IBM gives vision for future Macs

[~~NeYo~~]

i Don't know if this had been posted, but i thought i'd 'DROPPIT' for all you guys who hadn't seen it! ...


A look into the future of the chipset that Macs use is expected when IBM announce their new 64-bit Power4 processor, which is expected to support the Altivec multimedia extensions which have been set out by Motorola and Apple.

The announcement is expected at the Microprocessor Forum 2002, which will be held in California in a few days. Unlike the original Power4 this will only have one processor core and is expected to be fully compatible with PowerPC technology through extensions implemented by IBM.

A spokesman from The Microprocessor Report has said "Because it supports a full 32-bit environment, this chip should be able to boot the Mac OS just fine," and he has also commented on the likely impact the announcement at the show stating "I expect there will be a fair amount of discussion about this part".

The likelihood of Mac developing their system to the 64-bit architecture any time soon is unlikely as this is a major job but their recently release Xserver could run with it, although only in 32-bit mode. Apple aren’t expected to make any major announcements about this architecture until late next year but with "a vector processing unit implementing over 160 specialized vector instructions" to boost multimedia applications users can be quite optimistic about the future of the Mac.

NeYo : Don't Grill my ass if this got posted already!

Source: http://www.siliconstrategies.com/story/OEG20021009S0020

 

[sheepguy42]

Well, I for one have not seen it. What I have seen are a number of people swearing the Power4 is not the future of the Mac, because it does not have real AltiVev. I would guess they might have to change their tune now...
On a personal note, this processor sounds very promissing to me, but I am not sure how I will handle all the Windows people I know saying "Ha! You Mac people keep telling us how great the PowerPC is, and now Apple has realized it's not, and so you will just follow them to the next thing and believe it's great because they say so!" The problem is, explaining to them the similarities and how the Power4 has all that PowerPC had and more will be useless, so then they will listen even less. And of course, they never believe you when you tell them how the Pentium line is just as x86 as a 286.

 

[plastic]

Well, one thing for sure, I do hope that either IBM or MOTO will (1) come out with new chips (or anyone for that matter) to make the Mac faster. (2) stop making empty promises (3) give us loyal Mac users something to look forward to and get a high from it.

/me prays...

 

[LordOphidian]

This has been said before, but it bears repeating, the proc in question would not be a vanilla Power4 proc. It will be the GPuL (Giga Processor Ultra Light), which is a lower power, lower heat Power4 processor with a vector unit that, while it contains aproximately the same number of instuctions, is aparently not AltiVec compatable.

The chip sound quite impressive, and I personally hope that Apple makes the jump over to this proc soon, probably starting with Xserve and then moving it down the line.

 

[MacFreak]

Well, As I know that My friends who work at IBM and told me that GPUL do included AltiVec compatable and support 32bits environment. They already tested GPUL and Power4 on prototype PowerMac. They planned to released GPUL with new PowerMac not Power4. Because of the cost issues. Anyway, within few months new xServe will release with GPUL or Power4. Mmmmm

One problem that Apple is not sure about this because of costs about GPUL. I'ts not cheap. My thinking that Apple will offer two options to use processors of Motorola and IBM. That allow customers to decide which to pick PowerMac for their budget need.

In the past that when the first PowerPC released. Many people complained to Apple. Because all products is way expensive compared to PC. That why they need to be more careful how do Apple offer the right price and comfortable for customer to purchase right machine.

Don't forget many people really didn't understand what PowerPC Brand. means and MHZ too. They thought that PowerPC is a brand name of processor. Also they look at big numbers such as MHZ. For example at CompUSA. A guy looking at Gateway that have offered 2.x ghz. I told him come and look up at Apple and showed him new Dual 1.25 GHZ. He said "Nah it's slow and it have 1.25ghz!" I said "no it's Dual" He said oh but I prefer Gateway since it have one processor of 2.x ghz which is faster than one processor and cheaper. However many people don't understand what Apple have. I kept telling people that Apple is alike Porsche. While PC is alike VW Bettle. Because of costs. Anyway, I think Apple should change the number tell them it's 2.50ghz and forget say Dual 1.25 ghz. it confused people and they are not

We do know it while others dont. That one problem.

 

[Rhino_G3]

Everybody but me must have a "friend" in the industry that works at the top level of every computer related field.

I seriously doubt that apple would go nuts and completely change the architecture of the Xserve only a few months after it's release. unless the chip's vectoring unit is 100% compatable with the altivec, the OS and apps will have to be rewritten to be able to use this functionality. Even if the Chip was altivec compatable it would be running in a 32 bit compatability environment (ala the PPC conversion running 68k compatablity) Again, I don't think they would do this in their enterprise level server.

 

[nichrome]

The GPUL, by the way, is not an official codename for the processor part.

Some Mac/PPC enthusiasts were probably bored and tried to come up with a cool sounding name (failing miserably in the process).

The part currently has no official name.

 

[nichrome]

Originally posted by Rhino_G3
unless the chip's vectoring unit is 100% compatable with the altivec

Reportedly, it is. It's not going by the name AltiVec since the trademark belongs to Motorola. If this chip was ever used in a Mac, its vector unit would be branded "Velocity Engine". This is why Apple has its own name for AltiVec -- they can switch between compatible technologies and have the same marketable name. Too bad everyone refers to Velocity Engine with the Motorola VU part name "AltiVec".

the OS and apps will have to be rewritten to be able to use this functionality.

Incorrect, assuming the PowerPC compatibility extensions are not just smoke and mirrors.

Even if the Chip was altivec compatable it would be running in a 32 bit compatability environment (ala the PPC conversion running 68k compatablity)

Flat out wrong. A full (ie. full-speed) 32-bit environment in the chip has nothing to do with 68K emulation.

Again, I don't think they would do this in their enterprise level server.

That's actually where they're most likely to do it first, since biotech companies could benefit greatly from 64-bit.

 

[Rhino_G3]

Originally posted by nichrome
Reportedly, it is. It's not going by the name AltiVec since the trademark belongs to Motorola. If this chip was ever used in a Mac, its vector unit would be branded "Velocity Engine". This is why Apple has its own name for AltiVec -- they can switch between compatible technologies and have the same marketable name. Too bad everyone refers to Velocity Engine with the Motorola VU part name "AltiVec".

True, the only reason I was using the term AltiVec was to decrease Confusion... If the instructions were not exactly the same on both vector units The apps or libraries used would have to be rewritten to reflect the change

Incorrect, assuming the PowerPC compatibility extensions are not just smoke and mirrors.

Yes, the chip will have compatability extensions.
This is the same thing that transmeta has tried to do with the Crusoe processor and x86. The clock speed is there but real world performance is lacking. I will give you the fact that the way transmeta has gone about this is less than optimal, but it is still using the same ideology.

Flat outwrong. A full (ie. full-speed) 32-bit environment in the chip has nothing to do with 68K emulation.

You are right about this... I'm not sure what I was thinking when I wrote that. It was done through software.

That's actually where they're most likely to do it first, since biotech companies could benefit greatly from 64-bit.

I'm sure the Xserve will be the first out with this chip, if it does get used. As I said though, It's not economicaly feasible to ditch a complete architecture that's only a few months old. Also, the OS will have to have the 64 bit compatablity as well if you do want any benefits from a 64 bit chip. We won't see these chips in use until the OS is ready as well.

 

[Rhino_G3]

Originally posted by nichrome
Reportedly, it is. It's not going by the name AltiVec since the trademark belongs to Motorola. If this chip was ever used in a Mac, its vector unit would be branded "Velocity Engine". This is why Apple has its own name for AltiVec -- they can switch between compatible technologies and have the same marketable name. Too bad everyone refers to Velocity Engine with the Motorola VU part name "AltiVec".

True, the only reason I was using the term AltiVec was to decrease Confusion... If the instructions were not exactly the same on both vector units The apps or libraries used would have to be rewritten to reflect the change

Incorrect, assuming the PowerPC compatibility extensions are not just smoke and mirrors.

Yes, the chip will have compatability extensions. The only time I have seen this work properly, with good performance,was in alpha chips with X86 compatability.

This is the same thing that transmeta has tried to do with the Crusoe processor and x86. The clock speed is there but real world performance is lacking. I will give you the fact that the way transmeta has gone about this is less than optimal, but it is still using the same ideology.

Flat outwrong. A full (ie. full-speed) 32-bit environment in the chip has nothing to do with 68K emulation.

You are right about this... I'm not sure what I was thinking when I wrote that. It was done through software.

That's actually where they're most likely to do it first, since biotech companies could benefit greatly from 64-bit.

I'm sure the Xserve will be the first out with this chip, if it does get used. As I said though, It's not economicaly feasible to ditch a complete architecture that's only a few months old. Also, the OS will have to have the 64 bit compatablity as well if you do want any benefits from a 64 bit chip.

 

[Rhino_G3]

IBM server chip seen slimmed down for Apple Macs
Reuters, 10.13.02, 3:49 PM ET

ARMONK, N.Y. (Reuters) - International Business Machines Corp. Monday announced a microchip for personal computers that will crunch data in chunks twice as big as the current standard and is expected by industry watchers to be used by Apple Computer Inc.
IBM said its new PowerPC chip would go into production late next year and process 64 bits of data at a time at 1.8 Gigahertz, or 1.8 billion cycles per second

http://www.forbes.com/technology/newswire/2002/10/13/rtr749520.html

Regardless if they use it or not I don't think we'll be seeing them in macs for a while.

 

[MisterMe]

Originally posted by Rhino_G3
IBM server chip seen slimmed down for Apple Macs
Reuters, 10.13.02, 3:49 PM ET

ARMONK, N.Y. (Reuters) - International Business Machines Corp. Monday announced a microchip for personal computers that will crunch data in chunks twice as big as the current standard and is expected by industry watchers to be used by Apple Computer Inc.
IBM said its new PowerPC chip would go into production late next year and process 64 bits of data at a time at 1.8 Gigahertz, or 1.8 billion cycles per second

http://www.forbes.com/technology/newswire/2002/10/13/rtr749520.html

Regardless if they use it or not I don't think we'll be seeing them in macs for a while.

You did notice that the story is embargoed until April 1st, didn't you?

 

[Rhino_G3]

actualy, that is 12:01 Eastern time which translates to 0401 Greenwich mean time

 

[WaterLung]

Just to clear a few things up, as I understand them...

1. This new "GPUL" PPC will have an SIMD unit that is fully compatible with Altivec. Apple has far too much invested in Altivec to simply walk away from it. Furthermore, the AIM alliance allows for IBM to use an altivec compatible unit.

2. While 64 bit, the GPUL will be backwards compatible, meaning that it can be run in a "32 bit mode". In other words, Apple can begin using this chip with a 32 bit version of OS X and update to 64 bit at their leisure.

3. Apple reportedly has had working prototypes of the GPUL, running OS X, since this spring. It is very possible that Apple has already developed a 64 bit version of OS X.

4. The GPUL WILL be used in Macs, there is no question about it. The great unknown is WHEN. Many insiders are saying fall 2003, but for Apple's sake we must hope that it will arrive sooner. Alternatively, there are many who believe that the reason new Macs will not boot into OS 9 beginning in 2003 is that OS 9 will not be updated to run on the GPUL. This would put the introduction of the GPUL VERY soon, to be announced at MWSF 03, and to ship soon after (or several months after).
This is most of what I know based on water-cooler gossip, but I trust this info to be fairly reliable. The real unknown is Motorola. One scenario is that the GPUL will be used in Powermacs, and the G4 will be used in consumer macs and laptop macs. A more optimistic scenario has Moto introducing an updated G4 to be used in Powermacs beginning at MWSF03; updated with a faster fsb (RapidIO), and a smaller process (130 nm). Such a G4 would migrate to consumer Macs when the GPUL is introduced in the fall. This second scenario also fits with Apple's desire to use a unified motherboard architecture, because the system interface for both the new G4 and the GPUL could theoretically be the same. There is talk of an Apple version of RapidIO, dubbed ApplePi, that will interface with both the GPUL and this new G4++.

The last paragraph is mostly speculation, but the 4 points can be taken as fact. Barring any major setbacks, within a year Apple should be able to close the performance gap....and if the GPUL performs as well in real world usage as it does on paper, then Apple will create a new performance gap, but this time on the leading edge of the gap.

 

[nichrome]

The part name is PowerPC 970.

Ahh, now we no longer have to hear ignorant babble about "GPUL".

Giga Processor Ultra Lite. Geez. How can anyone think that IBM could use such a name -- they use short, concise product names like PowerPC, Power4 and Regatta.

 

[Buckey]

IMHO the biggest advantage this chip has is that it will have a bus speed of 900MHZ. Right now the max bus speed on a G4 is 167MHZ. Go ahead, reduce the cache size to save money. If Apple can make such a leap in bus speed, we will be far better off.

 

[Rhino_G3]

Originally posted by WaterLung
Just to clear a few things up, as I understand them...

2. While 64 bit, the GPUL will be backwards compatible, meaning that it can be run in a "32 bit mode". In other words, Apple can begin using this chip with a 32 bit version of OS X and update to 64 bit at their leisure.

This would serve no advantage to having the 64 bit chip though. without an OS able to use it, the full potential of the chip just sits there, unused.

3. Apple reportedly has had working prototypes of the GPUL, running OS X, since this spring. It is very possible that Apple has already developed a 64 bit version of OS X.

Apple has reportedly been working concurently on both a 32 and 64 bit version of OS X. I had thought it was still quite a ways off though.

4. the GPUL WILL be used in Macs, there is no question about it. The great unknown is WHEN. Many insiders are saying fall 2003, but for Apple's sake we must hope that it will arrive sooner. Alternatively, there are many who believe that the reason new Macs will not boot into OS 9 beginning in 2003 is that OS 9 will not be updated to run on the GPUL. This would put the introduction of the GPUL VERY soon, to be announced at MWSF 03, and to ship soon after (or several months after).

That does make good sense why the mac wouldn't boot OS 9 after 2003. if the chip was completely 32 bit compatable wouldn't the chip remain suitable for use in OS 9 as well?

Also, what about most stories saying this chip will actualy go into production late next year?

 

[lurk]

This would serve no advantage to having the 64 bit chip though. without an OS able to use it, the full potential of the chip just sits there, unused.

I have seen this idea before and whether the kernel is 64 bit is actually not very relevent. We went through this whole thing with move to 64 bit Sparcs. There is was possible to mix and match 32/64 bit kernels and 32/64 bit applications with little concern. The only issue I recall was that you could not allocate an address space larger than that adressable by 32 bit code on a 32 bit kernel.

It was still possible to use all of the user accessable 64 bit goodness under the 32 bit kernel. The only problem is you can't get at the whole address space.

The ability to run in a 32 bit address space is kept on the 64 bit kernel since only a few programs actually benifit from addressing the whole thing and there is an expence to using big pointers. If you know you will never need more than 4 Gigabytes of addressspace for your app then you get to use 32 bit pointers and not bother shuffeling all those wide pointers arround. This is the case for 98% of all applications.

-Eric

 

[malexgreen]

Originally posted by lurk
I have seen this idea before and whether the kernel is 64 bit is actually not very relevent. We went through this whole thing with move to 64 bit Sparcs. There is was possible to mix and match 32/64 bit kernels and 32/64 bit applications with little concern. The only issue I recall was that you could not allocate an address space larger than that adressable by 32 bit code on a 32 bit kernel.

It was still possible to use all of the user accessable 64 bit goodness under the 32 bit kernel. The only problem is you can't get at the whole address space.

The ability to run in a 32 bit address space is kept on the 64 bit kernel since only a few programs actually benifit from addressing the whole thing and there is an expence to using big pointers. If you know you will never need more than 4 Gigabytes of addressspace for your app then you get to use 32 bit pointers and not bother shuffeling all those wide pointers arround. This is the case for 98% of all applications.

-Eric

So if Apple starts building the PPC970 into systems before it has a native 64-bit MACOSX, will the new 970-based systems execute 32bit MACOSX and 32-bit MACOSX-based applications (like MS OFFICE v.X, which may never be upgraded to 64-bit) faster than a similarly clocked G4? I hope so.

 

[lurk]

The answer is most likely be yes but not for any reason of 64 bittedness. The new chips have a buss speed of 900MHz which is 4-5x faster than what the G4 has. Office would not have any reason to even know that it was running on a 32 bit G4 or a 64 bit PPC970. It just won't make use of the top 32 bits in any of the 64 bit regesters.

-Eric