Makes for an interesting thought, but if the 980 could achieve performance increases of 2 times or more!, do you think Apple would just say, Hey guys this PowerMac is a 3Gz G5, but its twice as fast as our shipping 2.5GHz - I doubt it. \
And consumers would probably be getting confused with usage of the G5 processor in other models, like the iMac or the PowerBook. "Is this slow, or fast G5??"
If changes are significant, and 970/980 architectures overlap each other in GHz-range, I hope for an easy way of differing them, as easy as G5 and G6.
If you look at how IBM is typically measuring this performance increase you'll see it revolves around things which are multi-threaded. Simply adding SMT to the POWER5 and doubling the clock rate will get the on the order of 3x improvement (over the 1.3 GHz POWER4 that they discuss). Add the FastPath stuff to avoid distracting the core(s), clean up any bottlenecks that 2 years of testing on the POWER4 showed up and blam!, you're running 4x faster./B]
Thanks for the info. I did not realize that some of the performance will be clock related.
Quote:
Originally posted by Programmer
Ya know, the guys over at Ars talk about how AltiVec was "tacked on" but I just don't get it. Yes they took the POWER4 and integrated a VMX unit into it, but in what way was it "tacked on"?
I myself moved from perpetual darkness (and perpetual f*****g rain) to Joy, Light and the odd Hurricane.
Well, in the Summer I know for certain that the UK has a much greater time between sun rise and sun set than Florida, so no perpertual darkness here. As to the rain, I quite like it. :-)
Quote:
Originally posted by fuzzylogic
As for fantasies, how do you figure I've got it wrong?
The fantasy is that within 12 months we're going to have another chip to replace the one that's not even shipping yet. I don't see any comments from me on your workflow.
Take the launch of the "G5". People were saying over a year ago that it would ship around August 2003, based on information that was pretty freely available and confirmable. That's being able to foresee the future with some degree of certainty.
All it takes is the right information and some thought.
Right now we've got people asking about G5 PowerBooks, the same kind of information says that these aren't going to happen until next year. I'll take bets on it if you doubt my powers as an oracle?
In the absence of hope/info, there's little to tie us all to the platform. No longer the case. whatever you may say, Clive.
Here's a bit of knowledge for you that I don't mind sharing:
Never place your trust in software or hardware vendors in terms of what they may or may not ship in the future.
You'll just end up on your arse.
To be really honest I don't care what Apple ships next year, as long as they are moving forward, with both hardware and OS. The tools I've going to be counting on to earn a living are the same ones I'm using today. If something new comes out then I'll take advantage of it, the same as anyone else would, but to plan your future on prayers is just foolish.
To be really honest I don't care what Apple ships next year, as long as they are moving forward, with both hardware and OS
Er.... I thought that was what I was saying. I'm really more concerned with the OS side of the equation at this point, now that the hardware has at least the appearance of legs for the next few years.
And no-one in the recording field could actually buy kit now whose use would be impaired by the non-appearance of a G6 in precisely 12 months. Could they?
Wrong, they changed the tower case with the intro of the 8600/9600, again with the G3 and again with the B&W G3. The 8600 was introduced in February 1997, the G3 in November 1997 and the B&W in Jan 1999 - three different cases in less than two years.
IIRC the 8600/9600 case was the same as the G3 MT (the green button on top) and the G3 DT was the same as the 7200 (and later derivatives). So, that makes one case in 1997 and a new one in 1999.
So, anyone fire up their Mayas yet and make mockups of the new G6 tower cases? Come on people, it's only 1 year away.
SiLK resins are a family of spin-on, low-k ILD materials that allow semiconductor manufacturers to increase the speed and performance of their chips. First-generation SiLK resin products offer k=2.6 for both copper damascene and aluminum/tungsten processing, a k-value reduction of more
than 35 percent versus conventional CVD oxide dielectric films. Dow offers a clear, extendible path to next-generation chip technologies with porous SiLK resin.
rickag: IIRC IBM is already using a low-k dieletric process at East Fishkill.
Programmer, regarding 970 scaling: you said you wouldn't expect more than 40-45% from the move to 0.09. 40-45% on the 2GHz 970 brings it to 2.8GHz - that's pretty damned close to the 3GHz number.
Since chips go through numerous small revisions in between major bumps, I don't see why a 970 with perhaps a slightly longer pipeline couldn't hit 3GHz at 0.09. IBM may have other process improvements in store for Fishkill too - remember, it just opened.
Regarding the 980: I wouldn't be expecting test silicon before this time next year. Linux has already booted on a very early POWER5 core, but I'm sure there's a lot of work left. As far as naming goes - was the 7400->7450 change enough to warrant dropping the G4 label? After all, Apollo was a fairly major redesign of the G4 core.
So for Apple hasn't upped their own marketing number except for fairly major changes - in the first G4, the introduction of AltiVec, and now all the changes in the G5. I doubt a 980 would be called a G6.
Apart from my weak attempt at some Antipodean humour a little earlier in this thread, I can't think of a good reason for Apple to not increment the names of their boxen with each new processor generation. In short, why wouldn't Apple call a 980 powered desktop a G6?
OK, I'll grant that the 7400 -> 7450 represented a major advance, but Motorola didn't label the 7450 a new generation CPU, so why would Apple? Obviously, there is a certain amount of fuzziness in the line between "generations" and mere "updates", but to date Apple seems to have followed the chip-makers' lead in their desktop nomenclature.
Programmer, regarding 970 scaling: you said you wouldn't expect more than 40-45% from the move to 0.09. 40-45% on the 2GHz 970 brings it to 2.8GHz - that's pretty damned close to the 3GHz number.
Close isn't 3 GHz.
Quote:
Since chips go through numerous small revisions in between major bumps, I don't see why a 970 with perhaps a slightly longer pipeline couldn't hit 3GHz at 0.09. IBM may have other process improvements in store for Fishkill too - remember, it just opened.
A change in pipeline length is not a minor revision. It would probably be enough to justify bumping the version number to 980 all by itself. Moto went from 7410 to 7450 when they did that, although Apple retained the label G4. Minor revisions typically are tweaks that are below the horizon of most non-technical users. Changing cache sizes, table sizes, fixing circuits with problematic timing, etc.
Quote:
Regarding the 980: I wouldn't be expecting test silicon before this time next year. Linux has already booted on a very early POWER5 core, but I'm sure there's a lot of work left. As far as naming goes - was the 7400->7450 change enough to warrant dropping the G4 label? After all, Apollo was a fairly major redesign of the G4 core.
So for Apple hasn't upped their own marketing number except for fairly major changes - in the first G4, the introduction of AltiVec, and now all the changes in the G5. I doubt a 980 would be called a G6.
Exactly. G4 -> G5 is a completely new chip. They are organized in completely different ways. They just happen to share the same instruction set and execution model, just like different cars share the same "user interface".
IIRC the 8600/9600 case was the same as the G3 MT (the green button on top) and the G3 DT was the same as the 7200 (and later derivatives). So, that makes one case in 1997 and a new one in 1999.
At a casual glance, yes, they look the same. But in fact the G3 case is shorter than the 8600. Still, the 8600 and B&W G3 cases were safely within two years of each other, which was previously said couldn't happen.
Actually he was being a bit pessimistic, 45% on 2GHz is 2.9GHz. It seems likely to me that if they can get that that gain from the process change, they are also likely to be able to get better yields and some process tweaks in to get to 3.0GHz.
From the bare figures so far they would only be 3% off, it seems remarkably unlikely that the G5 is at top speed in its first production run (G4 went from 350-1,400, G3 has gone from 233-1,000).
Depends on how many months you have to save up yields of 3GHz chips
Quote:
A change in pipeline length is not a minor revision. It would probably be enough to justify bumping the version number to 980 all by itself. Moto went from 7410 to 7450 when they did that, although Apple retained the label G4. Minor revisions typically are tweaks that are below the horizon of most non-technical users. Changing cache sizes, table sizes, fixing circuits with problematic timing, etc.
FWIW I meant "minor" as in "a change that's smaller than doubling the number of execution units on the core", as people are talking about the 980 as being. Didn't the 750FX take the G3 from a 5 to a 7 stage pipeline?
From the bare figures so far they would only be 3% off, it seems remarkably unlikely that the G5 is at top speed in its first production run (G4 went from 350-1,400, G3 has gone from 233-1,000).
Whoa, whoa, whoa. Just because Apple labels something as a G3 or a G4 doesn't mean that it's the same processor. The G4 has gone from 7400 to 7410 to 7450 to 7455 in that space, and the G3 has gone from 740 to 750 to 750CX to 750FX and now 750GX. These were not production runs of the same processor.
And you can't do a range based on lowest speed at introduction to highest speed now - you've gotta compare top speeds. Just a minor nitpick, but it makes it a little bit smaller.
With respect to Clive's feelings on the matter, my personal theory is that 980 could easily be announced in September 04, for shipment in professional desktops by November 04.
That give 970 a lifecycle of fifteen months in the professional line, but then diffuses that processor to the professional portable line (assuming a migration to 0.09) and the consumer line.
The professional desktop line needs to maintain performance parity with Intel's Tejas class which is due out in the second half of 2004, supporting DDR-II and a clock speed of 4.5 GHz. A 3.0 to 3.2 GHz 980 seems perfectly reasonable in that context.
I think we have to be very careful about being taken over by the euphoria of the 970 and the fact that the cork has been removed by the bottle, but also by constraining our imagination because Motorola could not find their collective ar*es using both hands and a mirror.
There truly is no point in using the last three years of pain and embarrasment as a guide for the future.
Here in England, I support a football team called Manchester United; for twenty five years, I suffered as our deadliest foes won everything in sight and we languished in horrendous medicrority; but since 1992, we have won the English league title more often than not (eight times I think) plus the FA Cup several times and the championship of Europe.
That hegemony will ultimately recede and be replaced with period of more modest achievement, and then there will be another period of ascendency and then depression and then who knows...
My point being is that all history teaches us is that things change, sometimes quickly, sometimes slowly, and sometimes impercepitably. As Mac users, we have had a period of the last of these, but could now be due for a period of almost unbelievably fast advances especially when placed in the context of the last three years.
I'm savouring the moment for now, and I'll wait for the future thanks very much.
I'm savouring the moment for now, and I'll wait for the future thanks very much.
Proving that not everybody from the UK is a miserable git.
((Mind you, I quite like rain too...no, really...it helps my lemon to grow...))
Nice post, Mark. A sound balance of past analysis and a cautious welcome to the future.
On a parallel development, there's no reason to think a 980 won't be here by early 05. Fall 04? That's about a year/year and a half from G5 announcement. Isn't that in line from the G3 to G4 transition? This is IBM now. They aren't starting these 9XX from scratch per se each time. That's surely got to cut down the time to market especially when you've learned already from doing the Power 4 to 970 cpu. It can only get quicker I'd have thought.
If Apple wants growth they're going to have to get aggressive with their cpu line. I'm quitely optimistic. It's up to Apple to get Panther out. Be aggressive on prices. Performance. Advertise. Increase unit sales. And get that 980 into the Mac line as soon as possible. Pile the pressure on the Wintel.
Personally, I think the 980 is gonna be a G6. The G5 will, I think, be aggressively pipelined into the consumer line and lower pro' line.
Originally posted by Mark- Card Carrying FanaticRealist
I'm savouring the moment for now, and I'll wait for the future thanks very much.
Yes. I really can't believe a serious pro makes platform decisions by "who is fastest now" logic. It is about software and the ability to upgrade adequately that is important to most shops with more money in big apps than in anything else. With Quark, whole houses will be upgrading and feeling just fine. Those that jumped ship because of 2 bad years...well they probably already jumped to InDesign anyway.
It is about the software....that is why Apple is still great in creative and crappy in the consumer sector.
[B]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?
Well, come on now, not JUST for hype value. When you are launching a major overhaul of your architecture you do need to let developers know about it before it ships to consumers. Otherwise, the software will lag way too far behind.
That being said, since announcing it to developers is going to let the cat out of the bag anyway, can you blame Apple for actualy hyping their achievement? It's called marketing, and companies either do it well or die.
Intel?s Pentium 4 processors have been showing excellent performance scalability for the last 18 months thanks to boosting megahertz as well as introduction of various internal improvements, such as 512KB L2 cache and the Hyper-Threading technology implementations. We ahve reported you many times already that Intel will continue this wise strategy and will gradually bring new technologies and features to the market along with newer microprocessor cores. Let us try to cast some light on what is believed to be the long-term Intel?s roadmap.
Later this year the world will witness the first PC microprocessors made using thin 90nm technology. The currently code-named Prescott CPUs will feature 800MHz Quad Pumped Bus, 1MB of L2 cache and the Hyper-Threading II technology in addition to new set of instructions known as PNI ? Prescott New Instructions that are proposed to accelerate processing of streams further. Apparently, the Prescott chips will live through 2004 and will be discontinued only in 2005. What is also very interesting to note is that there are discussions about implementing 1066MHz Quad Pumped Bus and new Prescott ?B? core to support the new PSB in early 2005. So, from the present point of view, the family of Prescott chips should look as follows:
Prescott 3.20GHz (800MHz QPB)
Prescott 3.40GHz (800MHz QPB)
Prescott 3.60GHz (800MHz QPB)
Prescott 3.80GHz (800MHz QPB)
Prescott 4.00GHz (800MHz QPB)
Prescott 4.20GHz (800MHz QPB)
Prescott 4.40GHz (800MHz QPB)
Prescott 4.60GHz (800MHz QPB)
Prescott 4.80GHz (800MHz QPB)
Prescott 4.80GHz (1066MHz QPB)
Prescott 5.00GHz (800MHz QPB)
Prescott 5.06GHz (1066MHz QPB)
Prescott 5.33GHz (1066MHz QPB)
The successor of the Prescott should be the code-named Tejas CPU, also the crown of the NetBurst architecture. The king Tejas should come to the market in late 2004 or early 2005. The novelty will again feature yet another set of new instructions (TNI, Tejas New Instuctions) as well as new incarnation of the Hyper-Threading technology. Firstly made using 90nm manufacturing process, the Tejas chips will be shrunk to 65nm technology in late 2005 or early 2006. The version manufactured using more advanced thinner process will also receive 2MB of L2 cache and 1200MHz Quad Pumped Bus in addition to higher core-clock.
Tejas 5.33GHz (L2=1MB, 1066MHz QPB)
Tejas 5.60GHz (L2=1MB, 1066MHz QPB)
Tejas 5.86GHz (L2=1MB, 1066MHz QPB)
Tejas 6.00GHz (L2=2MB, 1200MHz QPB)
Tejas 6.13GHz (L2=1MB, 1066MHz QPB)
Tejas 6.30GHz (L2=2MB, 1200MHz QPB)
Tejas 6.60GHz (L2=2MB, 1200MHz QPB)
Tejas 6.90GHz (L2=2MB, 1200MHz QPB)
Tejas 7.20GHz (L2=2MB, 1200MHz QPB)
After the Tejas, the brand-new Nehalem processor and architecture will be unveiled. But it is also a totally another story?
Prescott will arrive by Christmas at speeds of up to 3.6GHz. With a monster 1MB L2 cache at 3GHz+ and improved Hyperthreading, there's little doubt it will be the speed champ considering the dual 2.0GHz G5 has trouble keeping up with the 8-month-old 3.06GHz P4 (533MHZ FSB) on non-memory intensive tasks.
The good for Apple is that they'll have faster buses. The bad news is that Apple will still have to use the same RAM as PC manufacturers use and so memory speed will be the bottleneck. Already we have a 1GHz (effective bus) but are relegated to use the same memory Intel's 800MHz FSB motherboards use--dual channel DDR400.
Comments
Originally posted by hasapi
Makes for an interesting thought, but if the 980 could achieve performance increases of 2 times or more!, do you think Apple would just say, Hey guys this PowerMac is a 3Gz G5, but its twice as fast as our shipping 2.5GHz - I doubt it.
And consumers would probably be getting confused with usage of the G5 processor in other models, like the iMac or the PowerBook. "Is this slow, or fast G5??"
If changes are significant, and 970/980 architectures overlap each other in GHz-range, I hope for an easy way of differing them, as easy as G5 and G6.
Originally posted by Programmer
[B]
If you look at how IBM is typically measuring this performance increase you'll see it revolves around things which are multi-threaded. Simply adding SMT to the POWER5 and doubling the clock rate will get the on the order of 3x improvement (over the 1.3 GHz POWER4 that they discuss). Add the FastPath stuff to avoid distracting the core(s), clean up any bottlenecks that 2 years of testing on the POWER4 showed up and blam!, you're running 4x faster./B]
Thanks for the info. I did not realize that some of the performance will be clock related.
Originally posted by Programmer
Ya know, the guys over at Ars talk about how AltiVec was "tacked on" but I just don't get it. Yes they took the POWER4 and integrated a VMX unit into it, but in what way was it "tacked on"?
That is probably where I got that impression.
Originally posted by fuzzylogic
I myself moved from perpetual darkness (and perpetual f*****g rain) to Joy, Light and the odd Hurricane.
Well, in the Summer I know for certain that the UK has a much greater time between sun rise and sun set than Florida, so no perpertual darkness here. As to the rain, I quite like it. :-)
Originally posted by fuzzylogic
As for fantasies, how do you figure I've got it wrong?
The fantasy is that within 12 months we're going to have another chip to replace the one that's not even shipping yet. I don't see any comments from me on your workflow.
Originally posted by fuzzylogic
you're having a laugh, right?
No, you're just being disengenuous.
Take the launch of the "G5". People were saying over a year ago that it would ship around August 2003, based on information that was pretty freely available and confirmable. That's being able to foresee the future with some degree of certainty.
All it takes is the right information and some thought.
Right now we've got people asking about G5 PowerBooks, the same kind of information says that these aren't going to happen until next year. I'll take bets on it if you doubt my powers as an oracle?
Originally posted by fuzzylogic
In the absence of hope/info, there's little to tie us all to the platform. No longer the case. whatever you may say, Clive.
Here's a bit of knowledge for you that I don't mind sharing:
Never place your trust in software or hardware vendors in terms of what they may or may not ship in the future.
You'll just end up on your arse.
To be really honest I don't care what Apple ships next year, as long as they are moving forward, with both hardware and OS. The tools I've going to be counting on to earn a living are the same ones I'm using today. If something new comes out then I'll take advantage of it, the same as anyone else would, but to plan your future on prayers is just foolish.
Originally posted by Clive
To be really honest I don't care what Apple ships next year, as long as they are moving forward, with both hardware and OS
Er.... I thought that was what I was saying. I'm really more concerned with the OS side of the equation at this point, now that the hardware has at least the appearance of legs for the next few years.
And no-one in the recording field could actually buy kit now whose use would be impaired by the non-appearance of a G6 in precisely 12 months. Could they?
Originally posted by Clive
Wrong, they changed the tower case with the intro of the 8600/9600, again with the G3 and again with the B&W G3. The 8600 was introduced in February 1997, the G3 in November 1997 and the B&W in Jan 1999 - three different cases in less than two years.
IIRC the 8600/9600 case was the same as the G3 MT (the green button on top) and the G3 DT was the same as the 7200 (and later derivatives). So, that makes one case in 1997 and a new one in 1999.
So, anyone fire up their Mayas yet and make mockups of the new G6 tower cases? Come on people, it's only 1 year away.
kroehl
Unless of course it already uses this technology.
MORE information from the SiLKnet Alliance.
SiLK resins are a family of spin-on, low-k ILD materials that allow semiconductor manufacturers to increase the speed and performance of their chips. First-generation SiLK resin products offer k=2.6 for both copper damascene and aluminum/tungsten processing, a k-value reduction of more
than 35 percent versus conventional CVD oxide dielectric films. Dow offers a clear, extendible path to next-generation chip technologies with porous SiLK resin.
rickag: IIRC IBM is already using a low-k dieletric process at East Fishkill.
Programmer, regarding 970 scaling: you said you wouldn't expect more than 40-45% from the move to 0.09. 40-45% on the 2GHz 970 brings it to 2.8GHz - that's pretty damned close to the 3GHz number.
Since chips go through numerous small revisions in between major bumps, I don't see why a 970 with perhaps a slightly longer pipeline couldn't hit 3GHz at 0.09. IBM may have other process improvements in store for Fishkill too - remember, it just opened.
Regarding the 980: I wouldn't be expecting test silicon before this time next year. Linux has already booted on a very early POWER5 core, but I'm sure there's a lot of work left. As far as naming goes - was the 7400->7450 change enough to warrant dropping the G4 label? After all, Apollo was a fairly major redesign of the G4 core.
So for Apple hasn't upped their own marketing number except for fairly major changes - in the first G4, the introduction of AltiVec, and now all the changes in the G5. I doubt a 980 would be called a G6.
OK, I'll grant that the 7400 -> 7450 represented a major advance, but Motorola didn't label the 7450 a new generation CPU, so why would Apple? Obviously, there is a certain amount of fuzziness in the line between "generations" and mere "updates", but to date Apple seems to have followed the chip-makers' lead in their desktop nomenclature.
Originally posted by Anonymous Karma
Programmer, regarding 970 scaling: you said you wouldn't expect more than 40-45% from the move to 0.09. 40-45% on the 2GHz 970 brings it to 2.8GHz - that's pretty damned close to the 3GHz number.
Close isn't 3 GHz.
Since chips go through numerous small revisions in between major bumps, I don't see why a 970 with perhaps a slightly longer pipeline couldn't hit 3GHz at 0.09. IBM may have other process improvements in store for Fishkill too - remember, it just opened.
A change in pipeline length is not a minor revision. It would probably be enough to justify bumping the version number to 980 all by itself. Moto went from 7410 to 7450 when they did that, although Apple retained the label G4. Minor revisions typically are tweaks that are below the horizon of most non-technical users. Changing cache sizes, table sizes, fixing circuits with problematic timing, etc.
Regarding the 980: I wouldn't be expecting test silicon before this time next year. Linux has already booted on a very early POWER5 core, but I'm sure there's a lot of work left. As far as naming goes - was the 7400->7450 change enough to warrant dropping the G4 label? After all, Apollo was a fairly major redesign of the G4 core.
So for Apple hasn't upped their own marketing number except for fairly major changes - in the first G4, the introduction of AltiVec, and now all the changes in the G5. I doubt a 980 would be called a G6.
Exactly. G4 -> G5 is a completely new chip. They are organized in completely different ways. They just happen to share the same instruction set and execution model, just like different cars share the same "user interface".
Originally posted by kroehl
IIRC the 8600/9600 case was the same as the G3 MT (the green button on top) and the G3 DT was the same as the 7200 (and later derivatives). So, that makes one case in 1997 and a new one in 1999.
At a casual glance, yes, they look the same. But in fact the G3 case is shorter than the 8600. Still, the 8600 and B&W G3 cases were safely within two years of each other, which was previously said couldn't happen.
Originally posted by Programmer
Close isn't 3 GHz.
Actually he was being a bit pessimistic, 45% on 2GHz is 2.9GHz. It seems likely to me that if they can get that that gain from the process change, they are also likely to be able to get better yields and some process tweaks in to get to 3.0GHz.
From the bare figures so far they would only be 3% off, it seems remarkably unlikely that the G5 is at top speed in its first production run (G4 went from 350-1,400, G3 has gone from 233-1,000).
Originally posted by Programmer
Close isn't 3 GHz.
Depends on how many months you have to save up yields of 3GHz chips
A change in pipeline length is not a minor revision. It would probably be enough to justify bumping the version number to 980 all by itself. Moto went from 7410 to 7450 when they did that, although Apple retained the label G4. Minor revisions typically are tweaks that are below the horizon of most non-technical users. Changing cache sizes, table sizes, fixing circuits with problematic timing, etc.
FWIW I meant "minor" as in "a change that's smaller than doubling the number of execution units on the core", as people are talking about the 980 as being. Didn't the 750FX take the G3 from a 5 to a 7 stage pipeline?
Originally posted by Clive
From the bare figures so far they would only be 3% off, it seems remarkably unlikely that the G5 is at top speed in its first production run (G4 went from 350-1,400, G3 has gone from 233-1,000).
Whoa, whoa, whoa. Just because Apple labels something as a G3 or a G4 doesn't mean that it's the same processor. The G4 has gone from 7400 to 7410 to 7450 to 7455 in that space, and the G3 has gone from 740 to 750 to 750CX to 750FX and now 750GX. These were not production runs of the same processor.
And you can't do a range based on lowest speed at introduction to highest speed now - you've gotta compare top speeds. Just a minor nitpick, but it makes it a little bit smaller.
That give 970 a lifecycle of fifteen months in the professional line, but then diffuses that processor to the professional portable line (assuming a migration to 0.09) and the consumer line.
The professional desktop line needs to maintain performance parity with Intel's Tejas class which is due out in the second half of 2004, supporting DDR-II and a clock speed of 4.5 GHz. A 3.0 to 3.2 GHz 980 seems perfectly reasonable in that context.
I think we have to be very careful about being taken over by the euphoria of the 970 and the fact that the cork has been removed by the bottle, but also by constraining our imagination because Motorola could not find their collective ar*es using both hands and a mirror.
There truly is no point in using the last three years of pain and embarrasment as a guide for the future.
Here in England, I support a football team called Manchester United; for twenty five years, I suffered as our deadliest foes won everything in sight and we languished in horrendous medicrority; but since 1992, we have won the English league title more often than not (eight times I think) plus the FA Cup several times and the championship of Europe.
That hegemony will ultimately recede and be replaced with period of more modest achievement, and then there will be another period of ascendency and then depression and then who knows...
My point being is that all history teaches us is that things change, sometimes quickly, sometimes slowly, and sometimes impercepitably. As Mac users, we have had a period of the last of these, but could now be due for a period of almost unbelievably fast advances especially when placed in the context of the last three years.
I'm savouring the moment for now, and I'll wait for the future thanks very much.
I'm savouring the moment for now, and I'll wait for the future thanks very much.
Proving that not everybody from the UK is a miserable git.
Nice post, Mark. A sound balance of past analysis and a cautious welcome to the future.
On a parallel development, there's no reason to think a 980 won't be here by early 05. Fall 04? That's about a year/year and a half from G5 announcement. Isn't that in line from the G3 to G4 transition? This is IBM now. They aren't starting these 9XX from scratch per se each time. That's surely got to cut down the time to market especially when you've learned already from doing the Power 4 to 970 cpu. It can only get quicker I'd have thought.
If Apple wants growth they're going to have to get aggressive with their cpu line. I'm quitely optimistic. It's up to Apple to get Panther out. Be aggressive on prices. Performance. Advertise. Increase unit sales. And get that 980 into the Mac line as soon as possible. Pile the pressure on the Wintel.
Personally, I think the 980 is gonna be a G6. The G5 will, I think, be aggressively pipelined into the consumer line and lower pro' line.
The furture's bright, the future's Apple.
Lemon Bon Bon
Originally posted by Mark- Card Carrying FanaticRealist
I'm savouring the moment for now, and I'll wait for the future thanks very much.
Yes. I really can't believe a serious pro makes platform decisions by "who is fastest now" logic. It is about software and the ability to upgrade adequately that is important to most shops with more money in big apps than in anything else. With Quark, whole houses will be upgrading and feeling just fine. Those that jumped ship because of 2 bad years...well they probably already jumped to InDesign anyway.
It is about the software....that is why Apple is still great in creative and crappy in the consumer sector.
Originally posted by Clive
[B]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?
Well, come on now, not JUST for hype value. When you are launching a major overhaul of your architecture you do need to let developers know about it before it ships to consumers. Otherwise, the software will lag way too far behind.
That being said, since announcing it to developers is going to let the cat out of the bag anyway, can you blame Apple for actualy hyping their achievement? It's called marketing, and companies either do it well or die.
Intel?s Pentium 4 processors have been showing excellent performance scalability for the last 18 months thanks to boosting megahertz as well as introduction of various internal improvements, such as 512KB L2 cache and the Hyper-Threading technology implementations. We ahve reported you many times already that Intel will continue this wise strategy and will gradually bring new technologies and features to the market along with newer microprocessor cores. Let us try to cast some light on what is believed to be the long-term Intel?s roadmap.
Later this year the world will witness the first PC microprocessors made using thin 90nm technology. The currently code-named Prescott CPUs will feature 800MHz Quad Pumped Bus, 1MB of L2 cache and the Hyper-Threading II technology in addition to new set of instructions known as PNI ? Prescott New Instructions that are proposed to accelerate processing of streams further. Apparently, the Prescott chips will live through 2004 and will be discontinued only in 2005. What is also very interesting to note is that there are discussions about implementing 1066MHz Quad Pumped Bus and new Prescott ?B? core to support the new PSB in early 2005. So, from the present point of view, the family of Prescott chips should look as follows:
Prescott 3.20GHz (800MHz QPB)
Prescott 3.40GHz (800MHz QPB)
Prescott 3.60GHz (800MHz QPB)
Prescott 3.80GHz (800MHz QPB)
Prescott 4.00GHz (800MHz QPB)
Prescott 4.20GHz (800MHz QPB)
Prescott 4.40GHz (800MHz QPB)
Prescott 4.60GHz (800MHz QPB)
Prescott 4.80GHz (800MHz QPB)
Prescott 4.80GHz (1066MHz QPB)
Prescott 5.00GHz (800MHz QPB)
Prescott 5.06GHz (1066MHz QPB)
Prescott 5.33GHz (1066MHz QPB)
The successor of the Prescott should be the code-named Tejas CPU, also the crown of the NetBurst architecture. The king Tejas should come to the market in late 2004 or early 2005. The novelty will again feature yet another set of new instructions (TNI, Tejas New Instuctions) as well as new incarnation of the Hyper-Threading technology. Firstly made using 90nm manufacturing process, the Tejas chips will be shrunk to 65nm technology in late 2005 or early 2006. The version manufactured using more advanced thinner process will also receive 2MB of L2 cache and 1200MHz Quad Pumped Bus in addition to higher core-clock.
Tejas 5.33GHz (L2=1MB, 1066MHz QPB)
Tejas 5.60GHz (L2=1MB, 1066MHz QPB)
Tejas 5.86GHz (L2=1MB, 1066MHz QPB)
Tejas 6.00GHz (L2=2MB, 1200MHz QPB)
Tejas 6.13GHz (L2=1MB, 1066MHz QPB)
Tejas 6.30GHz (L2=2MB, 1200MHz QPB)
Tejas 6.60GHz (L2=2MB, 1200MHz QPB)
Tejas 6.90GHz (L2=2MB, 1200MHz QPB)
Tejas 7.20GHz (L2=2MB, 1200MHz QPB)
After the Tejas, the brand-new Nehalem processor and architecture will be unveiled. But it is also a totally another story?
Prescott will arrive by Christmas at speeds of up to 3.6GHz. With a monster 1MB L2 cache at 3GHz+ and improved Hyperthreading, there's little doubt it will be the speed champ considering the dual 2.0GHz G5 has trouble keeping up with the 8-month-old 3.06GHz P4 (533MHZ FSB) on non-memory intensive tasks.
The good for Apple is that they'll have faster buses. The bad news is that Apple will still have to use the same RAM as PC manufacturers use and so memory speed will be the bottleneck. Already we have a 1GHz (effective bus) but are relegated to use the same memory Intel's 800MHz FSB motherboards use--dual channel DDR400.