[zubro]

http://www-1.ibm.com/businesscenter/...av_id/emerging

The SONY PS3 should run with it next year, what about our PM?
Will they run with it before?
Anyway, Intell has nothing at all to compare on their side
So Windows shall die! (I will have such a party! )

[Chito]

I hardly think either the G5 or Windows will die. Even though I'd LIKE to see windows die.

[ChrisB]

As with any new processor design technology, it will take some time for the marketplace to determine what will succeed and what consumers are willing to pay for. Plus, it will take a year or two for mass production to catch up.

Don't expect Apple to jump anytime soon. They will need to test any new chip design (from IBM) in order to see that it fits into a seemless user experience.

That's part of the reason we don't have a PowerBook G5 yet - to support the chip today the form factor (and as a result the user experience) would have to change to accomidate more power and heat.

[Luca Rescigno]

Duplicate thread.

And not gonna happen in a Mac. The G5 and the Cell are totally different.

[Halfloaf]

I've just read the blurb on BBC website...Is this a "Adopt or die" scenario for Apple? Are they holding onto old technology?

I don't think the G5 will go anywhere soon because of the Xbox 2. (edit: if that's still what they plan on using)

-HL

[ChrisB]

When you think about it, you could say the same for Intel and AMD as well. As long as they choose to put out the P4 and RISC variants (AMD) then there is nothing wrong with Apple sticking with the G5 and up.

IBM may turn the G6 into a cell based processor, who knows. But the marketplace will decide as it always does.

I'm sure Apple knows what is going on and will evaluate accordingly - they're not stupid.

[Ganesha]

Originally posted by Halfloaf:
I've just read the blurb on BBC website...Is this a "Adopt or die" scenario for Apple? Are they holding onto old technology?

People have been saying that for years about Apple... yet somehow Apple remains here year after year.

Apple of course is hard to predict. Apple is a leader in the early adoption of technologies. Apple is also a trailer in the adoption of some established technologies (i.e. USB2).

Will Apple use Cell? It probably depends cost and on how well it performs at day to day computing.

[macaddict0001]

The g5 is not going anywhere for at least 12 years.
Not to mention cell is not going to be as scalable as they say.
I know this because:
1. No-one uses an underclocked desktop chip in a cellphone. mainly because of size restraints.
2. the chip will cost 200 dollars whether its going in a $20 inkjet or a $50000 server.
In my opinion it will be a line of chips built on the same architecture that come in all sizes and prices.

[Leonard]

As the article itself says, the Cell processor is a customized chip, just as the G5 and the future G5's are custmized chips for Apple and IBM. Besides, what makes a Cell processor, a Cell processor. Is a dual-core PowerPC 97x chip a Cell processor? The article also mentions virtualization (or running multiple OSes), which is a going to be a feature of the next G5 chip. In other words the G5 is IBM's customized chip for Apple and IBM. Frankly the definition of the Cell processor is too vague. IF IBM created a dual-core PowerPC 970x chip with an integrated memory controller, to me that could also be defined as a Cell chip.

[P]

Ars Technica has the beginnings of a rundown on the Cell:

http://arstechnica.com/articles/paedia/cpu/cell-1.ars

As you can see, it is a CPU that is not very suited to the very serial execution style of current OSes. It will certainly find room in specialized places, and maybe in the future software will be more parallellized so more things can use it, but it's not coming today. Also, the vector units are not Altivec-based (they're mcuh more trivial), so it's not a drop-in replacement.

[MORT A POTTY]

good god an integrated memory controller would be great. as would PCI-E, support for SLI, DDR2, and dual-dual core processors.....

but doubt we'll see that this year.


Cell ? G5

[Catfish_Man]

Originally posted by macaddict0001:
The g5 is not going anywhere for at least 12 years.
Not to mention cell is not going to be as scalable as they say.
I know this because:
1. No-one uses an underclocked desktop chip in a cellphone. mainly because of size restraints.
2. the chip will cost 200 dollars whether its going in a $20 inkjet or a $50000 server.
In my opinion it will be a line of chips built on the same architecture that come in all sizes and prices.

Um, huh? You have no idea what you're talking about. 12 years? The price can't vary (for example by adding or removing cores)? And what on earth does underclocked chips in cellphones have to do with anything?

[macaddict0001]

Originally posted by Catfish_Man:
Um, huh? You have no idea what you're talking about. 12 years? The price can't vary (for example by adding or removing cores)? And what on earth does underclocked chips in cellphones have to do with anything?

I am assuming that g5 variants although weak will be used in integrated devices eventually dwindling down to toasters. You do realize that chips like the 603e are still being used in devices. they have been made smaller but they are still essentially the same chip. I am assuming that it will be a full desktop chip and will have a set number or cores because if the number of cores change it will be a line of chips not one. And i doubt that the size will be able to be small enough to fit in a digital watch and also powerful enough to be in servers.

[Ganesha]

Originally posted by Catfish_Man:
Um, huh? You have no idea what you're talking about. 12 years? The price can't vary (for example by adding or removing cores)? And what on earth does underclocked chips in cellphones have to do with anything?

Not to mention 12 years ago in the Mac World...

68040 was the top of the line processor.. Not too many of those being sold. In 12 years the G5 may be quite dated.

[ChrisB]

It seems that the G4 may be the longest selling Apple related processor - it's still in low-end Mac's - we've been using it since 1999 right? That's almost 6 years, I don't think any other processor model that Apple has used has had that long a life span - has it? If that's the case, we'll be seeing the G5 for at least another 4-5 years in various other Apple products. Well in theory...

[MORT A POTTY]

Originally posted by ChrisB:
It seems that the G4 may be the longest selling Apple related processor - it's still in low-end Mac's - we've been using it since 1999 right? That's almost 6 years, I don't think any other processor model that Apple has used has had that long a life span - has it? If that's the case, we'll be seeing the G5 for at least another 4-5 years in various other Apple products. Well in theory...

why not? essentially, Intel is still using the Pentium III.

[Todd Madson]

I read the news of the cell processor with a raised eyebrow since
wasn't there rumors of a "dual core" G5 chip going on some months
back so a dual CPU machine would in effect be a four CPU machine
if they chose to use this dual core G5 chip. Which makes me ponder.

It does sound like a different architecture but perhaps some of
the ideas.

I could see Steve, at some future rollout:

"Dual dual dual! Dual CPU, Dual cores, dual video cards...etc."

[omar96]

Originally posted by MORT A POTTY:
why not? essentially, Intel is still using the Pentium III.

The P6 architecture is roughly 9 years old, I'll give you that, but it has been modified heavily for the Pentium M. Also, the Pentium III itself was released in July, 1999, so with that it's only 5.5 years old, roughly. It's got piles of nice, mobile enhancements like a longer pipeline, some intriguing power-saving circuitry, and using the P4's Netburst bus rather than the GTK+ that the Tualatin PIII uses. It'd be hard to call processors like Dothan completely new, but it'd be just as hard to call it a Pentium III...it'd be a real oversimplification.

If you're talking about a CPU family, I MIGHT give you that, but there are better analogies. Motorolla's 68k's are still found in phones and other appliances, and the original 68k is well over that 12 years old that macaddict0001 was giving as a timeframe for the G5, or even the PPC970FX's family.

[ChrisB]

Intel, IBM, and Motorola will always try to extend the life of a processor family as long as possible - saves R & D dollars, and after the chip and it's variants have been created, it becomes pure profit for them after they've recovered their R & D money. It's just smart business.

[P]

Originally posted by ChrisB:
It seems that the G4 may be the longest selling Apple related processor - it's still in low-end Mac's - we've been using it since 1999 right? That's almost 6 years, I don't think any other processor model that Apple has used has had that long a life span - has it? If that's the case, we'll be seeing the G5 for at least another 4-5 years in various other Apple products. Well in theory...

The 68000 was in the first Mac, in january 1984, and in the Classic, discontinued in september 1992. That's 8.5 years - plus it was in the Lisa before that. The 030 went from 88 (IIx) to 95 (LC 550) , also more than the G4. The G4 has just barely beat the G3, so it's not that old yet - 6 years is about par for the course, from the fastest Powermac to the slowest portable. Plus, the G4 is really two different cores: The 7400/7410 and the 7440/7450 cores, which really ought to be considered two different CPUs.

If you're going to talk about length, you can talk about the time the G4 was the fastest CPU Apple used, because that was a long time. At just below 4 years, it was the longest time in Macintosh history.

[P]

Originally posted by omar96:
The P6 architecture is roughly 9 years old, I'll give you that, but it has been modified heavily for the Pentium M. Also, the Pentium III itself was released in July, 1999, so with that it's only 5.5 years old, roughly. It's got piles of nice, mobile enhancements like a longer pipeline, some intriguing power-saving circuitry, and using the P4's Netburst bus rather than the GTK+ that the Tualatin PIII uses. It'd be hard to call processors like Dothan completely new, but it'd be just as hard to call it a Pentium III...it'd be a real oversimplification.

This sort of improvement is exactly what I would like to see the G4 get. A new FSB and possibly some pipeline rebalancing and there you have it, a mobile CPU for a few more years at least.

The core is identical form the PentPro up to the PIII, and then slightly modified for Pentium M. The 7450 G4s were a bigger change from the 7400s than the PIII to Pentium M, IMO.

[MORT A POTTY]

Originally posted by omar96:
The P6 architecture is roughly 9 years old, I'll give you that, but it has been modified heavily for the Pentium M. Also, the Pentium III itself was released in July, 1999, so with that it's only 5.5 years old, roughly. It's got piles of nice, mobile enhancements like a longer pipeline, some intriguing power-saving circuitry, and using the P4's Netburst bus rather than the GTK+ that the Tualatin PIII uses. It'd be hard to call processors like Dothan completely new, but it'd be just as hard to call it a Pentium III...it'd be a real oversimplification.

If you're talking about a CPU family, I MIGHT give you that, but there are better analogies. Motorolla's 68k's are still found in phones and other appliances, and the original 68k is well over that 12 years old that macaddict0001 was giving as a timeframe for the G5, or even the PPC970FX's family.

well, then I dont know what Apple is going to do when the e600 core gets past the 7448, because then it's got some major changes coming to it, so it technicly won't be a G4 in the sense that a Dothan isn't a traditional PIII.

so, will Apple call it a G4.5?

[omar96]

hell, i dunno...isn't that the one that's going to get a big bus boost? Besides, based on what I've read on Ars, the currently-shipping G4's are called G4e...

[MORT A POTTY]

Originally posted by omar96:
hell, i dunno...isn't that the one that's going to get a big bus boost? Besides, based on what I've read on Ars, the currently-shipping G4's are called G4e...

the current G4s are not the same as the original G4, no. they were stuck at 500Mhz forever because Moto couldn't ramp the clockspeed, so they altered the design a bit.

but yeah, the post 7448 processors will have up to 667Mhz FSB, DDR2 (I believe), dual core, and a ton of other tweaks

[Catfish_Man]

Originally posted by P:

The core is identical form the PentPro up to the PIII, and then slightly modified for Pentium M. The 7450 G4s were a bigger change from the 7400s than the PIII to Pentium M, IMO.

Eheheh, no.

7400 -> 7450: slightly improved vector issue capability, longer pipelines, slightly improved floating point unit, some cache changes, other minor tweaks

P3 -> Dothan: new bus, longer pipeline, new vector instructions set, major cache changes, microops fusion, redesigned power management, other minor tweaks

Dothan really is basically a new processor.

MORT: the 667MHz "bus" on the 8641* is internal to the chip. Externally it has an on chip memory controller, RapidIO, gigabit ethernet, and PCI-E, but no FSB.

[mad cow disease]

I'm skeptical...processor technology after processor technology has tried to unseat Intel, and even if they hiccup Intel still commands a lion's share of the CPU market. You're simply not going to unseat a company with a $115 billion+ market cap and billions of dollars in cash with one technology.

Combine that with the fact that Microsoft, Intel, and Dell are close allies and you have a virtual stranglehold on the PC market.

[Pierre B.]

Originally posted by P:
It will certainly find room in specialized places, and maybe in the future software will be more parallellized so more things can use it, but it's not coming today. Also, the vector units are not Altivec-based (they're mcuh more trivial), so it's not a drop-in replacement.

However, it is quite clear that the 64-bit Power-based core does have VMX capabilities.

This one makes me think that probably Apple would ask at some point IBM for a successor for the G5, featuring a VMX-enabled Power(5?) derivative, with a certain number of CELL-like SPUs. So something like the announced CELL, but with the robust G5 (or whatever) in the place of the 64-bit Power core that IBM presented this week. The processing of the upcoming Core Image/Video in Tiger is exactly the kind of work this new architecture has been conceived for.

The only difficulty for the moment is the programming model for the CELL architecture. We hardly know anything about this, but it seems reasonable to assume that it would be well placed in the general context of vector programming (like Altivec), or some variant. Perhaps Apple has already played with CELL and its development tools, just to see how CI/V would work on the CELL's SPUs.

So, if indeed Apple goes CELL (or something similar, adapted to the Macintosh market), what this gives: (1) a powerful G-something as principal processing unit, (2) a VMX unit to handle legacy vector code for Mac's Altivec, (3) a certain number of CELL-like SPUs, capable to handle perfectly well media processing (and who knows what else), and (4) an equally powerful GPU to handle graphics, perhaps in cooperation with the SPUs. Not bad .

[P]

Originally posted by Catfish_Man:
Eheheh, no.

7400 -> 7450: slightly improved vector issue capability, longer pipelines, slightly improved floating point unit, some cache changes, other minor tweaks

P3 -> Dothan: new bus, longer pipeline, new vector instructions set, major cache changes, microops fusion, redesigned power management, other minor tweaks

Dothan really is basically a new processor.

The 7450 redesigned the pipe completely, broke the 4-step idea for the first time ever on a Motorola chip, added 2 more vector units, and... it's a new core in all but name. The FPU is actually slightly worse...

The PentM twiddled with the decoder stage. That's where the longer pipe came from, that's where the uops fusion happens, that's where everything new happens. I'm not so certain exactly what it does with the FSB, and the vector instructions were just another bolt-on, like MMX and SSE.

What I'm saying is that the 7450 modified the execution core. The Pentium M did not, or at least very slightly, and instead focused on the surrounding parts. The Pentium Pro core, with a few minor bugfixes, still lives in the Pentium M, and that's remarkable.

[Eug Wanker]

Lots more information about Cell here.

Anyways, I'm just not convinced Apple will adopt Cell any time soon. I'm thinking updated PPC 9XX series, including possibly a variant with integrated memory controller (maybe of POWER5 derivation with SMT).

[Catfish_Man]

Originally posted by Eug Wanker:
Lots more information about Cell here.

Anyways, I'm just not convinced Apple will adopt Cell any time soon. I'm thinking updated PPC 9XX series, including possibly a variant with integrated memory controller (maybe of POWER5 derivation with SMT).

Wow, *great* info there

Looks like quite a radical departure from IBM's previous designs, since it's fully custom. Good to see that the SPEs have a double precision unit too. I wonder if the rounding differences in single precision mode would make it impossible for Apple to use SPEs/

[Big Mac]

I got to this thread late, but I wish I had the first response, because OP, you're a grade-A moron. If Cell delivers on its promises (and gets anywhere close to fulfilling the hype) that's a great thing for Apple. Cell is POWER technology. But I have a lot of doubts about Cell as a primary desktop MPU. From the information we have at this point, it appears to be a more generalized, powerful vector processor. It does not seem well suited to the desktop applications we use today. Cell seemingly requires applications that are extensively and specially multithreaded because its split execution cores are specialized (meant to perform specific functions) and require in order execution. Cell would most likely require a paradigm shift in Mac programming before we would ever see the fruits of it. If you want to read some higher quality arm-chair geek analysis of Cell (as well as a lot of inferior comments), check out the recent /. article.

[OtisWild]

... but it doesn't have real double-precision float in the SPEs, it takes a 10-fold hit in speed (from 256GFlops single to 20-25GFlops double according to the RWT article).

It could make a brilliant wicked fast coprocessor for Photoshop and FCP, I think, but as a general-purpose CPU I don't think it'll get much play in the Mac world.

[Carl Finlow]

Source : Hardware.fr
Intel presented the successor to their very high end CPU, the Itanium. The Montecito will hold no less than 1.7 billion transistors. At a clock speed of 2 GHz its dissipation should be of 100W.
In fact, this CPU includes 12 MB L3 cache for each of its two cores, which makes it 24 MB. This cache represents almost 1.2 billion transistors in itself. There are "only" 250 million per logical processor + L1 et L2 cache.
For a comparison, the Power 5, which is dual core, has about 280 millions transistors, cache included.


i thought i'd have seen more about this by now, thinking there would have been more news from the ISSCC

that's a staggering amount of both transisters and cache

[rtamesis]

January 9, 2007 San Francisco -Apple CEO Steve Jobs today unveiled Apple's newest desktop Macintosh, the parallel-processing Power Mac G6 running Mac OS 10.5 at his keynote address to 4000 MacWorld attendees at the Moscone Center. After his trademark "Oh, there's just one more thing...," he displayed a photograph of a cell to a wildly enthusiastic audience before bringing the desktop Mac on stage. The $3499 model reportedly uses four of IBM's 64-bit Cell 9000 chips, reaching top speeds of 5.5 Ghz with room up to 15 Gb of main memory. He announced that the computer is capable of achieving 1.2 teraflops.

[macaddict0001]

Originally posted by Carl Finlow:
Source : Hardware.fr
Intel presented the successor to their very high end CPU, the Itanium. The Montecito will hold no less than 1.7 billion transistors. At a clock speed of 2 GHz its dissipation should be of 100W.
In fact, this CPU includes 12 MB L3 cache for each of its two cores, which makes it 24 MB. This cache represents almost 1.2 billion transistors in itself. There are "only" 250 million per logical processor + L1 et L2 cache.
For a comparison, the Power 5, which is dual core, has about 280 millions transistors, cache included.


i thought i'd have seen more about this by now, thinking there would have been more news from the ISSCC

that's a staggering amount of both transisters and cache

amazing, but intel also said that the p4 would be at 6 ghz right now.

[Catfish_Man]

Originally posted by macaddict0001:
amazing, but intel also said that the p4 would be at 6 ghz right now.

Cache on a taped out processor is hardly the same thing as frequency on a future one. It's not really changeable. Montecito is real, the question is how well it performed. Personally, I'm expecting to be impressed, but perhaps I'm overestimating based on how well I2 has done performance-wise (not gonna argue market acceptance )

[memyselfandimac]

Originally posted by Pierre B.:
However, it is quite clear that the 64-bit Power-based core does have VMX capabilities.

This one makes me think that probably Apple would ask at some point IBM for a successor for the G5, featuring a VMX-enabled Power(5?) derivative, with a certain number of CELL-like SPUs. So something like the announced CELL, but with the robust G5 (or whatever) in the place of the 64-bit Power core that IBM presented this week. The processing of the upcoming Core Image/Video in Tiger is exactly the kind of work this new architecture has been conceived for.

The only difficulty for the moment is the programming model for the CELL architecture. We hardly know anything about this, but it seems reasonable to assume that it would be well placed in the general context of vector programming (like Altivec), or some variant. Perhaps Apple has already played with CELL and its development tools, just to see how CI/V would work on the CELL's SPUs.

So, if indeed Apple goes CELL (or something similar, adapted to the Macintosh market), what this gives: (1) a powerful G-something as principal processing unit, (2) a VMX unit to handle legacy vector code for Mac's Altivec, (3) a certain number of CELL-like SPUs, capable to handle perfectly well media processing (and who knows what else), and (4) an equally powerful GPU to handle graphics, perhaps in cooperation with the SPUs. Not bad .

I think thats correct, from what i've taken in about the cell, all Apple has to do is step in ans say we want cell that has a PPC with VMX that supports our Altivec ( the current VMX on the Cells do not), then they could say we want models with 8 sub processors for our top of the line, 6 SPE's for our mid level systems and ones with 2 SPE for the bottom tier computers. I can imagine a top of the line G6 with almost all if not all graphics processing done on chip. Not constraint by slow bus between the processor and todays video cards, a whole new way of thinking about the way PC will be built. Imagine a TV that can decode high def, run Playstation 3 games and run OS X all in one package. Combine this with HVD and you can almost see a new wave of computing on our doorstep.

[dke]

I wrote a deep analysis of the Cell Processor (PDF), and what it could mean to the Mac Market or the PC Market at large. (I have it mirrored a few places).

http://www.igeek.com/CellProcessor.pdf
http://homepage.mac.com/dke/.cv/dke/...pdf-binhex.hqx
http://www.mymac.com/fileupload/CellProcessor.pdf

[Catfish_Man]

Originally posted by memyselfandimac:
I think thats correct, from what i've taken in about the cell, all Apple has to do is step in ans say we want cell that has a PPC with VMX that supports our Altivec ( the current VMX on the Cells do not)

You're incorrect here. VMX is Altivec, Altivec is VMX. The PPE supports Altivec. The SPEs do not, they apparently support a non-strict subset of Altivec (according to the RWT article). I think the PPE actually looks reasonably plausible as a future Mac chip (probably multicore, since its single thread perf seems less than spectacular).

[moonmonkey]

Originally posted by dke:
I wrote a deep analysis of the Cell Processor (PDF), and what it could mean to the Mac Market or the PC Market at large. (I have it mirrored a few places).

http://www.igeek.com/CellProcessor.pdf
http://homepage.mac.com/dke/.cv/dke/...pdf-binhex.hqx
http://www.mymac.com/fileupload/CellProcessor.pdf

Very interesting, I think you should invest in iwork though.

[Johan Niklasson]

There will be a G5 Power Book. Apple is testing a new chip as we speak, the G6, also a chip made in cooperatin between Apple and IBM. Smaller size, less power, more cores, more cashe, higher speeds and more bandwidth. Not for the Power Book, yet, though. IBM makes the Cell and the G6 side by side.

The speeds should be between 3 and 5 GHz. Production at full capacity in June. Two months before that or maybe just before the introduction of the new Power Mac G6 workstations we will finally seee the Power Book G5.

I have waited for the G5 Power Book way too long now. Everyone at Apple does too. They feel the pressure and have tried to jump over one development stage to satisfy demand. It will come later but will be better.

Thanks Jobs for helping 5% of humanity to spend 50% less time doing their work on their computers.

[Simon]

Hmm, Johan, how would making a G6 for the Power Mac help get a G5 into the PowerBook? Care to enlighten us?

[Johan Niklasson]

Originally posted by Simon:
Hmm, Johan, how would making a G6 for the Power Mac help get a G5 into the PowerBook? Care to enlighten us?

Sorry, enlightenment is something everyone has to achieve themselves : )

The G5 PowerBook is the next step but with another G5 processor. The Power Mac can not wait for the Power Book upgrade, it will have to continue with the next generation. Maybe at one point in the future we will again see similar processor capacity in both Power Macs and Power Books but right now the development is splitting. The G5 speed problems immediately got IBM and Apple to take another road. Development started for the new G6 which will overcome the problems a faster G5 ran into.

The Cell is another interesting option for Power Mac. Maybe a version of it will end up in our Power Books. In any case the technology invented for the Cell is partly implemented in the new G6.

So the message is, dont expect same types of processors in Power Books and Power Macs, and, Yes, the Cell development will benefit the Apple product line.

[moonmonkey]

Originally posted by Johan Niklasson:
There will be a G5 Power Book. Apple is testing a new chip as we speak, the G6, also a chip made in cooperatin between Apple and IBM. Smaller size, less power, more cores, more cashe, higher speeds and more bandwidth. Not for the Power Book, yet, though. IBM makes the Cell and the G6 side by side.

The speeds should be between 3 and 5 GHz. Production at full capacity in June. Two months before that or maybe just before the introduction of the new Power Mac G6 workstations we will finally seee the Power Book G5.

I have waited for the G5 Power Book way too long now. Everyone at Apple does too. They feel the pressure and have tried to jump over one development stage to satisfy demand. It will come later but will be better.

Thanks Jobs for helping 5% of humanity to spend 50% less time doing their work on their computers.

You know too much.

[macdaemon]

I may be wrong, but it seems to me that Apple is going to crack this PC world with their amazing operation system, in the last interview to Forture, Jobs said that many PC vendors would like to licence MacOS X because their clients tired of problems with Microsoft software, so maybe Toshiba and Sony is actually that vendors, so Cell specially designed for easy way to port MacOS X. It's just my opinion, I could be wrong.

[memyselfandimac]

Originally posted by Catfish_Man:
You're incorrect here. VMX is Altivec, Altivec is VMX. The PPE supports Altivec. The SPEs do not, they apparently support a non-strict subset of Altivec (according to the RWT article). I think the PPE actually looks reasonably plausible as a future Mac chip (probably multicore, since its single thread perf seems less than spectacular).

And from arstechnica....
---------------------------------------------------------------------------
"The Cell and Apple


Finally, before signing off, I should clarify my earlier remarks to the effect that I don't think that Apple will use this CPU. I originally based this assessment on the fact that I knew that the SPUs would not use VMX/Altivec. However, the PPC core does have a VMX unit. Nonetheless, I expect this VMX to be very simple, and roughly comparable to the Altivec unit o the first G4. Everything on this processor is stripped down to the bare minimum, so don't expect a ton of VMX performance out of it, and definitely not anything comparable to the G5. Furthermore, any Altivec code written for the new G4 or G5 would have to be completely re-optimized due to inorder nature of the PPC core's issue.


So the short answer is, Apple's use of this chip is within the realm of concievability, but it's extremely unlikely in the short- and medium-term. Apple is just too heavily invested in Altivec, and this processor is going to be a relative weakling in that department. Sure, it'll pack a major SIMD punch, but that will not be a double-precision Alitvec-type punch."
---------------------------------------------------------------------------

So yes it is a VMX unit but NO, it does not fully support full blown Altivec, but if the later posters 411 is correct looks like Apple has already stepped up to the plate and redesigned a chip using cell like components but for sure their not going to water down Altivec to make it run on watered down VMX unit, they have to much invested in Altivec along with all the developers would be p'd off if they had to reoptimize all their apps for it.

[Catfish_Man]

Originally posted by memyselfandimac:
And from arstechnica....
---------------------------------------------------------------------------
"The Cell and Apple


Finally, before signing off, I should clarify my earlier remarks to the effect that I don't think that Apple will use this CPU. I originally based this assessment on the fact that I knew that the SPUs would not use VMX/Altivec. However, the PPC core does have a VMX unit. Nonetheless, I expect this VMX to be very simple, and roughly comparable to the Altivec unit o the first G4. Everything on this processor is stripped down to the bare minimum, so don't expect a ton of VMX performance out of it, and definitely not anything comparable to the G5. Furthermore, any Altivec code written for the new G4 or G5 would have to be completely re-optimized due to inorder nature of the PPC core's issue.


So the short answer is, Apple's use of this chip is within the realm of concievability, but it's extremely unlikely in the short- and medium-term. Apple is just too heavily invested in Altivec, and this processor is going to be a relative weakling in that department. Sure, it'll pack a major SIMD punch, but that will not be a double-precision Alitvec-type punch."
---------------------------------------------------------------------------

So yes it is a VMX unit but NO, it does not fully support full blown Altivec, but if the later posters 411 is correct looks like Apple has already stepped up to the plate and redesigned a chip using cell like components but for sure their not going to water down Altivec to make it run on watered down VMX unit, they have to much invested in Altivec along with all the developers would be p'd off if they had to reoptimize all their apps for it.

Um, what you just quoted supports my point, not yours. The original G4's Altivec unit was just as complete as the G4+'s or G5. What he's saying is that it may not be terribly impressive Altivec. Not sure about the "double precision punch" comment, and iirc later posters were confused about that as well (Altivec/VMX doesn't support DP).

[Eug Wanker]

Cell's VMX = Altivec.

ie. Full support. It's just not as fast per MHz.

[telem]

Originally posted by P:
The 68000 was in the first Mac, in january 1984, and in the Classic, discontinued in september 1992. That's 8.5 years - plus it was in the Lisa before that. The 030 went from 88 (IIx) to 95 (LC 550) , also more than the G4. The G4 has just barely beat the G3, so it's not that old yet - 6 years is about par for the course, from the fastest Powermac to the slowest portable. Plus, the G4 is really two different cores: The 7400/7410 and the 7440/7450 cores, which really ought to be considered two different CPUs.

If you're going to talk about length, you can talk about the time the G4 was the fastest CPU Apple used, because that was a long time. At just below 4 years, it was the longest time in Macintosh history.

The 68k chips were made until at least 1995 so that's a very long time. The 68060 was the last.

[Eug Wanker]

Another article:

Each SPE consumes about 1 W when clocked at 2 GHz, 2 W at 3 GHz, and 4 W at 4 GHz. Including the eight SPEs, the PPE, and other logic, the CELL processor will dissipate close to 15 W at 2 GHz, about double that at 3 GHz, and perhaps double that again at 4 GHz.

60 Watts at 4 GHz... and the 8 SPEs are 32 Watts. Take away the SPEs and you've got a 4 GHz chip under 30 Watts. If that's max power (and anywhere near to being accurate), that's pretty impressive.