Let's not turn this thread into a realtime raytracing-tread, but I do suggest that you try searching Google with these criteria: "realtime"+"raytracing"... lots of examples found easily...
This story is complete bollocks. Apple just launched a completely new range of machines (and can't even ship them) and you're contemplating that they're going to have another completely new range of chips within a year!?
Can't even ship them? If they don't start shipping when Apple said they will, then you can make that comment. They had good reasons to pre-announce when they did.
This story is not bollocks, at least not for the reason that you give. The 970 is based on the POWER4, so that is really 1.5 year old technology at this point. The 980 is based on the POWER5 which should arrive sometime in Q2 '04, and this story says that the 980 will follow the POWER5 closely which is possible because they've done the groundwork to scale down the POWER5s. The transition from 970 -> 980 will be roughly equivalent to the transition from 601 -> 604 which took from '94 to '95, or the 603e -> G3 which (IIRC) was about a year or year and half. Just because we've been stuck on the G4 for several years thanks to Motorola doesn't mean Apple and IBM can't deliver successive chip revisions on an aggressive timetable. Indeed, even Motorola wasn't as bad as everybody complains about because they delivered the 7400, 7410, 7450, 7455, and soon the 7457. The 7400->7450 transition was a fairly major redesign with the lengthening of the pipelines.
So if you're going to try to shoot down a story, at least use a reasonable argument which has some historical support behind it!
This article cant even get the specs of the 970 correct.The 970 has 4 altivec units(simple integer,complex integer,floating point,and permute),not 2 as they claim.Dont believe a word of it.
This article cant even get the specs of the 970 correct.The 970 has 4 altivec units(simple integer,complex integer,floating point,and permute),not 2 as they claim.Dont believe a word of it.
Sorry, you're wrong. As it is now, the altivec IS split in two, one part for the Vector Permute Unit, and one part for the Vector Simple Integer, Complex Integer and Floating Point units.
The disadvantage is; since each of the two parts of the AV-unit can fetch 2 IOPs per clock per group to the Vector Issue Queues, and issue only one IOP to the execution-units per clock. This disables you from, say, executing anything simultaneously on the VCIU and VFPU.
Edit: With pic taken from arstechnica.com's excellent article on the 970.
Actually the article is still wrong as it says "pipelines" not issue queues.Maybe it was poorly worded but it doesnt increase my confidence in the source.Why should we give the benefit of a doubt to a rumor like this anyway? Maybe they just made it up based on the arstechnica artcle.It has happened before I'm sure.The idea they will have a new chip architecture in the powermacs in a year is pretty outrageous,especially considering how powerful the 970 is.And why should anyone care? Is everyone dissapointed in the G5 already even before it ships.Strange!
I believe the new 9x0 is being designed concurrently with the Power5. Should IBM adopt this rational design approach, the Power5 and 9x0 will be variations of the same design. This may reduce the time to market considerably.
The 970 was a redesign and adaptation of the Power4 and considerable reengineering was necessary. This took time. It seems possible that future Powerx and 9x0 CPUs will be introduced simultaneously.
This technical convergence of CPUs also suggests a marketing and OS convergence.
Heavy iron servers and workstations for big business from IBM and light iron servers, high performance desktops and portables for small business and individuals from Apple; all using OSX with both IBM and Apple systems using compatible software.
I believe the new 9x0 is being designed concurrently with the Power5. Should IBM adopt this rational design approach, the Power5 and 9x0 will be variations of the same design. This may reduce the time to market considerably.
It could get very interesting indeed!
People forget, such an approach would help IBM as much as it helps Apple and that it would surely be cheaper and more beneficial in 'time to market' if done in this 'simultaneous' way.
This 'rumour' should not be underestimated. For IBM? The gloves are off in its war with Intel. This goes deeper in some ways than the relationship with Apple. Intel are coming after IBM with the Itannic. I don't think IBM would be foolish to take that threat lying down.
We're so used to Intel and AMD being aggressive. Why not Apple and IBM? Now the cause is just, the reason sound. I don't see the 980 taking as long to market. The 970 is a clear indicator that Apple and IBM have forged a new alliance. Why would a 980 take another four years? IBM have indicated prototypes of the 'next gen' exist in their labs. Frightening. Time to market? Well, the Power 5 is hitting early/spring 04? IF it is...then maybe we can suppose that a 980 may only be a half year/year behind tops! Looks like it could be early 05 at the latest.
The time is ripe for PPC to put up or shut up. PPC is just a spring chicken compared to x86 land... Apple can't afford any more G4 debacles. Look what it did to G4 tower sales. Exactly!
Hmmm 10ghz front side bus? Does anyone here know what's next after DDR400? Now that the memory is the bottleneck it would be interesting to see what we have coming for memory and when.
I just reread the rumor, and I suspect that when they talk about the four integer and FP units, they might be talking about a heavily updated Altivec unit. Now, the unit has 2 Integer units, and 1 FP unit. Might this be 256-bit altivec?
Oh, and it seems very likely that they can actually make it to a june '04-schedule. A 980 identical to the standard POWER5 (which implies that POWER5 actually has altivec), with the exeption of half the cores should be easy enough, compared to the POWER4->970-transition.
Hmmm 10ghz front side bus? Does anyone here know what's next after DDR400? Now that the memory is the bottleneck it would be interesting to see what we have coming for memory and when.
No 10GHz FSB, just an on-die memory-controller. And after DDR400, there's DDRII (Quad-pumped).
There is one problem with all us Mac-heads fanning the flames of these rumors toward one another. We tend to compare our beloved PowerPC processors -- which may be available some months or years in the future -- with the PC processors which are available in quantity today! When the PPC 601 was first boasted, it sounded great against Intel's shipping iron. A year (or so) later a Mac with the 601 eventually shipped.
If Apple ships more than 1000 Dual 2 GHz G5 machines in August, I'll be really, really, really surprised.
There is one problem with all us Mac-heads fanning the flames of these rumors toward one another. We tend to compare our beloved PowerPC processors -- which may be available some months or years in the future -- with the PC processors which are available in quantity today!
I always compare with what I've seen on roadmaps, or expect to be out the same time as the new, shiny, hairy mac with the new&glossy processor. Comparisions with nowaday models are useless, mostly
I think the perception that the POWER4 & POWER5 are truly different architectures is incorrect. My guess is that they started from the POWER4 and began making improvements. The POWER4->POWER5 transition (and hence 970->980) will be more akin to the PentiumPro -> PentiumII -> PentiumIII transitions. Bigger than the 604 -> 604e (that'll be like 970 -> 970+), but potentially much like the various PentiumIV revisions Intel has been doing.
We'll hopefully see a 3Ghz PM the same day a 2Ghz PM is shipping this year.
We can analyze better once we see how the 3Ghz came into fruition but for right now i think we can assume that Apple wants to space out premieres of chips. they will probably stay on the 970 for atleast 2 more years. That probably means they make 970 on the .90nm process and move from there making other improvements along the way such as a improved vector unit (read hanibal's 970 analysis for more).
In late 2004 or 2005, when Intel makes an announcement of the P5... then i think we can start seeing Apple rev up for a fight. By that time i think it may benefit Apple to make some headlines with the rumored 980 vs. the P5. I think it is way too premature to say that the rumored 980 will be here next year at this time.
I think the perception that the POWER4 & POWER5 are truly different architectures is incorrect. My guess is that they started from the POWER4 and began making improvements. The POWER4->POWER5 transition (and hence 970->980) will be more akin to the PentiumPro -> PentiumII -> PentiumIII transitions. Bigger than the 604 -> 604e (that'll be like 970 -> 970+), but potentially much like the various PentiumIV revisions Intel has been doing.
I agree. My guess is that if (when) Apple begins using the 980s, they will still be branded G5.... especially if they come out next summer. The PowerMac G5 is created for upward mobility. They made that beast to be current for many years, just like El Capitan. They wouldn't change the brand name on a (by then) recognizable product. Nor would they change the design of a case (that's made for several years) that is merely one or two years old...just because there's a processor revision.
Can't even ship them? If they don't start shipping when Apple said they will, then you can make that comment. They had good reasons to pre-announce when they did.
Bollocks. They pre-announced for hype value. It looks like the sole purpose was to beat AMD to the 64-bit chip announcement. Well, all that counts in this game is shipping, so we'll see who ships first?
Quote:
Originally posted by Programmer
So if you're going to try to shoot down a story, at least use a reasonable argument which has some historical support behind it!
Apple has never shipped a new processor for a line of products within a year of shipping a new processor for that same line of products.
Your own post tells you the logic here. 970 in a Mac 18 months after Power4 (by your figures), Power5 not shipping yet, 980 in a Mac... 18 months after Power5?
Comments
Originally posted by Zapchud
Realtime Raytracing is something we do have already.
Realtime radiosity is something I look forward to
Real time ray-tracing, of what?
Originally posted by Bigc
Real time ray-tracing, of what?
Let's not turn this thread into a realtime raytracing-tread, but I do suggest that you try searching Google with these criteria: "realtime"+"raytracing"... lots of examples found easily...
Originally posted by Clive
This story is complete bollocks. Apple just launched a completely new range of machines (and can't even ship them) and you're contemplating that they're going to have another completely new range of chips within a year!?
Can't even ship them? If they don't start shipping when Apple said they will, then you can make that comment. They had good reasons to pre-announce when they did.
This story is not bollocks, at least not for the reason that you give. The 970 is based on the POWER4, so that is really 1.5 year old technology at this point. The 980 is based on the POWER5 which should arrive sometime in Q2 '04, and this story says that the 980 will follow the POWER5 closely which is possible because they've done the groundwork to scale down the POWER5s. The transition from 970 -> 980 will be roughly equivalent to the transition from 601 -> 604 which took from '94 to '95, or the 603e -> G3 which (IIRC) was about a year or year and half. Just because we've been stuck on the G4 for several years thanks to Motorola doesn't mean Apple and IBM can't deliver successive chip revisions on an aggressive timetable. Indeed, even Motorola wasn't as bad as everybody complains about because they delivered the 7400, 7410, 7450, 7455, and soon the 7457. The 7400->7450 transition was a fairly major redesign with the lengthening of the pipelines.
So if you're going to try to shoot down a story, at least use a reasonable argument which has some historical support behind it!
Originally posted by Zapchud
Of course, precise raytracing of complex scenes is another beast
Raytracing of anything other than complex scenes is nothing but a toy or benchmarking tool.
Originally posted by cuneglasus
This article cant even get the specs of the 970 correct.The 970 has 4 altivec units(simple integer,complex integer,floating point,and permute),not 2 as they claim.Dont believe a word of it.
Sorry, you're wrong. As it is now, the altivec IS split in two, one part for the Vector Permute Unit, and one part for the Vector Simple Integer, Complex Integer and Floating Point units.
The disadvantage is; since each of the two parts of the AV-unit can fetch 2 IOPs per clock per group to the Vector Issue Queues, and issue only one IOP to the execution-units per clock. This disables you from, say, executing anything simultaneously on the VCIU and VFPU.
Edit: With pic taken from arstechnica.com's excellent article on the 970.
The 970 was a redesign and adaptation of the Power4 and considerable reengineering was necessary. This took time. It seems possible that future Powerx and 9x0 CPUs will be introduced simultaneously.
This technical convergence of CPUs also suggests a marketing and OS convergence.
Heavy iron servers and workstations for big business from IBM and light iron servers, high performance desktops and portables for small business and individuals from Apple; all using OSX with both IBM and Apple systems using compatible software.
This could get interesting.
I believe the new 9x0 is being designed concurrently with the Power5. Should IBM adopt this rational design approach, the Power5 and 9x0 will be variations of the same design. This may reduce the time to market considerably.
It could get very interesting indeed!
People forget, such an approach would help IBM as much as it helps Apple and that it would surely be cheaper and more beneficial in 'time to market' if done in this 'simultaneous' way.
This 'rumour' should not be underestimated. For IBM? The gloves are off in its war with Intel. This goes deeper in some ways than the relationship with Apple. Intel are coming after IBM with the Itannic. I don't think IBM would be foolish to take that threat lying down.
We're so used to Intel and AMD being aggressive. Why not Apple and IBM? Now the cause is just, the reason sound. I don't see the 980 taking as long to market. The 970 is a clear indicator that Apple and IBM have forged a new alliance. Why would a 980 take another four years? IBM have indicated prototypes of the 'next gen' exist in their labs. Frightening. Time to market? Well, the Power 5 is hitting early/spring 04? IF it is...then maybe we can suppose that a 980 may only be a half year/year behind tops! Looks like it could be early 05 at the latest.
The time is ripe for PPC to put up or shut up. PPC is just a spring chicken compared to x86 land... Apple can't afford any more G4 debacles. Look what it did to G4 tower sales. Exactly!
Lemon Bon Bon
Oh, and it seems very likely that they can actually make it to a june '04-schedule. A 980 identical to the standard POWER5 (which implies that POWER5 actually has altivec), with the exeption of half the cores should be easy enough, compared to the POWER4->970-transition.
Originally posted by nebcon65
Hmmm 10ghz front side bus? Does anyone here know what's next after DDR400? Now that the memory is the bottleneck it would be interesting to see what we have coming for memory and when.
No 10GHz FSB, just an on-die memory-controller. And after DDR400, there's DDRII (Quad-pumped).
If Apple ships more than 1000 Dual 2 GHz G5 machines in August, I'll be really, really, really surprised.
Please, continue...
Originally posted by Locomotive
There is one problem with all us Mac-heads fanning the flames of these rumors toward one another. We tend to compare our beloved PowerPC processors -- which may be available some months or years in the future -- with the PC processors which are available in quantity today!
I always compare with what I've seen on roadmaps, or expect to be out the same time as the new, shiny, hairy mac with the new&glossy processor. Comparisions with nowaday models are useless, mostly
We'll hopefully see a 3Ghz PM the same day a 2Ghz PM is shipping this year.
We can analyze better once we see how the 3Ghz came into fruition but for right now i think we can assume that Apple wants to space out premieres of chips. they will probably stay on the 970 for atleast 2 more years. That probably means they make 970 on the .90nm process and move from there making other improvements along the way such as a improved vector unit (read hanibal's 970 analysis for more).
In late 2004 or 2005, when Intel makes an announcement of the P5... then i think we can start seeing Apple rev up for a fight. By that time i think it may benefit Apple to make some headlines with the rumored 980 vs. the P5. I think it is way too premature to say that the rumored 980 will be here next year at this time.
Originally posted by Programmer
I think the perception that the POWER4 & POWER5 are truly different architectures is incorrect. My guess is that they started from the POWER4 and began making improvements. The POWER4->POWER5 transition (and hence 970->980) will be more akin to the PentiumPro -> PentiumII -> PentiumIII transitions. Bigger than the 604 -> 604e (that'll be like 970 -> 970+), but potentially much like the various PentiumIV revisions Intel has been doing.
I agree. My guess is that if (when) Apple begins using the 980s, they will still be branded G5.... especially if they come out next summer. The PowerMac G5 is created for upward mobility. They made that beast to be current for many years, just like El Capitan. They wouldn't change the brand name on a (by then) recognizable product. Nor would they change the design of a case (that's made for several years) that is merely one or two years old...just because there's a processor revision.
Exciting stuff, though!
Originally posted by Programmer
Can't even ship them? If they don't start shipping when Apple said they will, then you can make that comment. They had good reasons to pre-announce when they did.
Bollocks. They pre-announced for hype value. It looks like the sole purpose was to beat AMD to the 64-bit chip announcement. Well, all that counts in this game is shipping, so we'll see who ships first?
Originally posted by Programmer
So if you're going to try to shoot down a story, at least use a reasonable argument which has some historical support behind it!
Apple has never shipped a new processor for a line of products within a year of shipping a new processor for that same line of products.
Your own post tells you the logic here. 970 in a Mac 18 months after Power4 (by your figures), Power5 not shipping yet, 980 in a Mac... 18 months after Power5?