage and no money. All their other ills are just symptoms.
As I mentioned, Apple can contribute their not inconsiderable brain trust to Motorola's efforts in order to produce a high performance PowerPC. Really, the G4 isn't that far from fitting that description. Replace MaxBus with an onboard memory controller and a fast packet-based bus (RapidIO or HT, or even ApplePI - the controller could be made modular via OCEAN, so that Mot could make RIO and API versions of the same processor); add another FPU. Done. It would still lose to the 970 running most compiler-output code, but it could beat the 970 in the embedded space, where code is hand-tuned (and also in any other circumstance where the code was hand-tuned - Apple is quite fond of doing that, actually).
Don't forget that IBM is incredibly expensive, and Motorola is not exactly swimming in cash right now. They've used IBM as a last-resort backup for a reason.
Quote:
What is more, they seam to have a vested interest, at least for now, in the success of the PowerPC in the server/desktop arena. Motorolla dosnt, and hasnt since Apple killed the clones. A worst case scenerio, IBM takes us through the 980, at which time Apple has a large enough adoption of OS X to have a truely "portable" OS, and development kit to automatically build FAT bionaries for multiple hardware platforms...The biggest problem that I see for Apple is keeping the momentum in the low end arena for the next 12 months....but we will see.
IBM had a considerable amount of interest in CHRP/PREP, and in OpenDoc (based, like MS' COM, on IBM's SOM technology), and in Pink, and in Taligent. I can't remember for sure, but they might have gotten OS/2 booted on the PowerPC. They even designed and published a CHRP board. I don't think that at this point anyone is still aching over the cancellation of the clones. AIM is dead as a formal alliance, but it seems to be doing well as an informal confederacy.
Mot should stick around for three simple reasons: Their design approach is different than IBM's, and thus the two combined offer Apple more options than either one could; Apple is a huge and steady and needed customer for Mot SPS, which, combined with SPS's brain drain, gives Apple a great deal of leverage and influence in CPU design; and finally, two vendors are better than one.
As I've been saying for a while now, Mot is trying to turn around, and they've been showing results. They've got an alliance together that obviates their fab problems and gives them a clear path to 90nm and smaller processes. I'm with Programmer in betting that the next few years won't be like the last 5 for Motorola SPS, and believe me, I'm skeptical of Motorola myself. But we'll see soon enough.
It'd be cool to see IBM and Mot in a war over Apple. IBM would probably crush Mot like a bug, but it'd be nice to see where Mot gets before the bug crushing.
They could, however, easily take the memory controller and RIO interface from their 8xxx series host processors and mate them to the 7457 on a 9nm process, and that would result in a fast low power processor.
I like it, its just that they cannot even fab 130nm 7457's with any real success?, I agree this would be ideal in several of apple's consumer products and good for apple/us to have vendor options. I suspect, however, that the horse has already bolted. Firstly, the 90nm 970 will be out by years end (approx), that would satisfy all of apples products except for the ibook which currently uses a <10W 750.
The G3 chip will most likely scale up ok, for now and until apple desides on when to use the 970+ in the imacs/emacs/pb's we expect G4's to continue. How Moto intends to maintain those orders, i dont know, but certainly looks shaky from where im sitting, despite amorphs very valid points.
I like it, its just that they cannot even fab 130nm 7457's with any real success?
Source? So far no problems have been reported besides some doubtfull rumors. Robert from Powerlogix seems to be very pleased with the 7457 and is confident that Mot will deliver. (http://www.cubeowner.com/forums//ind...19&t=3011&st=0)
Kickaha and Amorph couldn't moderate themselves out of a paper bag. Abdicate responsibility and succumb to idiocy. Two years of letting a member make personal attacks against others, then stepping aside when someone won't put up with it. Not only that but go ahead and shut down my posting priviledges but not the one making the attacks. Not even the common decency to abide by their warning (afer three days of absorbing personal attacks with no mods in sight), just shut my posting down and then say it might happen later if a certian line is crossed. Bullshit flag is flying, I won't abide by lying and coddling of liars who go off-site, create accounts differing in a single letter from my handle with the express purpose to decieve and then claim here that I did it. Everyone be warned, kim kap sol is a lying, deceitful poster.
Now I guess they should have banned me rather than just shut off posting priviledges, because kickaha and Amorph definitely aren't going to like being called to task when they thought they had it all ignored *cough* *cough* I mean under control. Just a couple o' tools.
Don't worry, as soon as my work resetting my posts is done I'll disappear forever.
I was refering the 7457 time line. Motorola isn't the only company that's having problems with 0.13, TSMC was struggling aswell. Major fault was to skip 0.15 altogether, this caused a big gap as Mot didn't see risk in the big jump.
Pushing the 7455 to 1.42GHz on 0.18 is quite an achievement which aleast raised my hope for faster progression once 0.13 is up and running. Of course we are talking about a firm which missed many timelines, but judging from the past I'd say if small companies like Powerlogix has already got samples for some time now, release shouldn't be far away.
While its true that they lost many engineers, they didn't lose them all. We don't seem to be disagreeing here -- I'm not predicting that they're going to suddenly come out with a G5 class processor. They could, however, easily take the memory controller and RIO interface from their 8xxx series host processors and mate them to the 7457 on a 9nm process, and that would result in a fast low power processor.
I wonder, if you build a machine that looks like this
[RAM] == [CPU] --rio-- [core logic] -- AGP
or
[RAM] == [CPU] --HT-- [AGP-HT bridge] -- AGP
(where the RAM controller is built onto the processor)
The RAM is now kinda far away from the AGP side of the machine. Quartz Extreme could cause flooding of the CPU with AGP DMA requests to read the RAM contents as windows are moved and refreshed.. probably not good for the poor CPU trying to get its own work done.
Of course if the window server evolves so that it can keep window backing stores in VRAM then that issue might fade. But now you have a need for a lot of VRAM which makes it tougher to get into those low end machines or laptops.
Kickaha and Amorph couldn't moderate themselves out of a paper bag. Abdicate responsibility and succumb to idiocy. Two years of letting a member make personal attacks against others, then stepping aside when someone won't put up with it. Not only that but go ahead and shut down my posting priviledges but not the one making the attacks. Not even the common decency to abide by their warning (afer three days of absorbing personal attacks with no mods in sight), just shut my posting down and then say it might happen later if a certian line is crossed. Bullshit flag is flying, I won't abide by lying and coddling of liars who go off-site, create accounts differing in a single letter from my handle with the express purpose to decieve and then claim here that I did it. Everyone be warned, kim kap sol is a lying, deceitful poster.
Now I guess they should have banned me rather than just shut off posting priviledges, because kickaha and Amorph definitely aren't going to like being called to task when they thought they had it all ignored *cough* *cough* I mean under control. Just a couple o' tools.
Don't worry, as soon as my work resetting my posts is done I'll disappear forever.
The RAM is now kinda far away from the AGP side of the machine. Quartz Extreme could cause flooding of the CPU with AGP DMA requests to read the RAM contents as windows are moved and refreshed.. probably not good for the poor CPU trying to get its own work done.
This isn't as much of an issue as you might think. AGP/GPU transactions tend to be very heavily pipelined and designed for long latencies. AGP 8x only uses about 2 GB/sec of bandwidth, so a decent memory setup (DDR333 or better) can support that and still keep the CPU fed. The CPU tends to want lower-latency access to memory so this is actually better arrangement. And just because the CPU is on the same chip as the memory controller doesn't mean that the CPU's operation is in any way interfered with by external memory accesses (aside from the bandwidth consumption which the CPU mitigates via its L1/L2 caches).
Not a sure thing at all, just look at the G4 intro debacle. Plenty of good performing samples (out of how many 100's of wafers we may never know), enough to do all the R&D for the 500 Mhz G4 introduction. Then Mot was unable to fab 500Mhz units for several months and forced Apple to do the infamous "Downgrade". Onsies and twosies are meaningless when thousands are the enrtering argumant for a viable product.
Yes, but my point above was that it is no longer Motorola doing the fabbing. It'll be somebody that actually knows how to run a fab.
They could, however, easily take the memory controller and RIO interface from their 8xxx series host processors and mate them to the 7457 on a 9nm process, and that would result in a fast low power processor.
If they fab a 7457 on a 9nm process they'll leave the entire industry in the dust.
Not a sure thing at all, just look at the G4 intro debacle. Plenty of good performing samples (out of how many 100's of wafers we may never know), enough to do all the R&D for the 500 Mhz G4 introduction. Then Mot was unable to fab 500Mhz units for several months and forced Apple to do the infamous "Downgrade". Onsies and twosies are meaningless when thousands are the enrtering argumant for a viable product.
One profound difference between Moto then and IBM now is that IBM wants to get into the chip fab business. It is in IBM's interests to show that they can do die shrinks, get production up, etc. IBM is motivated because if they do well, then they stand to win chip outsourcing contracts from TSMC and other Taiwanese fabbing companies. IBM wants to compete, and they want to get to 90 nm ASAP. This is in direct contrast to Moto which obviously wants to strangle its semiconductor division and pretend that they never made chips.
IBM wants to compete, and they want to get to 90 nm ASAP. This is in direct contrast to Moto.
Umm, Moto wants to get to 90nm so badly that they were talking openly late last year about skipping the 130nm node altogether, and they partnered with several other companies on a 90nm fab (Crolles).
Granted, the motivation is different, because Mot SPS is just trying to avoid getting axed as unprofitable. But companies (and company divisions) can be dangerous when they're up against a wall...
I don't believe that you can seperate chip design from fab so easily.
From what I understand, chips need to be designed for manufacturability.
Just saying that Mot's design may actually have something to do with their fab problems, so having someone else fab them isn't necessarily a magic cure-all.
I don't believe that you can seperate chip design from fab so easily.
From what I understand, chips need to be designed for manufacturability.
Just saying that Mot's design may actually have something to do with their fab problems, so having someone else fab them isn't necessarily a magic cure-all.
I don't believe that you can seperate chip design from fab so easily.
From what I understand, chips need to be designed for manufacturability.
The great bulk of them, however, are not designed well enough to be manufactured amid dustbunnies. Mot actually shut off the fans in their fabs to save money!
You do have a point in that Mot has made some gambles with process technology that didn't pay off, and they're currently struggling to get that tech working on 130nm. So it's not an either/or thing.
Quote:
Just saying that Mot's design may actually have something to do with their fab problems, so having someone else fab them isn't necessarily a magic cure-all.
As I understand it, the original 500MHz ceiling was imposed by a design flaw in the L1 cache, and the 7450 bowed in with a laundry list of errata, so this is certainly possible. But there are more mundane reasons for Mot's inability to fabricate CPUs in quantity.
Kickaha and Amorph couldn't moderate themselves out of a paper bag. Abdicate responsibility and succumb to idiocy. Two years of letting a member make personal attacks against others, then stepping aside when someone won't put up with it. Not only that but go ahead and shut down my posting priviledges but not the one making the attacks. Not even the common decency to abide by their warning (afer three days of absorbing personal attacks with no mods in sight), just shut my posting down and then say it might happen later if a certian line is crossed. Bullshit flag is flying, I won't abide by lying and coddling of liars who go off-site, create accounts differing in a single letter from my handle with the express purpose to decieve and then claim here that I did it. Everyone be warned, kim kap sol is a lying, deceitful poster.
Now I guess they should have banned me rather than just shut off posting priviledges, because kickaha and Amorph definitely aren't going to like being called to task when they thought they had it all ignored *cough* *cough* I mean under control. Just a couple o' tools.
Don't worry, as soon as my work resetting my posts is done I'll disappear forever.
Source? So far no problems have been reported besides some doubtfull rumors. Robert from Powerlogix seems to be very pleased with the 7457 and is confident that Mot will deliver. (http://www.cubeowner.com/forums//ind...19&t=3011&st=0)
End of Line
Source?, Sure! apple.com, can you find an apple product that is running >1GHz, that is not in PMac?. And how long has apple gleefuly remained at 1GHZ for the majority of its products. Oh yeah, Moto's doing a great job, everyone give them a round of applause!
Future-wise though, (and back in synch with your sub-thread) even with the fab at Crolles, I don't know what the chances are for Mot's recovery in PPC production. My impression was much of that fab was dedicated to joint chip and process R&D pursuits in the telecom sector, a bit of a stretch to use it as a life-ring for one partners major product line, although anythings possible.
Maybe Moto will just design newer G4's and outsource production to IBM's Fishkill plant. No, they're probably too stupid to do something sensible like that.
I'm really looking forward to the day when Steve gets up on stage and announces that Moto is no more, and then cusses them out (probably not going to happen either).
Comments
Originally posted by @homenow
If Moto would contract out their fabs to IBM, who has proven processes I might feel better about a roadmap that they put out. But Moto dosnt just have a fab problem, it seams to me that they are yet to fully recover from the "brain drain" that I remember talked about a few years ago when Intel hired a bunch of their engineers away.
Philips and STM are not newcomers to fab tech, and for that matter neither is Motorola. Mot SPS's chief problems have been malignant, metastatic Dilbertoma, lots of excess baggage and no money. All their other ills are just symptoms.
As I mentioned, Apple can contribute their not inconsiderable brain trust to Motorola's efforts in order to produce a high performance PowerPC. Really, the G4 isn't that far from fitting that description. Replace MaxBus with an onboard memory controller and a fast packet-based bus (RapidIO or HT, or even ApplePI - the controller could be made modular via OCEAN, so that Mot could make RIO and API versions of the same processor); add another FPU. Done. It would still lose to the 970 running most compiler-output code, but it could beat the 970 in the embedded space, where code is hand-tuned (and also in any other circumstance where the code was hand-tuned - Apple is quite fond of doing that, actually).
Don't forget that IBM is incredibly expensive, and Motorola is not exactly swimming in cash right now. They've used IBM as a last-resort backup for a reason.
What is more, they seam to have a vested interest, at least for now, in the success of the PowerPC in the server/desktop arena. Motorolla dosnt, and hasnt since Apple killed the clones. A worst case scenerio, IBM takes us through the 980, at which time Apple has a large enough adoption of OS X to have a truely "portable" OS, and development kit to automatically build FAT bionaries for multiple hardware platforms...The biggest problem that I see for Apple is keeping the momentum in the low end arena for the next 12 months....but we will see.
IBM had a considerable amount of interest in CHRP/PREP, and in OpenDoc (based, like MS' COM, on IBM's SOM technology), and in Pink, and in Taligent. I can't remember for sure, but they might have gotten OS/2 booted on the PowerPC. They even designed and published a CHRP board. I don't think that at this point anyone is still aching over the cancellation of the clones. AIM is dead as a formal alliance, but it seems to be doing well as an informal confederacy.
Mot should stick around for three simple reasons: Their design approach is different than IBM's, and thus the two combined offer Apple more options than either one could; Apple is a huge and steady and needed customer for Mot SPS, which, combined with SPS's brain drain, gives Apple a great deal of leverage and influence in CPU design; and finally, two vendors are better than one.
As I've been saying for a while now, Mot is trying to turn around, and they've been showing results. They've got an alliance together that obviates their fab problems and gives them a clear path to 90nm and smaller processes. I'm with Programmer in betting that the next few years won't be like the last 5 for Motorola SPS, and believe me, I'm skeptical of Motorola myself. But we'll see soon enough.
Originally posted by Programmer
They could, however, easily take the memory controller and RIO interface from their 8xxx series host processors and mate them to the 7457 on a 9nm process, and that would result in a fast low power processor.
I like it, its just that they cannot even fab 130nm 7457's with any real success?, I agree this would be ideal in several of apple's consumer products and good for apple/us to have vendor options. I suspect, however, that the horse has already bolted. Firstly, the 90nm 970 will be out by years end (approx), that would satisfy all of apples products except for the ibook which currently uses a <10W 750.
The G3 chip will most likely scale up ok, for now and until apple desides on when to use the 970+ in the imacs/emacs/pb's we expect G4's to continue. How Moto intends to maintain those orders, i dont know, but certainly looks shaky from where im sitting, despite amorphs very valid points.
Originally posted by hasapi
I like it, its just that they cannot even fab 130nm 7457's with any real success?
Source? So far no problems have been reported besides some doubtfull rumors. Robert from Powerlogix seems to be very pleased with the 7457 and is confident that Mot will deliver. (http://www.cubeowner.com/forums//ind...19&t=3011&st=0)
End of Line
Originally posted by User Tron
Source? So far no problems have been reported besides some doubtfull rumors.
From Chip Architect.com in Nov 2000, MacCentral in Apr 2001, The Register in Apr 2001. It states that AMD/Motorola will bring HiP7 into service in 2001. AMD did i Dresden, but Motorola haven't shipped ANY production processor using HiP7 as far as i know.
Now I guess they should have banned me rather than just shut off posting priviledges, because kickaha and Amorph definitely aren't going to like being called to task when they thought they had it all ignored *cough* *cough* I mean under control. Just a couple o' tools.
Don't worry, as soon as my work resetting my posts is done I'll disappear forever.
Originally posted by Henriok
From Chip Architect.com in Nov 2000, MacCentral in Apr 2001, The Register in Apr 2001. It states that AMD/Motorola will bring HiP7 into service in 2001. AMD did i Dresden, but Motorola haven't shipped ANY production processor using HiP7 as far as i know.
I was refering the 7457 time line. Motorola isn't the only company that's having problems with 0.13, TSMC was struggling aswell. Major fault was to skip 0.15 altogether, this caused a big gap as Mot didn't see risk in the big jump.
Pushing the 7455 to 1.42GHz on 0.18 is quite an achievement which aleast raised my hope for faster progression once 0.13 is up and running. Of course we are talking about a firm which missed many timelines, but judging from the past I'd say if small companies like Powerlogix has already got samples for some time now, release shouldn't be far away.
End of Line
Originally posted by Programmer
While its true that they lost many engineers, they didn't lose them all. We don't seem to be disagreeing here -- I'm not predicting that they're going to suddenly come out with a G5 class processor. They could, however, easily take the memory controller and RIO interface from their 8xxx series host processors and mate them to the 7457 on a 9nm process, and that would result in a fast low power processor.
I wonder, if you build a machine that looks like this
[RAM] == [CPU] --rio-- [core logic] -- AGP
or
[RAM] == [CPU] --HT-- [AGP-HT bridge] -- AGP
(where the RAM controller is built onto the processor)
The RAM is now kinda far away from the AGP side of the machine. Quartz Extreme could cause flooding of the CPU with AGP DMA requests to read the RAM contents as windows are moved and refreshed.. probably not good for the poor CPU trying to get its own work done.
Of course if the window server evolves so that it can keep window backing stores in VRAM then that issue might fade. But now you have a need for a lot of VRAM which makes it tougher to get into those low end machines or laptops.
Now I guess they should have banned me rather than just shut off posting priviledges, because kickaha and Amorph definitely aren't going to like being called to task when they thought they had it all ignored *cough* *cough* I mean under control. Just a couple o' tools.
Don't worry, as soon as my work resetting my posts is done I'll disappear forever.
Originally posted by boots
The RAM is now kinda far away from the AGP side of the machine. Quartz Extreme could cause flooding of the CPU with AGP DMA requests to read the RAM contents as windows are moved and refreshed.. probably not good for the poor CPU trying to get its own work done.
This isn't as much of an issue as you might think. AGP/GPU transactions tend to be very heavily pipelined and designed for long latencies. AGP 8x only uses about 2 GB/sec of bandwidth, so a decent memory setup (DDR333 or better) can support that and still keep the CPU fed. The CPU tends to want lower-latency access to memory so this is actually better arrangement. And just because the CPU is on the same chip as the memory controller doesn't mean that the CPU's operation is in any way interfered with by external memory accesses (aside from the bandwidth consumption which the CPU mitigates via its L1/L2 caches).
Originally posted by AirSluf
Not a sure thing at all, just look at the G4 intro debacle. Plenty of good performing samples (out of how many 100's of wafers we may never know), enough to do all the R&D for the 500 Mhz G4 introduction. Then Mot was unable to fab 500Mhz units for several months and forced Apple to do the infamous "Downgrade". Onsies and twosies are meaningless when thousands are the enrtering argumant for a viable product.
Yes, but my point above was that it is no longer Motorola doing the fabbing. It'll be somebody that actually knows how to run a fab.
Originally posted by Programmer
They could, however, easily take the memory controller and RIO interface from their 8xxx series host processors and mate them to the 7457 on a 9nm process, and that would result in a fast low power processor.
If they fab a 7457 on a 9nm process they'll leave the entire industry in the dust.
Eight processor iBooks, here we come!
Originally posted by AirSluf
Not a sure thing at all, just look at the G4 intro debacle. Plenty of good performing samples (out of how many 100's of wafers we may never know), enough to do all the R&D for the 500 Mhz G4 introduction. Then Mot was unable to fab 500Mhz units for several months and forced Apple to do the infamous "Downgrade". Onsies and twosies are meaningless when thousands are the enrtering argumant for a viable product.
One profound difference between Moto then and IBM now is that IBM wants to get into the chip fab business. It is in IBM's interests to show that they can do die shrinks, get production up, etc. IBM is motivated because if they do well, then they stand to win chip outsourcing contracts from TSMC and other Taiwanese fabbing companies. IBM wants to compete, and they want to get to 90 nm ASAP. This is in direct contrast to Moto which obviously wants to strangle its semiconductor division and pretend that they never made chips.
Originally posted by Yevgeny
IBM wants to compete, and they want to get to 90 nm ASAP. This is in direct contrast to Moto.
Umm, Moto wants to get to 90nm so badly that they were talking openly late last year about skipping the 130nm node altogether, and they partnered with several other companies on a 90nm fab (Crolles).
Granted, the motivation is different, because Mot SPS is just trying to avoid getting axed as unprofitable. But companies (and company divisions) can be dangerous when they're up against a wall...
From what I understand, chips need to be designed for manufacturability.
Just saying that Mot's design may actually have something to do with their fab problems, so having someone else fab them isn't necessarily a magic cure-all.
Originally posted by Transcendental Octothorpe
I don't believe that you can seperate chip design from fab so easily.
From what I understand, chips need to be designed for manufacturability.
Just saying that Mot's design may actually have something to do with their fab problems, so having someone else fab them isn't necessarily a magic cure-all.
good point.
Originally posted by Transcendental Octothorpe
I don't believe that you can seperate chip design from fab so easily.
From what I understand, chips need to be designed for manufacturability.
The great bulk of them, however, are not designed well enough to be manufactured amid dustbunnies. Mot actually shut off the fans in their fabs to save money!
You do have a point in that Mot has made some gambles with process technology that didn't pay off, and they're currently struggling to get that tech working on 130nm. So it's not an either/or thing.
Just saying that Mot's design may actually have something to do with their fab problems, so having someone else fab them isn't necessarily a magic cure-all.
As I understand it, the original 500MHz ceiling was imposed by a design flaw in the L1 cache, and the 7450 bowed in with a laundry list of errata, so this is certainly possible. But there are more mundane reasons for Mot's inability to fabricate CPUs in quantity.
Now I guess they should have banned me rather than just shut off posting priviledges, because kickaha and Amorph definitely aren't going to like being called to task when they thought they had it all ignored *cough* *cough* I mean under control. Just a couple o' tools.
Don't worry, as soon as my work resetting my posts is done I'll disappear forever.
Originally posted by User Tron
Source? So far no problems have been reported besides some doubtfull rumors. Robert from Powerlogix seems to be very pleased with the 7457 and is confident that Mot will deliver. (http://www.cubeowner.com/forums//ind...19&t=3011&st=0)
End of Line
Source?, Sure! apple.com, can you find an apple product that is running >1GHz, that is not in PMac?. And how long has apple gleefuly remained at 1GHZ for the majority of its products. Oh yeah, Moto's doing a great job, everyone give them a round of applause!
Originally posted by AirSluf
Future-wise though, (and back in synch with your sub-thread) even with the fab at Crolles, I don't know what the chances are for Mot's recovery in PPC production. My impression was much of that fab was dedicated to joint chip and process R&D pursuits in the telecom sector, a bit of a stretch to use it as a life-ring for one partners major product line, although anythings possible.
Maybe Moto will just design newer G4's and outsource production to IBM's Fishkill plant. No, they're probably too stupid to do something sensible like that.
I'm really looking forward to the day when Steve gets up on stage and announces that Moto is no more, and then cusses them out (probably not going to happen either).