I want you to call back and say the exact opposite of everything you just said!
JYD:
[quote] The Power4 family is no pipedream: a scaled-down derivative is being prepared for Apple. IBM is getting serious now, and they'll soon be way ahead of Motorola in desktop performance. The current G3 already has much better process technology than the G4, and Altivec-compatible SIMD is coming. Oh, and Motorola's hygiene is beyond reproach!<hr></blockquote>
Ok, I'm betting the next Apple processor is coming from IBM. Moki is possibly alluding to this and we have backup from the ATI guy over at xlr8 [edit: I remember now, it was MrNSX].
As you may remember, several months ago he said that IBM was very motivated in taking the tech from the Power4 and using it across the board. In other words they were going to use that tech to build an entire range of processors. Hmmm...son of Power4 anyone? This would also tie in with the rumors that the new processor will go by a non G name.
For the PowerPC from other sources than Motorola-discussion I would like to point to the abandonning of ATI. In case of the CPU it`s much more complicated to switch the supplier but if that was initiated in 1999 or early 2000 then there could be a result in the summer of 2002.
> but the fact of the matter is that the Amiga's "wonderful" motherboard technology was completely surpassed circa 1994 and hasn't been relevant since.
I disagree. Instead the Amiga was already obsolete in 1990. The rumor was that production cost per unit was 10,-$ back then. There was a decent redesign but due to mismanagement Commodore decided for a lukewarm overpriced try: The A3000. There was a prototype later with a Hewlett Packard RISC-Co-Processor for 3d calculation (not texture mapping like the first 3dfx voodoo cards but for geometric calculations). That was years before the first 3D card (without geometrics) arrived.
Today the architecture is obsolete no matter what CPU you would use. But for the demands of that time (multimedia was mopre bitmap-centered now it`s centered around 3D and the CPU`s ability to decompress crunched files) it was accurate.
The current PS2 architecture is the same from a console-point of view.
What smart architecture beyond the current 1980-ies office-computer desktop-design would be I don`t know.
With Quartz extreme it seems Apple is now up to think of moving forward.
It would certainly be promising to see the motherboard with a closer connection of CPU and GPU aswell as Hypertransport. It was posted here a while a go that a test-machine with such a motherboard would exist.
>I wouldn't hold my breath for some magical motherboard technology that hasn't already been discussed in these forums -- in fact it would be better if such a magical technology did not appear because that would Apple out on its own again, being incompatible and unsupportable by the rest of the electronics industry. Powerful solutions (RapidIO, HyperTransport, etc) are being developed by companies that are really good at that sort of thing, Apple just has to jump on them and use their powers of integration to build a compelling end product... that's what they are good at.
Yes I agree. Although I`m wondering if a simple powerfull CPU and a few up-to-date coprocessors (opposed to on-CPU Altivec) all with direct memory access (direct connection to RAM possible without bothering the CPU) would be any good today. Maybe that would just cause more bandwith-problems - or maybe not. I don`t know. At least specialised co-processors are limited for their purposes. That might be more yesterdays computing than a current SIMD that can be used universally.
Though I would certainly want to take a look at the motherboard designs of high-end servers and "supercomputers" for inspiration if I was in charge.
This quote seems to point in a similar direction for CPU`s:
>As you may remember, several months ago he said that IBM was very motivated in taking the tech from the Power4 and using it across the board. In other words they were going to use that tech to build an entire range of processors. Hmmm...son of Power4 anyone? This would also tie in with the rumors that the new processor will go by a non G name.
.. analogy: New materials and components in car-design are often developed in motor-sports. The manufacturers then go shopping for ideas in their motorsport-shops.
Junkyard Dawg:
>>The G3 sucks compared to the G4, anyone can see that. Apple needs something faster than the G4, and something that uses Altivec, and the only place they can get something like that is from Moto.
>> Game over.
moki:
>You'll find you are mistaken... about a great many things.
So we`ll see the old silent birth PPC 620 then at 5Ghz in June 2002? )
What I would really hope for is the current G4 revision (or another revision upwards of the G4) slightly clocked faster but Dual-CPU as standard on all Power Macs and a motherboard with better bandwith. That would give Apple enough time to develop new CPU`s. I don`t see that happening (and maybe there`s reasons like Motorola couldn`t produce enough volume) but it would be good from my point of view.
> Apple buying-out the PPC assets?
That`s indeed interesting. Quiet about that, too.
> 2-button mouse
Don`t tempt me in believing this. That`s my dream, although I already have a 3-Button mouse.
<strong>I think the powermacs will get a MB bump (166 mhz bus) DDR ram, firewire 2 (800mbps). As noted before, the economy still is shaky and Steve said he doesn't expect it to fully recover until late 2002. So anything with brand new tech will be scrapped until then (Rapid IO, G5, DDR333, firewire 2 optical connect, ect.). So the next "big" MW is MWSF 03, or the G5 might come out at a event near MWSF (think 99 DV iMacs).
But Steve also said that APPL would innovate its way through the tough times. Not releasing improved hardware goes against the grain. Also, there are R&D dollars tied up in the new Pmacs. APPL needs to start to see a return on that investment as soon as they can. They are doing well, but not *that* well.
<strong>As you may remember, several months ago he said that IBM was very motivated in taking the tech from the Power4 and using it across the board. In other words they were going to use that tech to build an entire range of processors. Hmmm...son of Power4 anyone? This would also tie in with the rumors that the new processor will go by a non G name.</strong><hr></blockquote>
Let us think, cracking PowerPC instructions into mini-ops, deep pipeline for quick clock speed scaling, designed for MP (which is on their PowerPC roadmap BTW). Hmm, none of that requires an expensive proc in and of itself. IBM is ramping a new .09 process (on which they will be shipping product in 03). Put it all together, and you have.....a possibility
Since Moto loves the PPC for the embedded market, they likely wouldn't particularly want to sell their IP in PPC. They might, however, be willing to sell the patents in Altivec, since it isn't really used that much for embedded applications (from what I recall).
So, if Apple were to buy the IP of Moto's Altivec technology, and if a separate Altivec co-processor (with Hyper-Transport interconnects?) is a possibility, then Apple would be in a much better position to use alternate suppliers for CPUs. Basically, any CPU using the PPC instruction set could more easily be modified for PowerMac use, if Apple had the ability to tack on a VelocityEngine co-proc onto other chips.
Is there any technical reasons why having SIMD on a co-processor wouldn't be possible? I suspect you'd need another controller to divert vectorized code to the co-proc., but this would be possible, would it not?
<strong>Is there any technical reasons why having SIMD on a co-processor wouldn't be possible? I suspect you'd need another controller to divert vectorized code to the co-proc., but this would be possible, would it not?</strong><hr></blockquote>
Wouldn't you end up with a idle main cpu while it's waiting for the altivec code to be processed? The code moving back and forth the two processing units might be more of a performance hit than it would help, I think. But I am no engineer...
I would hope people don't think a 17" flatscreen on the imac is a BIG announcement.
I'd like to see new Powermacs with a speed increase and the use of DDR at a 200-400mhz bus or better.
I can dream can't I?
Really lets look at the developments that have arrived in recent months.
1. Apple has devoted itself to bluetooth
2. Quicktime 6 is on the verge of release.
3. Anyone can buy the eMac now
4. RUMORS of Apple and AMD? While nothing is known or in most cases believed, both companies names appear together in one listing while others don't in the end credits of Star Wars EP2.
5. I found nothing on Apple buying the assets of PowerPC. But then again i didn't look at any rumor sites only the companies involved.
6. iTunes is supposed to release this summer for windoze.
Use this to deduct what you will but don't get your hopes up when we get iMacs with 17" flats as revolutionary change.
[quote]4. RUMORS of Apple and AMD? While nothing is known or in most cases believed, both companies names appear together in one listing while others don't in the end credits of Star Wars EP2.<hr></blockquote>
Both companies' names start with 'A'. Grasping detected
Since Moto loves the PPC for the embedded market, they likely wouldn't particularly want to sell their IP in PPC. They might, however, be willing to sell the patents in Altivec, since it isn't really used that much for embedded applications (from what I recall).
So, if Apple were to buy the IP of Moto's Altivec technology, and if a separate Altivec co-processor (with Hyper-Transport interconnects?) is a possibility, then Apple would be in a much better position to use alternate suppliers for CPUs. Basically, any CPU using the PPC instruction set could more easily be modified for PowerMac use, if Apple had the ability to tack on a VelocityEngine co-proc onto other chips.
Is there any technical reasons why having SIMD on a co-processor wouldn't be possible? I suspect you'd need another controller to divert vectorized code to the co-proc., but this would be possible, would it not?
Sit back and relax, que sera, sera. <img src="graemlins/smokin.gif" border="0" alt="[Chilling]" />
Steve's strategy for the consumer market is not aimed at speed, instead, Apple focuses at the 'Digital hub' thing. If anything, we should expect a new iDevice at the NY event. Remember Steve's most recent keynote, is it all there? <img src="confused.gif" border="0">
<strong>Since Moto loves the PPC for the embedded market, they likely wouldn't particularly want to sell their IP in PPC. They might, however, be willing to sell the patents in Altivec, since it isn't really used that much for embedded applications (from what I recall).
So, if Apple were to buy the IP of Moto's Altivec technology, and if a separate Altivec co-processor (with Hyper-Transport interconnects?) is a possibility, then Apple would be in a much better position to use alternate suppliers for CPUs. Basically, any CPU using the PPC instruction set could more easily be modified for PowerMac use, if Apple had the ability to tack on a VelocityEngine co-proc onto other chips.
Is there any technical reasons why having SIMD on a co-processor wouldn't be possible? I suspect you'd need another controller to divert vectorized code to the co-proc., but this would be possible, would it not?</strong><hr></blockquote>
No, this is not possible.
Yes AltiVec is used in the embedded markets where the G4 is used -- its a big part of the reason it does so well in those markets, as it aleviates the need for DSPs.
Comments
Moki:
I want you to call back and say the exact opposite of everything you just said!
JYD:
[quote] The Power4 family is no pipedream: a scaled-down derivative is being prepared for Apple. IBM is getting serious now, and they'll soon be way ahead of Motorola in desktop performance. The current G3 already has much better process technology than the G4, and Altivec-compatible SIMD is coming. Oh, and Motorola's hygiene is beyond reproach!<hr></blockquote>
</>
As you may remember, several months ago he said that IBM was very motivated in taking the tech from the Power4 and using it across the board. In other words they were going to use that tech to build an entire range of processors. Hmmm...son of Power4 anyone? This would also tie in with the rumors that the new processor will go by a non G name.
[ 06-11-2002: Message edited by: Arty50 ]</p>
> but the fact of the matter is that the Amiga's "wonderful" motherboard technology was completely surpassed circa 1994 and hasn't been relevant since.
I disagree. Instead the Amiga was already obsolete in 1990. The rumor was that production cost per unit was 10,-$ back then. There was a decent redesign but due to mismanagement Commodore decided for a lukewarm overpriced try: The A3000. There was a prototype later with a Hewlett Packard RISC-Co-Processor for 3d calculation (not texture mapping like the first 3dfx voodoo cards but for geometric calculations). That was years before the first 3D card (without geometrics) arrived.
Today the architecture is obsolete no matter what CPU you would use. But for the demands of that time (multimedia was mopre bitmap-centered now it`s centered around 3D and the CPU`s ability to decompress crunched files) it was accurate.
The current PS2 architecture is the same from a console-point of view.
What smart architecture beyond the current 1980-ies office-computer desktop-design would be I don`t know.
With Quartz extreme it seems Apple is now up to think of moving forward.
It would certainly be promising to see the motherboard with a closer connection of CPU and GPU aswell as Hypertransport. It was posted here a while a go that a test-machine with such a motherboard would exist.
>I wouldn't hold my breath for some magical motherboard technology that hasn't already been discussed in these forums -- in fact it would be better if such a magical technology did not appear because that would Apple out on its own again, being incompatible and unsupportable by the rest of the electronics industry. Powerful solutions (RapidIO, HyperTransport, etc) are being developed by companies that are really good at that sort of thing, Apple just has to jump on them and use their powers of integration to build a compelling end product... that's what they are good at.
Yes I agree. Although I`m wondering if a simple powerfull CPU and a few up-to-date coprocessors (opposed to on-CPU Altivec) all with direct memory access (direct connection to RAM possible without bothering the CPU) would be any good today. Maybe that would just cause more bandwith-problems - or maybe not. I don`t know. At least specialised co-processors are limited for their purposes. That might be more yesterdays computing than a current SIMD that can be used universally.
Though I would certainly want to take a look at the motherboard designs of high-end servers and "supercomputers" for inspiration if I was in charge.
This quote seems to point in a similar direction for CPU`s:
>As you may remember, several months ago he said that IBM was very motivated in taking the tech from the Power4 and using it across the board. In other words they were going to use that tech to build an entire range of processors. Hmmm...son of Power4 anyone? This would also tie in with the rumors that the new processor will go by a non G name.
.. analogy: New materials and components in car-design are often developed in motor-sports. The manufacturers then go shopping for ideas in their motorsport-shops.
Junkyard Dawg:
>>The G3 sucks compared to the G4, anyone can see that. Apple needs something faster than the G4, and something that uses Altivec, and the only place they can get something like that is from Moto.
>> Game over.
moki:
>You'll find you are mistaken... about a great many things.
So we`ll see the old silent birth PPC 620 then at 5Ghz in June 2002?
What I would really hope for is the current G4 revision (or another revision upwards of the G4) slightly clocked faster but Dual-CPU as standard on all Power Macs and a motherboard with better bandwith. That would give Apple enough time to develop new CPU`s. I don`t see that happening (and maybe there`s reasons like Motorola couldn`t produce enough volume) but it would be good from my point of view.
> Apple buying-out the PPC assets?
That`s indeed interesting. Quiet about that, too.
> 2-button mouse
Don`t tempt me in believing this. That`s my dream, although I already have a 3-Button mouse.
<strong>
You'll find you are mistaken... about a great many things.</strong><hr></blockquote>
This and other one-liner posts by Moki just cracked me up. So much so I had to register so I could post too :cool:
Next he will tell us that he "finds our lack of faith ... disturbing"
<strong>I think the powermacs will get a MB bump (166 mhz bus) DDR ram, firewire 2 (800mbps). As noted before, the economy still is shaky and Steve said he doesn't expect it to fully recover until late 2002. So anything with brand new tech will be scrapped until then (Rapid IO, G5, DDR333, firewire 2 optical connect, ect.). So the next "big" MW is MWSF 03, or the G5 might come out at a event near MWSF (think 99 DV iMacs).
[ 06-02-2002: Message edited by: Crusader ]
[ 06-02-2002: Message edited by: Crusader ]</strong><hr></blockquote>
But Steve also said that APPL would innovate its way through the tough times. Not releasing improved hardware goes against the grain. Also, there are R&D dollars tied up in the new Pmacs. APPL needs to start to see a return on that investment as soon as they can. They are doing well, but not *that* well.
Just my .02
<strong>As you may remember, several months ago he said that IBM was very motivated in taking the tech from the Power4 and using it across the board. In other words they were going to use that tech to build an entire range of processors. Hmmm...son of Power4 anyone? This would also tie in with the rumors that the new processor will go by a non G name.</strong><hr></blockquote>
Let us think, cracking PowerPC instructions into mini-ops, deep pipeline for quick clock speed scaling, designed for MP (which is on their PowerPC roadmap BTW). Hmm, none of that requires an expensive proc in and of itself. IBM is ramping a new .09 process (on which they will be shipping product in 03). Put it all together, and you have.....a possibility
<strong>...a possibility</strong><hr></blockquote>
Even the possibility would be great now.
So, if Apple were to buy the IP of Moto's Altivec technology, and if a separate Altivec co-processor (with Hyper-Transport interconnects?) is a possibility, then Apple would be in a much better position to use alternate suppliers for CPUs. Basically, any CPU using the PPC instruction set could more easily be modified for PowerMac use, if Apple had the ability to tack on a VelocityEngine co-proc onto other chips.
Is there any technical reasons why having SIMD on a co-processor wouldn't be possible? I suspect you'd need another controller to divert vectorized code to the co-proc., but this would be possible, would it not?
<strong>Is there any technical reasons why having SIMD on a co-processor wouldn't be possible? I suspect you'd need another controller to divert vectorized code to the co-proc., but this would be possible, would it not?</strong><hr></blockquote>
Wouldn't you end up with a idle main cpu while it's waiting for the altivec code to be processed? The code moving back and forth the two processing units might be more of a performance hit than it would help, I think. But I am no engineer...
I'd like to see new Powermacs with a speed increase and the use of DDR at a 200-400mhz bus or better.
I can dream can't I?
Really lets look at the developments that have arrived in recent months.
1. Apple has devoted itself to bluetooth
2. Quicktime 6 is on the verge of release.
3. Anyone can buy the eMac now
4. RUMORS of Apple and AMD? While nothing is known or in most cases believed, both companies names appear together in one listing while others don't in the end credits of Star Wars EP2.
5. I found nothing on Apple buying the assets of PowerPC. But then again i didn't look at any rumor sites only the companies involved.
6. iTunes is supposed to release this summer for windoze.
Use this to deduct what you will but don't get your hopes up when we get iMacs with 17" flats as revolutionary change.
Both companies' names start with 'A'. Grasping detected
Sorry but this damn ****ed UBB censored me
[ 06-10-2002: Message edited by: powerdoc ]</p>
So, if Apple were to buy the IP of Moto's Altivec technology, and if a separate Altivec co-processor (with Hyper-Transport interconnects?) is a possibility, then Apple would be in a much better position to use alternate suppliers for CPUs. Basically, any CPU using the PPC instruction set could more easily be modified for PowerMac use, if Apple had the ability to tack on a VelocityEngine co-proc onto other chips.
Is there any technical reasons why having SIMD on a co-processor wouldn't be possible? I suspect you'd need another controller to divert vectorized code to the co-proc., but this would be possible, would it not?
Steve's strategy for the consumer market is not aimed at speed, instead, Apple focuses at the 'Digital hub' thing. If anything, we should expect a new iDevice at the NY event. Remember Steve's most recent keynote, is it all there? <img src="confused.gif" border="0">
<strong>Would that be a "CONFIRMED" possibility.</strong><hr></blockquote>
I wouldn't want to downgrade it that much
<strong>Since Moto loves the PPC for the embedded market, they likely wouldn't particularly want to sell their IP in PPC. They might, however, be willing to sell the patents in Altivec, since it isn't really used that much for embedded applications (from what I recall).
So, if Apple were to buy the IP of Moto's Altivec technology, and if a separate Altivec co-processor (with Hyper-Transport interconnects?) is a possibility, then Apple would be in a much better position to use alternate suppliers for CPUs. Basically, any CPU using the PPC instruction set could more easily be modified for PowerMac use, if Apple had the ability to tack on a VelocityEngine co-proc onto other chips.
Is there any technical reasons why having SIMD on a co-processor wouldn't be possible? I suspect you'd need another controller to divert vectorized code to the co-proc., but this would be possible, would it not?</strong><hr></blockquote>
No, this is not possible.
Yes AltiVec is used in the embedded markets where the G4 is used -- its a big part of the reason it does so well in those markets, as it aleviates the need for DSPs.