I think that CPU yields will determine whether or not Apple goes for Dual CPU's. Apple knows full well that the first 970 machines are going to sell like mad, and so the first release of such chips will probably be single CPU's. The question would then be what the price is for the single CPU boxes. I wouldn't get your expectations up over dual 970's (except maybe on a rather expensive benchmark crushing uberprofessional desktop machine).</strong><hr></blockquote>
Single processor 970's, even at speeds of 1.8 Ghz, at the same price as the current Dual G4's will be a step backward in closing the price performance gap with Intel. It is in Apples best interest to close this gap as soon as possable, and the only way to do this that I can see today (without going to intel/AMD) is to use dual 970's. A single 1.8 Ghz will be a hard sell agenst a 3.4+ Intel without undesputed independent benchmarks and killer reviews.
I think that CPU yields will determine whether or not Apple goes for Dual CPU's. Apple knows full well that the first 970 machines are going to sell like mad, and so the first release of such chips will probably be single CPU's. The question would then be what the price is for the single CPU boxes. I wouldn't get your expectations up over dual 970's (except maybe on a rather expensive benchmark crushing uberprofessional desktop machine).</strong><hr></blockquote>
I agree. It would be safer for Apple to make the initial 970 run as Single processors and then follow up with 90 nanometer PPC 970's in Dual configurations in late 2004. This sounds like it would keep pace with What intel should have by them. Yumm...Dual 2.4Ghz PPC 970 systems with PCI Express. I can hardly wait.
[quote] Single processor 970's, even at speeds of 1.8 Ghz, at the same price as the current Dual G4's will be a step backward in closing the price performance gap with Intel. It is in Apples best interest to close this gap as soon as possable, and the only way to do this that I can see today (without going to intel/AMD) is to use dual 970's. A single 1.8 Ghz will be a hard sell agenst a 3.4+ Intel without undesputed independent benchmarks and killer reviews. <hr></blockquote>
You have a point there. Based on IBM's estimations the 1.8Ghz PPC 970 will perform somewhere between a P4 2.6Ghz-3.06Ghz range. Intel will be even faster come fall of this year. A Dual PPC970 setup would contain the clout to battle back. Apple's top Powermac at $2800 needs to run Dual 1.8Ghz chips. I could see the entry and midrange packing Single procs but the top has to be dual just so that we can thumb our noses at x86 for once.
I read somewhere (though I forget where... groan!) that the Power5 would have VMX, as will all future PPC9XXs. The idea is to simplify the architecture for optimisation/compilation. Also the benefits of altivec/VMX are huge on a fast processor with mad bandwidth. IBM don't want their PowerX to be smoked badly by PowerPC XXXs.
You have a point there. Based on IBM's estimations the 1.8Ghz PPC 970 will perform somewhere between a P4 2.6Ghz-3.06Ghz range. Intel will be even faster come fall of this year. A Dual PPC970 setup would contain the clout to battle back. Apple's top Powermac at $2800 needs to run Dual 1.8Ghz chips. I could see the entry and midrange packing Single procs but the top has to be dual just so that we can thumb our noses at x86 for once.</strong><hr></blockquote>
Remeber, the 1.8 will be the top end at introduction, or slightly after the introduction. Also a Dual processor CAN NEVER achieve 2X the speed of the processor. I might be wrong here, but I think the best they can achieve is somewhere around 80% of the second processors speed for multi-threaded tasks, due to overhead in managing the tasks(or aproximatly equivalent to 3.24 ghz processor), and that effeciency is reduced with each additional processor that is added.
Remeber, the 1.8 will be the top end at introduction, or slightly after the introduction. Also a Dual processor CAN NEVER achieve 2X the speed of the processor. I might be wrong here, but I think the best they can achieve is somewhere around 80% of the second processors speed for multi-threaded tasks, due to overhead in managing the tasks(or aproximatly equivalent to 3.24 ghz processor), and that effeciency is reduced with each additional processor that is added.</strong><hr></blockquote>
That's true in 99.9% of cases. For some compute-bound cases running two independent threads, a dual processor arrangement can very slightly outrun a single processor because there's no overhead from switching contexts.
This is a fairly exotic case, but I saw a real-world example with the RC5 client (if memory serves): A DP machine ran the program just a bit over twice as fast as a SP machine at the same clock speed.
My guess on price of the IBM 970 is about $350 or even less. IBM and Apple both want this to sell, and get the PPC platforms moving. A single in the low end PowerMac and dual 970s in the top. Hopefully dual 970s in the middle PowerMac too. Considering chip size and transistor count, I believe the yields will be good. IBM is likely to price this on anticipated yields and manufacturing cost too, not on initial yields. It would be far better to make less profit at first and get a good jump start in the market, with a lot of interest and enthusiasm. Hopefully the pricing will make duals, quads and even octos attractive. Run them at lower power and get them into high end consumer Macs too, and certainly the bigger Powerbooks.
Yeah I'm aware of that but at 937 for SpecInt and 1037 for Specfp x2 you still have a healthy margin of advantage with a Dual PPC 970 1.8Ghz. I mean can you imagine Pegging the CPU meters on that box running full throttle?
iDunno about any of the techno jargon, sorry. But from an effieciency standpoint, even the G4 doesn't look that bad. Too bad it doesn't have the FSB to make anything of 4+ CPU's. 970 has FSB, FSB, FSB and very decent power characteristics considering the kind of benches it puts up. Would this be the chip, the design paradigm to plunk 2 or 4 or even 8 into a standardish ATX case and get to work crunching numbers big time?
I think we should all be thinking MP aware everything.
970 in all PowerMac's in 2003 ... (Dual or otherwise)
Power5 in new Workstation class Mac's in late 2003/early 2004.
There's also a possibility that a "Workstation" class Mac in the latter part of this year could be run on 970, the same as PowerMac's, except it's motherboard would be souped up to handle oodles of thru-put, with perhaps the option of Quads.
Hey, ya gotta Shake somethin', er uh, run Shake on something, rather ...
Don't forget that IBM's POWER series is designed for durability rather than speed, at considerable cost. Considering that the POWER5 resembles the 970 in terms of its design goals, I wouldn't be surprised if a 980 emerges that's basically the POWER5 built to consumer-(and Intel-) grade specs. The resulting faster, cheaper processor would be a good candidate to replace the 970, obviously.
The odds of Apple using the POWER5 itself are slim to none, IMO. Unless Apple decides to get into seriously upscale hardware (not merely in performance, but also in terms of reliability and cost), there's no point in adopting it. The POWER5 should be much easier to adapt into a chip that Apple can use than the POWER4 was.
[quote]While we do not know what the new ultra-pro Mac line will be named, we do have some clues indicating the class of features and performance to be offered. It appears the new line's introduction will coincide with a major new release of OS X, as both quad and 8-way multiple processors are mentioned as earmarked for the new platform. We can only speculate that these processors will be of the IBM 970 family, as not even a whisper has come our way on that topic. But, if so, we may see the PowerMac line step up to Motorola's newest 7457 G4 processors, and stay with those chips for the foreseeable future, moving to a 1.8GHz clock speed as we move into 2004. As faster 970's appear in 2004, the PowerMac should then migrate to the slower versions of the IBM chip, leaving the new ultra-pro towers to enjoy the hottest 970 version then available.<hr></blockquote>
Okay now...get some tissue and wipe up that drool <img src="graemlins/lol.gif" border="0" alt="[Laughing]" />
<strong>Power5 in new Workstation class Mac's in late 2003/early 2004.
</strong><hr></blockquote>
POWER5 availablility has always been described as "2004". Why would you even mention 2003? Especially in an Apple machine, as opposed to the latest high end IBM servers?
POWER5 availablility has always been described as "2004". Why would you even mention 2003? Especially in an Apple machine, as opposed to the latest high end IBM servers?</strong><hr></blockquote>
Good question ... ok, late 2003 is probably way too optomistic, but here's the other reason why 2004 might not be, especially if Apple does go for a Shake-y workstation type machine:
[quote]Posted on InfoWorld <strong>
In 2004, IBM will roll out its Power5 processor, which will in some ways complete an overhaul of the company's entire Unix server line. Just as the Power4 chip revitalized IBM's Unix servers on the high end, the company is hoping the Power5 chip can boost the performance of midrange and low-end systems, said Ravi Arimilli, IBM fellow and chief architect, during a recent interview. With chips tuned for each class of Unix server it sells, IBM is looking to keep the heat on Sun Microsystems and stop users from defecting to Intel's Itanium processor.
"The Power5 chip is more of a midrange or low end design that can drive up to the high end and then down to things like blades," Arimilli said. "You don't see that with Power4. Two years between chips is enough time to run away and do some dramatic things."
The Power5 chip will replace the Power4 across the board, but improvements to the chip design and more attention to heat issues will help the new chip scale further down IBM's server line.
</strong><hr></blockquote>
The link to the rest of the article is posted elsewhere in this thread. I agree this is very optomistic, and that yes, this is ratcheting up the expectations quite a bit, but I wanted to speculate here so I could get a sense of the possibilities from folks who know chips a lot better than I do.
SO ...
If this is wayyyy impossible, it's better to know sooner, than after we all fill our drool buckets in 6 months.
Hmm. This might be at long last the fruition of the various acquisitions Apple has made in the last few years. (I'm thinking Raycer, for example.) This also might coincide with something I read on MacRumours' Page Two section, which I remember saying that Apple was working with Avid on an über-video editing workstation to be released this year. If the machines are one and the same, the date given might be off...but nevertheless it's certainly something tantalizing to think about.
[QB]Despite my teasing with posting THAT Intel roadmap, let's not be too pessimistic.
For a start, the PPC situation CANNOT get any worse than it is now. Simply not possible. That is, proviso, if Apple goes with IBM.
<hr></blockquote>
Oh, it could be worse: Motorolla declares bankrupcy and haults all production of processors to save costs. Apple scrambling, must use G3s in all product lines until the 970 is released in August. Due to technical problems, however, the 970 must be underclocked at 1 gzh for the first edition of chips. In dispair Apple releases OS X for intel and halts all hardware production. While this is all going on al-Quada explode nukes in New York and Washington on the second aniversery of September 11. A finacial disaster ensues. With revenue dropping Apple declares bankrupcy at the Macworld San Fransisco 2004. By late 2004 world wide revolution takes place and capitalism is driven from the globe. Crazed anarchist fire all of the US's nukes at one location on the equater pushing the earth off orbit. On what is suppose to be Macworld 2005, the earth burns up in the sun . . .
<strong>Yeah I'm aware of that but at 937 for SpecInt and 1037 for Specfp x2 you still have a healthy margin of advantage with a Dual PPC 970 1.8Ghz. I mean can you imagine Pegging the CPU meters on that box running full throttle?</strong><hr></blockquote>
With the "safe" scenario you posted earlier Apple gets killed by the press for another "too little to late" upgrade if they release the single 970's, becouse you dont have that "healthy margine" with the singles that you do with the duals...
Another poster, and MacWispers hinting at workstation level Macs...while I do see the point to Apple releaseing them for many reasons, top on the list is finding new markets for a larger market share, Apple despiratly needs to address shortcomings in their core (PowerMac...Graphics) industries ASAP. The "safest" bet for increasing revenues is to offer computers to thier past customers that make them want to upgrade today. They do this by offering them a hardware/software solution which will save the customer money by increasing productivity enough to offset the cost of purchasing the new hardware/software.
At the same time Apple does need to address the needs of other markets, such as audio/video (3-D, and film workstation class computers...Quads+) , gamers and weekend "Geeks" (iCube, eTower, Headless iMac, whatever). This is the surest way to attract customers to switch from another platform that wont consider it today becouse their needs arnt being addressed by Apples current offerings.
<strong>Don't forget that IBM's POWER series is designed for durability rather than speed, at considerable cost. Considering that the POWER5 resembles the 970 in terms of its design goals, I wouldn't be surprised if a 980 emerges that's basically the POWER5 built to consumer-(and Intel-) grade specs. The resulting faster, cheaper processor would be a good candidate to replace the 970, obviously.
The odds of Apple using the POWER5 itself are slim to none, IMO. Unless Apple decides to get into seriously upscale hardware (not merely in performance, but also in terms of reliability and cost), there's no point in adopting it. The POWER5 should be much easier to adapt into a chip that Apple can use than the POWER4 was.</strong><hr></blockquote>
Yeah, good point ...
I get the impression that when Ravi Arimilli is talking "Power5", he seems to be talking about a family of chips that will "scale", with far less change than the Power4, from the big iron to the blade ... due to this possible similarity between the low and hi end, maybe a 98x line of chips will be considered "low end Power5"?
Comments
<strong>
I think that CPU yields will determine whether or not Apple goes for Dual CPU's. Apple knows full well that the first 970 machines are going to sell like mad, and so the first release of such chips will probably be single CPU's. The question would then be what the price is for the single CPU boxes. I wouldn't get your expectations up over dual 970's (except maybe on a rather expensive benchmark crushing uberprofessional desktop machine).</strong><hr></blockquote>
Single processor 970's, even at speeds of 1.8 Ghz, at the same price as the current Dual G4's will be a step backward in closing the price performance gap with Intel. It is in Apples best interest to close this gap as soon as possable, and the only way to do this that I can see today (without going to intel/AMD) is to use dual 970's. A single 1.8 Ghz will be a hard sell agenst a 3.4+ Intel without undesputed independent benchmarks and killer reviews.
<img src="confused.gif" border="0">
<strong>
I think that CPU yields will determine whether or not Apple goes for Dual CPU's. Apple knows full well that the first 970 machines are going to sell like mad, and so the first release of such chips will probably be single CPU's. The question would then be what the price is for the single CPU boxes. I wouldn't get your expectations up over dual 970's (except maybe on a rather expensive benchmark crushing uberprofessional desktop machine).</strong><hr></blockquote>
I agree. It would be safer for Apple to make the initial 970 run as Single processors and then follow up with 90 nanometer PPC 970's in Dual configurations in late 2004. This sounds like it would keep pace with What intel should have by them. Yumm...Dual 2.4Ghz PPC 970 systems with PCI Express. I can hardly wait.
[quote] Single processor 970's, even at speeds of 1.8 Ghz, at the same price as the current Dual G4's will be a step backward in closing the price performance gap with Intel. It is in Apples best interest to close this gap as soon as possable, and the only way to do this that I can see today (without going to intel/AMD) is to use dual 970's. A single 1.8 Ghz will be a hard sell agenst a 3.4+ Intel without undesputed independent benchmarks and killer reviews. <hr></blockquote>
You have a point there. Based on IBM's estimations the 1.8Ghz PPC 970 will perform somewhere between a P4 2.6Ghz-3.06Ghz range. Intel will be even faster come fall of this year. A Dual PPC970 setup would contain the clout to battle back. Apple's top Powermac at $2800 needs to run Dual 1.8Ghz chips. I could see the entry and midrange packing Single procs but the top has to be dual just so that we can thumb our noses at x86 for once.
Time will tell though.
Edit: It may have been the Power6 I read about.
[ 02-18-2003: Message edited by: 1337_5L4Xx0R ]</p>
<strong>
You have a point there. Based on IBM's estimations the 1.8Ghz PPC 970 will perform somewhere between a P4 2.6Ghz-3.06Ghz range. Intel will be even faster come fall of this year. A Dual PPC970 setup would contain the clout to battle back. Apple's top Powermac at $2800 needs to run Dual 1.8Ghz chips. I could see the entry and midrange packing Single procs but the top has to be dual just so that we can thumb our noses at x86 for once.</strong><hr></blockquote>
Remeber, the 1.8 will be the top end at introduction, or slightly after the introduction. Also a Dual processor CAN NEVER achieve 2X the speed of the processor. I might be wrong here, but I think the best they can achieve is somewhere around 80% of the second processors speed for multi-threaded tasks, due to overhead in managing the tasks(or aproximatly equivalent to 3.24 ghz processor), and that effeciency is reduced with each additional processor that is added.
<strong>
Remeber, the 1.8 will be the top end at introduction, or slightly after the introduction. Also a Dual processor CAN NEVER achieve 2X the speed of the processor. I might be wrong here, but I think the best they can achieve is somewhere around 80% of the second processors speed for multi-threaded tasks, due to overhead in managing the tasks(or aproximatly equivalent to 3.24 ghz processor), and that effeciency is reduced with each additional processor that is added.</strong><hr></blockquote>
That's true in 99.9% of cases. For some compute-bound cases running two independent threads, a dual processor arrangement can very slightly outrun a single processor because there's no overhead from switching contexts.
This is a fairly exotic case, but I saw a real-world example with the RC5 client (if memory serves): A DP machine ran the program just a bit over twice as fast as a SP machine at the same clock speed.
I think we should all be thinking MP aware everything.
970 in all PowerMac's in 2003 ... (Dual or otherwise)
Power5 in new Workstation class Mac's in late 2003/early 2004.
There's also a possibility that a "Workstation" class Mac in the latter part of this year could be run on 970, the same as PowerMac's, except it's motherboard would be souped up to handle oodles of thru-put, with perhaps the option of Quads.
Hey, ya gotta Shake somethin', er uh, run Shake on something, rather ...
The odds of Apple using the POWER5 itself are slim to none, IMO. Unless Apple decides to get into seriously upscale hardware (not merely in performance, but also in terms of reliability and cost), there's no point in adopting it. The POWER5 should be much easier to adapt into a chip that Apple can use than the POWER4 was.
[quote]While we do not know what the new ultra-pro Mac line will be named, we do have some clues indicating the class of features and performance to be offered. It appears the new line's introduction will coincide with a major new release of OS X, as both quad and 8-way multiple processors are mentioned as earmarked for the new platform. We can only speculate that these processors will be of the IBM 970 family, as not even a whisper has come our way on that topic. But, if so, we may see the PowerMac line step up to Motorola's newest 7457 G4 processors, and stay with those chips for the foreseeable future, moving to a 1.8GHz clock speed as we move into 2004. As faster 970's appear in 2004, the PowerMac should then migrate to the slower versions of the IBM chip, leaving the new ultra-pro towers to enjoy the hottest 970 version then available.<hr></blockquote>
Okay now...get some tissue and wipe up that drool <img src="graemlins/lol.gif" border="0" alt="[Laughing]" />
<strong>Power5 in new Workstation class Mac's in late 2003/early 2004.
</strong><hr></blockquote>
POWER5 availablility has always been described as "2004". Why would you even mention 2003? Especially in an Apple machine, as opposed to the latest high end IBM servers?
<strong>
POWER5 availablility has always been described as "2004". Why would you even mention 2003? Especially in an Apple machine, as opposed to the latest high end IBM servers?</strong><hr></blockquote>
Good question ... ok, late 2003 is probably way too optomistic, but here's the other reason why 2004 might not be, especially if Apple does go for a Shake-y workstation type machine:
[quote]Posted on InfoWorld <strong>
In 2004, IBM will roll out its Power5 processor, which will in some ways complete an overhaul of the company's entire Unix server line. Just as the Power4 chip revitalized IBM's Unix servers on the high end, the company is hoping the Power5 chip can boost the performance of midrange and low-end systems, said Ravi Arimilli, IBM fellow and chief architect, during a recent interview. With chips tuned for each class of Unix server it sells, IBM is looking to keep the heat on Sun Microsystems and stop users from defecting to Intel's Itanium processor.
"The Power5 chip is more of a midrange or low end design that can drive up to the high end and then down to things like blades," Arimilli said. "You don't see that with Power4. Two years between chips is enough time to run away and do some dramatic things."
The Power5 chip will replace the Power4 across the board, but improvements to the chip design and more attention to heat issues will help the new chip scale further down IBM's server line.
</strong><hr></blockquote>
The link to the rest of the article is posted elsewhere in this thread. I agree this is very optomistic, and that yes, this is ratcheting up the expectations quite a bit, but I wanted to speculate here so I could get a sense of the possibilities from folks who know chips a lot better than I do.
SO ...
If this is wayyyy impossible, it's better to know sooner, than after we all fill our drool buckets in 6 months.
[QB]MacWhispers had this to say today...
Hmm. This might be at long last the fruition of the various acquisitions Apple has made in the last few years. (I'm thinking Raycer, for example.) This also might coincide with something I read on MacRumours' Page Two section, which I remember saying that Apple was working with Avid on an über-video editing workstation to be released this year. If the machines are one and the same, the date given might be off...but nevertheless it's certainly something tantalizing to think about.
[QB]Despite my teasing with posting THAT Intel roadmap, let's not be too pessimistic.
For a start, the PPC situation CANNOT get any worse than it is now. Simply not possible. That is, proviso, if Apple goes with IBM.
<hr></blockquote>
Oh, it could be worse: Motorolla declares bankrupcy and haults all production of processors to save costs. Apple scrambling, must use G3s in all product lines until the 970 is released in August. Due to technical problems, however, the 970 must be underclocked at 1 gzh for the first edition of chips. In dispair Apple releases OS X for intel and halts all hardware production. While this is all going on al-Quada explode nukes in New York and Washington on the second aniversery of September 11. A finacial disaster ensues. With revenue dropping Apple declares bankrupcy at the Macworld San Fransisco 2004. By late 2004 world wide revolution takes place and capitalism is driven from the globe. Crazed anarchist fire all of the US's nukes at one location on the equater pushing the earth off orbit. On what is suppose to be Macworld 2005, the earth burns up in the sun . . .
Now this is a bad road map.
<img src="graemlins/lol.gif" border="0" alt="[Laughing]" />
<strong>Yeah I'm aware of that but at 937 for SpecInt and 1037 for Specfp x2 you still have a healthy margin of advantage with a Dual PPC 970 1.8Ghz. I mean can you imagine Pegging the CPU meters on that box running full throttle?</strong><hr></blockquote>
With the "safe" scenario you posted earlier Apple gets killed by the press for another "too little to late" upgrade if they release the single 970's, becouse you dont have that "healthy margine" with the singles that you do with the duals...
Another poster, and MacWispers hinting at workstation level Macs...while I do see the point to Apple releaseing them for many reasons, top on the list is finding new markets for a larger market share, Apple despiratly needs to address shortcomings in their core (PowerMac...Graphics) industries ASAP. The "safest" bet for increasing revenues is to offer computers to thier past customers that make them want to upgrade today. They do this by offering them a hardware/software solution which will save the customer money by increasing productivity enough to offset the cost of purchasing the new hardware/software.
At the same time Apple does need to address the needs of other markets, such as audio/video (3-D, and film workstation class computers...Quads+) , gamers and weekend "Geeks" (iCube, eTower, Headless iMac, whatever). This is the surest way to attract customers to switch from another platform that wont consider it today becouse their needs arnt being addressed by Apples current offerings.
<strong>Don't forget that IBM's POWER series is designed for durability rather than speed, at considerable cost. Considering that the POWER5 resembles the 970 in terms of its design goals, I wouldn't be surprised if a 980 emerges that's basically the POWER5 built to consumer-(and Intel-) grade specs. The resulting faster, cheaper processor would be a good candidate to replace the 970, obviously.
The odds of Apple using the POWER5 itself are slim to none, IMO. Unless Apple decides to get into seriously upscale hardware (not merely in performance, but also in terms of reliability and cost), there's no point in adopting it. The POWER5 should be much easier to adapt into a chip that Apple can use than the POWER4 was.</strong><hr></blockquote>
Yeah, good point ...
I get the impression that when Ravi Arimilli is talking "Power5", he seems to be talking about a family of chips that will "scale", with far less change than the Power4, from the big iron to the blade ... due to this possible similarity between the low and hi end, maybe a 98x line of chips will be considered "low end Power5"?
Anyway ... at this point ... who knows ...
[edit - grammar, what else?]
[ 02-18-2003: Message edited by: OverToasty ]</p>
<strong>
Now this is a bad road map.
<img src="graemlins/lol.gif" border="0" alt="[Laughing]" /> </strong><hr></blockquote>
You forgot the part where Steve Jobs rips off the face mask to reveal Bill Gates, screaming in a self-satisfied hi-pitched whine:
"There never was any hope! bwa ha ha ha" ...
... and THEN the asteriod hits.