[QB]I'm just surprised all the NDA goodness from WWDC hasn't leaked out. I don't expect a G5 at MWNY 02, but expect something very, very major by MWSF 03.
Codewarior
You are so right. We will see G5 in July 03. It is all about Motorolla and its problems along with Apple's excellent customer loyalty. There are plenty of Mac users that will buy G4's at pitifull upgrades for one more year.
The reference actually means that these ARE the G5s we're looking for. (Jedi mind tricks only work on the weak-minded.) </strong><hr></blockquote>
That is a rather astute observation, DCQ. Perhaps moki is attempting to trick weak-minded Apple lawyers while clueing us into the truth? A little bit of optimism in this bleak world never hurts, right?
Moki has all but said that the G5 is going to be a scaled down version of IBM's Power4 processor. Not sure if I believe this, but if true then it just could be Apple's meal ticket for the next several years.
Actually, to me, the rumor mill seems awfully quiet for a month before a macworld. Is it me or does anybody else think there should be more rumors floating around?</strong><hr></blockquote>
Yeah, the rumour mill is a little weak this time, someone has put the kybash on leaks, wonder why??
Heh, I wouldn't be surprised to see a post from Moki saying, "Difficult to see, the future is ... the Dark Side, clouds everything it does". So goes the saga of the mythical G5.
Here I am again, with some more information about the G5 and other hardware. I must say that it is very hard to get inside info, everything related to the G5 and RapidIO is covered by NDAs at Motorola. Unfortunately I have no inside information from Apple, so everything I can report is based on info by people that work for companies close to the PowerPC development.
- I have found someone who has worked for Motorola on the G4 and G5 designs at Somerset, but he refused to provide any inside information. Apparently there are a lot of people taking the NDAs very seriously.
- I finally had the chance to take a look at an MVP prototype motherboard by Motorola. The board should be about one year old an uses an ASIC by Marvell Technology to interconnect the CPUs on the MPX bus with other interfaces. This board still connects the RAM via MPX, although I was told the Apple Xserve architecture is based on this design - the ASIC Apple uses should be basically a Marvell ASIC with an additional DDR memory controller. The CPU daughtercard features two MPC7450 CPUs, although be board should also work with two MPC7455 processors without modifications. The I/O architecture of the board is very different from Apple's prototypes, it features PC-like serial and parallel ports (as well as USB, Ethernet and IDE onboard). PCI is 2x 32 bits and 2x 64 bits, and the board is neither green nor red, but blue (can anyone confirm if Motorola prototype boards are always blue please?).
- Marvell Technology has not only designed the MVP ASIC, but also has strong relations with Apple computer. They are the ones that have developed the Gigabit Ethernet controller used in the Powerbook G4. And Marvell is also member of the HyperTransport Consortium, same as Apple. As far as I know Marvell is working on HyperTransport circuits that could be integrated in the ASICs of the MVP and also in the Xserve ASIC. This is a bit confusing as both Motorola and IBM will use RapidIO as CPU interconnect, none of them will use HyperTransport. Maybe there are plans to interconnect the CPUs using RapidIO and to use HyperTransport to connect other elements to the central ASIC.
- Both Marvell and another company (which I will not name now) are currently working an RapidIO circuits to connect forthcoming Motorola CPUs to their ASICs. As far as I can say there are designs that are in late development, I think they shall be ready for the next MVP generation that might be launched with MPC8540 samples. Apparently Motorola is currently working on new MVP prototypes, which tells us that G5 prototypes may appear soon. But don't misunderstand me now: I'm talking about prototypes by Motorola that will probably use MPC8540 processors, I don't think we will ever see Apple machines using this technology.
- I can definately say that Motorola is currently working on at least two 'G5' designs using RapidIO, one of them being the MPC8540. I don't know what the other design(s) is(are), but considering the info I got it is highly probable that a desktop version of the G5 (which will probably be called MPC7500) is in late state of development. Unfortunately I do not know if one of these chips is already sampling.
- The RapidIO PowerPC related information embargo at Motorola and other companies related to the PowerPC is impressive. Maybe this is due to the rumored involvement by Apple Computer, I can only say that Motorola has been more open in the last few years. I could definately be that a desktop G5 processor is in a late state of development without anyone outside Motorola, Apple and a few other companies knowing about it. Nevertheless I was able to get some inside information about the MPC8540 (which I posted some time ago), and I can tell you that I was very impressed when I compared the embedded design with the performance expectations - the MIPS / complexity performance is very impressive (+25% to +50% compared to the current G4), and the overall design should allow higher than current clockrates (another +25% to +50% compared to the current G4). A 1.5 Ghz beating a 3 Ghz Pentium might be a realistic estimation for early 2003.
<strong>After the new G4 arrives Xserve can be transitioned over as its System Controller is already DDR-compatible. </strong><hr></blockquote>
Now here is something I am a bit confussed about. Wouldn't the CPU side of the ASIC chip be made for the current 133Mhz 1 GB/s chip interface. I mean the memory side is the right speed, but what about the CPU side?
I'm somewhat amazed by the excitement over the idea that Motorola is involved with the G5. These guys really aren't up to the major league level here. They have been good at design, but terrible in manufacturing - both performance and cost. On the other hand, IBM is equally capable, if not actually profoundly superior to Intel and AMD with respect to technologies.
If you were Steve Jobs, you would have been looking for a supplier besides Motorola since the G4 stalled out at 500MHz. Your problem would be that IBM is the only other member of your alliance. Well today, lo and behold, a PowerPC instruction set processor is undoubtedly the fastest processor in the world. It is called the Power4!
It incorporates on-chip multiprocessing which means the two chips communicate with one another at the clock speed of the processor, it modernizes PowerPC instruction handling, uses incredibly advanced manufacturing technologies and has behind it the company that invented the transistor and has had its employees and discoveries win many nobel prizes.
I personally would be dissappointed if Apple sticks with Motorola for another five years or so as its high end chip manufacturer. How long would we be stuck at 1.6 GHz?
<strong>Wouldn't the CPU side of the ASIC chip be made for the current 133Mhz 1 GB/s chip interface. I mean the memory side is the right speed, but what about the CPU side?</strong><hr></blockquote>
Most NorthBridges are asynchronous in this respect: the clock rates between it and the CPU and between it and the memory can be different, as it is the case in the Xserve and the new Athlon mobos sporting DDR333 (Athlon's FSB is still locked at 266 MHz, and it looks like AMD is not going to change that for the time being), just to name a few of the instances in which this occurs.
Somebody slap me down if I am wrong, but from IBM's specs on this Power4 processor, it is purely intended for the server market. Wouldn't it require a major rev of the architecture for it to be accomodated in a Mac centric system? There is no mention of Altivec or a comporable unit, what about all the work that Apple has done in optimizing all the iAppz and OSX for Altivec? Another variable has to be cost; while IBM makes these bad assed systems, they generally sell for mucho dinero$$, could these practically find their way into Apples lineup? Just wanted to raise some questions as to how realistic these really are for the Mac market..
I would just like to see Apple have multiple CPU suppliers so that they have more flexibility in the products they offer. A major motivation behind AIM was to ensure Apple multiple sources for this most critical of components. Since the arrival of the G4 Apple has been boxed in to using Motorola supplied processors (again). Yes IBM supplies the G3s, but its lack of SIMD and MP support limits its usefulness to Apple in anything except the iBook and the 1st gen iMac.
IBM and Motorola have both made it their stated direction to adopt RapidIO as the system interconnect, and BookE as the spec for future PowerPCs. IBM has decided to adopt SIMD finally, and while they are being coy about whether it is AltiVec-compatible I've got to believe that Apple would push really hard to ensure that it is. My guess is that they will both be shipping high performance desktop parts at some point in the next 2 years, and Apple will be able to use the whole range of available products due to the common use of RapidIO.
<strong>Somebody slap me down if I am wrong, but from IBM's specs on this Power4 processor, it is purely intended for the server market. Wouldn't it require a major rev of the architecture for it to be accomodated in a Mac centric system? There is no mention of Altivec or a comporable unit, what about all the work that Apple has done in optimizing all the iAppz and OSX for Altivec? Another variable has to be cost; while IBM makes these bad assed systems, they generally sell for mucho dinero$$, could these practically find their way into Apples lineup? Just wanted to raise some questions as to how realistic these really are for the Mac market.. </strong><hr></blockquote>
Very unrealistic if you consider the POWER4 as it presently is. It's for systems ranging from 200K+ USD to seven-digit figures. It'h a huge CPU, draws tons of power, runs hotter than a grill and it's most certainly optimized for multi-module configurations (every module made of up to 8 cores in 4 double-core units). And as you pointed out it lacks AltiVec, which--love it or hate it--is a key part of Apple's SW offering.
But what about a sort of scaled down version? Maybe just a single or double-core CPU, less cache, with the addition af an AltiVec-compatible unit? That sounds far more likely IMHO... We'll see... But by Moki's comments I'm pretty sure something very intriguing along these lines is about to come out...
<strong>Somebody slap me down if I am wrong, but from IBM's specs on this Power4 processor, it is purely intended for the server market. </strong><hr></blockquote>
I'm a peaceful man so I won't do any slapping. I will however point out that the specs on this processor would do just fine for Macs.
The real issue as you point out later is price. However, the high price of a processor like this is actually a function of amortizing the very high fixed costs of design and manufacturing capability rather than the actual variable costs of a little bit of copper and silicon.
In fact, the change in your pocket contains more commodity value (copper, tin, nickel etc.) than would a power4 processor (I'm assuming you have change in your pocket). IBM has to charge a ton for each processor because the only system they can put it in is their server system, of which they only sell thousands or tens of thousands of units (not hundreds of thousands or millions).
Manufacturing this chip in high volume would bring the per chip manufacturing costs (including R&D and equipment amortization and depreciation) to levels comparable with the G4.
In short, this chip is a viable alternative to the G4 and a far superiour chip to anything else out there, even running at twice the MHz.
Keep in mind, most technologies that started off in the highest end workstations and servers eventually worked their way down to even the lowliest consumer systems, and Apple has consistently been among the first PC manufacturers to incorporate these types of technologies at prices which are at a premium, but not an order of magnitude more expensive that PCs.
Here's a short list to refresh your memory: GUI, Color monitors, multitasking, RISC architecture, built-in ethernet, SCSI, Wireless networking, Gigabit ethernet, Firewire. Do you think we'll be adding multi-core processors to that list soon? I sure hope so.
<strong>Here's a short list to refresh your memory: GUI, Color monitors, multitasking, RISC architecture, built-in ethernet, SCSI, Wireless networking, Gigabit ethernet, Firewire. Do you think we'll be adding multi-core processors to that list soon? I sure hope so.</strong><hr></blockquote>
I hope so, I have some money saved that's burning a hole in my pocket. Don't get your hopes up, though. If you're slightly pessimistic as to what will come out then you'll be happier when you're proven wrong.
Comments
<strong>These aren't the G5's you're looking for... move along.</strong><hr></blockquote>
Funny.
The reference actually means that these ARE the G5s we're looking for. (Jedi mind tricks only work on the weak-minded.)
[QB]I'm just surprised all the NDA goodness from WWDC hasn't leaked out. I don't expect a G5 at MWNY 02, but expect something very, very major by MWSF 03.
Codewarior
You are so right. We will see G5 in July 03. It is all about Motorolla and its problems along with Apple's excellent customer loyalty. There are plenty of Mac users that will buy G4's at pitifull upgrades for one more year.
<strong>
Funny.
The reference actually means that these ARE the G5s we're looking for. (Jedi mind tricks only work on the weak-minded.)
That is a rather astute observation, DCQ. Perhaps moki is attempting to trick weak-minded Apple lawyers while clueing us into the truth? A little bit of optimism in this bleak world never hurts, right?
<strong>
Actually, to me, the rumor mill seems awfully quiet for a month before a macworld. Is it me or does anybody else think there should be more rumors floating around?</strong><hr></blockquote>
Yeah, the rumour mill is a little weak this time, someone has put the kybash on leaks, wonder why??
- I have found someone who has worked for Motorola on the G4 and G5 designs at Somerset, but he refused to provide any inside information. Apparently there are a lot of people taking the NDAs very seriously.
- I finally had the chance to take a look at an MVP prototype motherboard by Motorola. The board should be about one year old an uses an ASIC by Marvell Technology to interconnect the CPUs on the MPX bus with other interfaces. This board still connects the RAM via MPX, although I was told the Apple Xserve architecture is based on this design - the ASIC Apple uses should be basically a Marvell ASIC with an additional DDR memory controller. The CPU daughtercard features two MPC7450 CPUs, although be board should also work with two MPC7455 processors without modifications. The I/O architecture of the board is very different from Apple's prototypes, it features PC-like serial and parallel ports (as well as USB, Ethernet and IDE onboard). PCI is 2x 32 bits and 2x 64 bits, and the board is neither green nor red, but blue (can anyone confirm if Motorola prototype boards are always blue please?).
- Marvell Technology has not only designed the MVP ASIC, but also has strong relations with Apple computer. They are the ones that have developed the Gigabit Ethernet controller used in the Powerbook G4. And Marvell is also member of the HyperTransport Consortium, same as Apple. As far as I know Marvell is working on HyperTransport circuits that could be integrated in the ASICs of the MVP and also in the Xserve ASIC. This is a bit confusing as both Motorola and IBM will use RapidIO as CPU interconnect, none of them will use HyperTransport. Maybe there are plans to interconnect the CPUs using RapidIO and to use HyperTransport to connect other elements to the central ASIC.
- Both Marvell and another company (which I will not name now) are currently working an RapidIO circuits to connect forthcoming Motorola CPUs to their ASICs. As far as I can say there are designs that are in late development, I think they shall be ready for the next MVP generation that might be launched with MPC8540 samples. Apparently Motorola is currently working on new MVP prototypes, which tells us that G5 prototypes may appear soon. But don't misunderstand me now: I'm talking about prototypes by Motorola that will probably use MPC8540 processors, I don't think we will ever see Apple machines using this technology.
- I can definately say that Motorola is currently working on at least two 'G5' designs using RapidIO, one of them being the MPC8540. I don't know what the other design(s) is(are), but considering the info I got it is highly probable that a desktop version of the G5 (which will probably be called MPC7500) is in late state of development. Unfortunately I do not know if one of these chips is already sampling.
- The RapidIO PowerPC related information embargo at Motorola and other companies related to the PowerPC is impressive. Maybe this is due to the rumored involvement by Apple Computer, I can only say that Motorola has been more open in the last few years. I could definately be that a desktop G5 processor is in a late state of development without anyone outside Motorola, Apple and a few other companies knowing about it. Nevertheless I was able to get some inside information about the MPC8540 (which I posted some time ago), and I can tell you that I was very impressed when I compared the embedded design with the performance expectations - the MIPS / complexity performance is very impressive (+25% to +50% compared to the current G4), and the overall design should allow higher than current clockrates (another +25% to +50% compared to the current G4). A 1.5 Ghz beating a 3 Ghz Pentium might be a realistic estimation for early 2003.
<strong>After the new G4 arrives Xserve can be transitioned over as its System Controller is already DDR-compatible. </strong><hr></blockquote>
Now here is something I am a bit confussed about. Wouldn't the CPU side of the ASIC chip be made for the current 133Mhz 1 GB/s chip interface. I mean the memory side is the right speed, but what about the CPU side?
<strong>
Funny.
The reference actually means that these ARE the G5s we're looking for. (Jedi mind tricks only work on the weak-minded.)
moki's a jedi? <img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
Steve Jobs has always wanted to have the world's most powerful processor in his computers and the Power4 really fits this bill ( <a href="http://www.digit-life.com/articles/ibmpower4/"; target="_blank">http://www.digit-life.com/articles/ibmpower4/</a>; ).
I'm somewhat amazed by the excitement over the idea that Motorola is involved with the G5. These guys really aren't up to the major league level here. They have been good at design, but terrible in manufacturing - both performance and cost. On the other hand, IBM is equally capable, if not actually profoundly superior to Intel and AMD with respect to technologies.
If you were Steve Jobs, you would have been looking for a supplier besides Motorola since the G4 stalled out at 500MHz. Your problem would be that IBM is the only other member of your alliance. Well today, lo and behold, a PowerPC instruction set processor is undoubtedly the fastest processor in the world. It is called the Power4!
It incorporates on-chip multiprocessing which means the two chips communicate with one another at the clock speed of the processor, it modernizes PowerPC instruction handling, uses incredibly advanced manufacturing technologies and has behind it the company that invented the transistor and has had its employees and discoveries win many nobel prizes.
I personally would be dissappointed if Apple sticks with Motorola for another five years or so as its high end chip manufacturer. How long would we be stuck at 1.6 GHz?
[ 06-10-2002: Message edited by: vr6 ]</p>
<strong>Wouldn't the CPU side of the ASIC chip be made for the current 133Mhz 1 GB/s chip interface. I mean the memory side is the right speed, but what about the CPU side?</strong><hr></blockquote>
Most NorthBridges are asynchronous in this respect: the clock rates between it and the CPU and between it and the memory can be different, as it is the case in the Xserve and the new Athlon mobos sporting DDR333 (Athlon's FSB is still locked at 266 MHz, and it looks like AMD is not going to change that for the time being), just to name a few of the instances in which this occurs.
ZoSo
IBM and Motorola have both made it their stated direction to adopt RapidIO as the system interconnect, and BookE as the spec for future PowerPCs. IBM has decided to adopt SIMD finally, and while they are being coy about whether it is AltiVec-compatible I've got to believe that Apple would push really hard to ensure that it is. My guess is that they will both be shipping high performance desktop parts at some point in the next 2 years, and Apple will be able to use the whole range of available products due to the common use of RapidIO.
<strong>Somebody slap me down if I am wrong, but from IBM's specs on this Power4 processor, it is purely intended for the server market. Wouldn't it require a major rev of the architecture for it to be accomodated in a Mac centric system? There is no mention of Altivec or a comporable unit, what about all the work that Apple has done in optimizing all the iAppz and OSX for Altivec? Another variable has to be cost; while IBM makes these bad assed systems, they generally sell for mucho dinero$$, could these practically find their way into Apples lineup? Just wanted to raise some questions as to how realistic these really are for the Mac market..
Very unrealistic if you consider the POWER4 as it presently is. It's for systems ranging from 200K+ USD to seven-digit figures. It'h a huge CPU, draws tons of power, runs hotter than a grill and it's most certainly optimized for multi-module configurations (every module made of up to 8 cores in 4 double-core units). And as you pointed out it lacks AltiVec, which--love it or hate it--is a key part of Apple's SW offering.
But what about a sort of scaled down version? Maybe just a single or double-core CPU, less cache, with the addition af an AltiVec-compatible unit? That sounds far more likely IMHO... We'll see... But by Moki's comments I'm pretty sure something very intriguing along these lines is about to come out...
ZoSo
[ 06-10-2002: Message edited by: ZoSo ]</p>
<strong>Somebody slap me down if I am wrong, but from IBM's specs on this Power4 processor, it is purely intended for the server market. </strong><hr></blockquote>
I'm a peaceful man so I won't do any slapping. I will however point out that the specs on this processor would do just fine for Macs.
The real issue as you point out later is price. However, the high price of a processor like this is actually a function of amortizing the very high fixed costs of design and manufacturing capability rather than the actual variable costs of a little bit of copper and silicon.
In fact, the change in your pocket contains more commodity value (copper, tin, nickel etc.) than would a power4 processor (I'm assuming you have change in your pocket). IBM has to charge a ton for each processor because the only system they can put it in is their server system, of which they only sell thousands or tens of thousands of units (not hundreds of thousands or millions).
Manufacturing this chip in high volume would bring the per chip manufacturing costs (including R&D and equipment amortization and depreciation) to levels comparable with the G4.
In short, this chip is a viable alternative to the G4 and a far superiour chip to anything else out there, even running at twice the MHz.
<img src="graemlins/smokin.gif" border="0" alt="[Chilling]" />
Keep in mind, most technologies that started off in the highest end workstations and servers eventually worked their way down to even the lowliest consumer systems, and Apple has consistently been among the first PC manufacturers to incorporate these types of technologies at prices which are at a premium, but not an order of magnitude more expensive that PCs.
Here's a short list to refresh your memory: GUI, Color monitors, multitasking, RISC architecture, built-in ethernet, SCSI, Wireless networking, Gigabit ethernet, Firewire. Do you think we'll be adding multi-core processors to that list soon? I sure hope so.
<strong>Here's a short list to refresh your memory: GUI, Color monitors, multitasking, RISC architecture, built-in ethernet, SCSI, Wireless networking, Gigabit ethernet, Firewire. Do you think we'll be adding multi-core processors to that list soon? I sure hope so.</strong><hr></blockquote>
I hope so, I have some money saved that's burning a hole in my pocket. Don't get your hopes up, though. If you're slightly pessimistic as to what will come out then you'll be happier when you're proven wrong.