We will not see a "G6" until at least 2005 in my opinion. The new chip model numbers will be as significant as the different chip model numbers of the Motorola's....just more frequent.
I think Apple now have caught the high-end of the home computer market.
Like the Armari and BoXX - What they should do now is try and build specific machines for the high-end video market (£10,000 sort of range) built specificly for one purpose! Like a SGI sort of thing!
IBM started developing the 970 after the Power 4 was in production. It's my impression that they will be doing them in parallel from now on, because of Apple's commitment to use these chips and IBM's desire to use them in blade servers. We likely don't hear much about the Power5 derivative because Apple likes it that way. Lots of scenarios are possible. I'd like to see a mobile G5 developed and introduced when the mini-Power5 comes out, whatever they call it.
Like the Armari and BoXX - What they should do now is try and build specific machines for the high-end video market (£10,000 sort of range) built specificly for one purpose! Like a SGI sort of thing!
Some of their high-end customers - AJA, for one - are lobbying Apple for just that sort of thing.
Given that Apple really wants the high-end video market, it'll be interesting to see how they respond.
There has been some indications from IBM that there will be revisions of the 970 that lie somewhere between the 970 and the POwer5 derived chip. There is a very good possibility that the next generation may not be a Power5 based chip at all.
The pwoer5 based chip, what ever it is called at the time would become the basis for the G6. The G6 could very well mean a major revision to the motherboard. As such I don't think we will see the 970 based machines go away to soon.
That does not mean though that 970+ or 980 couldn't have significant enhancements. It is all a question of econmics and how fast Power5 derived hardware can be generated. Lets face it, it would not take much to enhance the 970 and generate significant pay offs. A larger cache for instance would do wonders as would and enhanced vector unit. Both of these enhancements can be handled without a major impact to the core.
Dave
Quote:
Originally posted by Addison
I don't see why not. It is the next generation, and IBM have acknowledged that by calling the design on which it is based the Power5, the 980 is a derivitave of the Power5.
Jumping from the G5 to the G6 in just over a year would show the world that Apple is on the move. Intel moved quickly from the Pentium to the Pentum II. The 980 based machine is clearly going to be a much more compattitive machine thatthe G5. The G5 is intouch buit isn't all conquring. The G6 just might really have that crown.
I'd like to see the next processor in the 970 family restore the missing "pseudo little-endian mode", the lack of which is the reason why Virtual PC doesn't run on the current G5 machines.
Of course, by time such a chip comes out, a new version of VPC that doesn't need PLEM could be here. If, however, the next version of VPC still uses PLEM when available (VPC will suck worse than it does now on a G4 if PLEM is totally scrapped -- a move I can see Microsoft making to "simplify product development and testing"), a G5/G6/G-whatever with PLEM will almost certainly run VPC significantly faster than it would without.
Remember, Power4 and Power5 are not standalone chips at all -- their power comes from being in the multi-chip modules. If the 980 (if it's even going to be called that) introduces dual-core (or more...but really unlikely) symmetric multithreading, then the G6 designation is completely warranted. But does it even matter a gosh-darn bit? NO! They don't even put "Power Mac" whatever on the actual machines anymore, so who cares?
I'd like to see the next processor in the 970 family restore the missing "pseudo little-endian mode", the lack of which is the reason why Virtual PC doesn't run on the current G5 machines.
Of course, by time such a chip comes out, a new version of VPC that doesn't need PLEM could be here. If, however, the next version of VPC still uses PLEM when available (VPC will suck worse than it does now on a G4 if PLEM is totally scrapped -- a move I can see Microsoft making to "simplify product development and testing"), a G5/G6/G-whatever with PLEM will almost certainly run VPC significantly faster than it would without.
There are two levels of little-endian support in the PowerPC processor. The one that was removed from the 970 was a processor mode where all memory loads were automatically byte-swapped (except for the instructions). This is probably the mode used by VirtualPC.
The other feature, which is still in the 970, is a set of "load reversed" operations. These are approximately the same speed as the normal load instructions, but compilers typically can't generate them. Since an emulator is usually hand coded assembly (or JIT generated assembly) it shouldn't be a big deal to generate the load reversed instructions instead of the normal loads and it won't impact performance significantly. There is no doubt a lot of code that they have to change and test, but it should be perfectly doable.
Some of their high-end customers - AJA, for one - are lobbying Apple for just that sort of thing.
Given that Apple really wants the high-end video market, it'll be interesting to see how they respond.
... cue Avid shaking in their boots.
The days of proprietary DSP hardware doing all the work are coming to an end - so such a killer multi-core system would be mega-spiff .... however, what about the Grid option? (which Apple clearly also seems to be working on)
Hmmmmm ... lessee ... by the time a G5 Xserve cluster with Grid aware versions of FCP and Shake are available (probably the middle of next year) it'll be all over but the shouting.
Or if you are like me, I edit feature films on a 733Mhz G4
I kick your butt. I edit on FCP 3 on a G4 Cube 450mhz 384mb of ram, 16mb of VRAM.
Back on topic. I don't think it would be smart for apple to brand the 980 (or whatever it ends up becoming) a G6. Having a newer generation processor in a portable opens up a whole world of problems, like what to do with the G5. Sure, put it in an iMac, eMac, new system...ect...but, than Apple will be forced to scrap the G5 tower they have now. Three words for that: REALLY bad idea. The new chip should be a G5, I see no reason for it not to be. I don't know all the technical stuff like a lot of ya'll obviously do, but from a marketing and sales standpoint...worst idea ever.
There are two levels of little-endian support in the PowerPC processor. The one that was removed from the 970 was a processor mode where all memory loads were automatically byte-swapped (except for the instructions). This is probably the mode used by VirtualPC.
The other feature, which is still in the 970, is a set of "load reversed" operations. These are approximately the same speed as the normal load instructions, but compilers typically can't generate them. Since an emulator is usually hand coded assembly (or JIT generated assembly) it shouldn't be a big deal to generate the load reversed instructions instead of the normal loads and it won't impact performance significantly. There is no doubt a lot of code that they have to change and test, but it should be perfectly doable.
Ah, good! I wonder if Microsoft has anyone working on VPC... or anyone who can.
Comments
Originally posted by Outsider
I meant SIMD unit and assumed it would be VMX. But I do recall there being mention of a SIMD unit. Must do research...
Apple won't use it if it doesn't have Altivec (VMX).
Originally posted by discstickers
Apple won't use it if it doesn't have Altivec (VMX).
Why would Apple want to use the POWER5?
Originally posted by Zapchud
Why would Apple want to use the POWER5?
Ah c'mon...you know he means a POWER5 derivative, not an actual POWER5.
Originally posted by discstickers
Apple won't use it if it doesn't have Altivec (VMX).
Exactly, but we were referring to the POWER5, a chip Apple will most likely not use.
Originally posted by jouster
Ah c'mon...you know he means a POWER5 derivative, not an actual POWER5.
Yes, of course I assumed so, but the "it" we were talking about was the POWER5
Like the Armari and BoXX - What they should do now is try and build specific machines for the high-end video market (£10,000 sort of range) built specificly for one purpose! Like a SGI sort of thing!
Hey just an Idea. What do yu think?
Originally posted by wancarlos
Like the Armari and BoXX - What they should do now is try and build specific machines for the high-end video market (£10,000 sort of range) built specificly for one purpose! Like a SGI sort of thing!
Some of their high-end customers - AJA, for one - are lobbying Apple for just that sort of thing.
Given that Apple really wants the high-end video market, it'll be interesting to see how they respond.
Or if you are like me, I edit feature films on a 733Mhz G4
The pwoer5 based chip, what ever it is called at the time would become the basis for the G6. The G6 could very well mean a major revision to the motherboard. As such I don't think we will see the 970 based machines go away to soon.
That does not mean though that 970+ or 980 couldn't have significant enhancements. It is all a question of econmics and how fast Power5 derived hardware can be generated. Lets face it, it would not take much to enhance the 970 and generate significant pay offs. A larger cache for instance would do wonders as would and enhanced vector unit. Both of these enhancements can be handled without a major impact to the core.
Dave
Originally posted by Addison
I don't see why not. It is the next generation, and IBM have acknowledged that by calling the design on which it is based the Power5, the 980 is a derivitave of the Power5.
Jumping from the G5 to the G6 in just over a year would show the world that Apple is on the move. Intel moved quickly from the Pentium to the Pentum II. The 980 based machine is clearly going to be a much more compattitive machine thatthe G5. The G5 is intouch buit isn't all conquring. The G6 just might really have that crown.
Of course, by time such a chip comes out, a new version of VPC that doesn't need PLEM could be here. If, however, the next version of VPC still uses PLEM when available (VPC will suck worse than it does now on a G4 if PLEM is totally scrapped -- a move I can see Microsoft making to "simplify product development and testing"), a G5/G6/G-whatever with PLEM will almost certainly run VPC significantly faster than it would without.
Originally posted by shetline
I'd like to see the next processor in the 970 family restore the missing "pseudo little-endian mode", the lack of which is the reason why Virtual PC doesn't run on the current G5 machines.
Of course, by time such a chip comes out, a new version of VPC that doesn't need PLEM could be here. If, however, the next version of VPC still uses PLEM when available (VPC will suck worse than it does now on a G4 if PLEM is totally scrapped -- a move I can see Microsoft making to "simplify product development and testing"), a G5/G6/G-whatever with PLEM will almost certainly run VPC significantly faster than it would without.
There are two levels of little-endian support in the PowerPC processor. The one that was removed from the 970 was a processor mode where all memory loads were automatically byte-swapped (except for the instructions). This is probably the mode used by VirtualPC.
The other feature, which is still in the 970, is a set of "load reversed" operations. These are approximately the same speed as the normal load instructions, but compilers typically can't generate them. Since an emulator is usually hand coded assembly (or JIT generated assembly) it shouldn't be a big deal to generate the load reversed instructions instead of the normal loads and it won't impact performance significantly. There is no doubt a lot of code that they have to change and test, but it should be perfectly doable.
Originally posted by Amorph
Some of their high-end customers - AJA, for one - are lobbying Apple for just that sort of thing.
Given that Apple really wants the high-end video market, it'll be interesting to see how they respond.
... cue Avid shaking in their boots.
The days of proprietary DSP hardware doing all the work are coming to an end - so such a killer multi-core system would be mega-spiff .... however, what about the Grid option? (which Apple clearly also seems to be working on)
Hmmmmm ... lessee ... by the time a G5 Xserve cluster with Grid aware versions of FCP and Shake are available (probably the middle of next year) it'll be all over but the shouting.
either way ...
"Longhorn this Bill"
Originally posted by tak1108
They Already have one, it's called the G5!
Or if you are like me, I edit feature films on a 733Mhz G4
I kick your butt. I edit on FCP 3 on a G4 Cube 450mhz 384mb of ram, 16mb of VRAM.
Back on topic. I don't think it would be smart for apple to brand the 980 (or whatever it ends up becoming) a G6. Having a newer generation processor in a portable opens up a whole world of problems, like what to do with the G5. Sure, put it in an iMac, eMac, new system...ect...but, than Apple will be forced to scrap the G5 tower they have now. Three words for that: REALLY bad idea. The new chip should be a G5, I see no reason for it not to be. I don't know all the technical stuff like a lot of ya'll obviously do, but from a marketing and sales standpoint...worst idea ever.
Originally posted by Programmer
There are two levels of little-endian support in the PowerPC processor. The one that was removed from the 970 was a processor mode where all memory loads were automatically byte-swapped (except for the instructions). This is probably the mode used by VirtualPC.
The other feature, which is still in the 970, is a set of "load reversed" operations. These are approximately the same speed as the normal load instructions, but compilers typically can't generate them. Since an emulator is usually hand coded assembly (or JIT generated assembly) it shouldn't be a big deal to generate the load reversed instructions instead of the normal loads and it won't impact performance significantly. There is no doubt a lot of code that they have to change and test, but it should be perfectly doable.
Ah, good! I wonder if Microsoft has anyone working on VPC... or anyone who can.
The X1.