Did I miss something and IBM has even mentioned the number "980"? Last I heard it was entirely rumour and speculation, but people seem to be treating it an awful lot like fact. The POWER5 is fact, but that is entirely a different animal and it isn't even out yet -- and won't be until next year so assuming that the 980 will arrive shortly thereafter seems aggressive.
Or did I miss an official roadmap from IBM?
John Kelly said something about IBM having already built the prototypes for the "next generation" of PPC chips (does this mean the G6??) in the G5 intro video - I guess it depends on your reading of his comment - is he referring to (1) a boost in speed for the 970; or (2) a 90NM 970; or (3) the 980?
Moore's law is is a statement about the economics of integrated circuit manufacture and states absolutely nothing about performance. Transistors do not necessarily directly relate to performance or clock speed.
I did not say different.
However, in the x86 camp, the advance in clockspeed has been faster than 2x in 18 month for some time now.
I disagree. The 970 isn't interim anything, and it should scale quite well, especially with the move to 90nm. I don't expect to see the 980 in anything (ie, a shipping Apple box) until at least 2005, likely 2006.
Once again, I find moki and me in complete agreement. How nice for a change The way the 970 has already exceeded initial expectations, I see this chip scale further than most of us anticipate now.
Once again, I find moki and me in complete agreement. How nice for a change The way the 970 has already exceeded initial expectations, I see this chip scale further than most of us anticipate now.
G-News
The 980 was laid out in some "roadmap" PDFs from IBM that were on the web at one point. I do not know if there are still there or not. In any event, if the roadmaps are to be believed, the 980 is a real project and is underway with planned introduction in '04. Whether Apple will adopt it is open to question, but IBM has stated their intention to do so.
Although the 970+ provides some of the process benefits that were initially thought to be first incorporated in the 980, the architectual changes in the 980 core are still significant, according to various reports.
Luckily, I never said the 980 wasn't real. I just said that it might take longer to make it into a PowerMac, than some of you seem to think (ie Xmas 2003 or so).
John Kelly said something about IBM having already built the prototypes for the "next generation" of PPC chips (does this mean the G6??) in the G5 intro video
I noticed that too. I tend to believe that "next generation" means something substantially different from the 970. And in the current IBM offerings, this should correspond to a POWER5 derivative (980?).
On the other hand, this does not prove that we will see something like the 980 next year nor that the 970 will have so short life. If I remember well, Jobs said that this chip "has legs".
But it is good anyway to hear that IBM already works on the next generation chip.
While Googling, I found the following. Some of it conflicts, some confirms, but then of course I haven't found anything directly from IBM yet.
Quote:
- Power5/PowerPC980 work is on target.
- Power5 due later this year, PowerPC 980 in Q3 2004 @ 90nm process.
-PowerPC 980 (4.5GHz-5GHz top speeds)
- In 2006: Power 6, PowerPC 990. 990 is targeted at 8GHz.
OSNews Forums:
Quote:
If IBM is really commited to it's new 9x0 line, expect a single core PowerPC 980 chip a few months after the intoduction of the POWER5 (rumored around April/May 2004).
The POWER5 will have 2 new main features compared to the POWER4+:
- SMT (Simultaneous MultiThreading) and new caches (more associative)
- FastPath, hardware assist to do some high level tasks (TCP/IP processing, table walks...).
Geek.com
Quote:
Future IBM PowerPC chips include the Power5-based PowerPC 980 at around 3GHz arriving in Q3 2004, and in 2006 the Power6-based PowerPC 990.
Going from 2 to 3 Ghz in 12 month sounds impressive - unless you consider that Moores law predicts a doubling of transistor density (translating in a doubling of chip speed) every 18 month.
Completely untrue -- Moore's Law, as you say, relates to transistor density which does not correlate to clock rate. I repeat, it does not correlate to clock rate. Shall I repeat it again?
Quote:
IBM does not seem to be able to achieve this. With their predicted rate, they only gain 75% over the next 18 month. As far as I know, intel cpus had seen an even faster gain in clockspeed over the last years.
Well you don't know very much then. The clock rate gains of Intel CPUs are an anomoly anyhow -- Intel's Pentium4 has been pushing clock rate to get performance, and that is not necessarily the best way to go about it. Even Intel has been backing off the advance of clock rate in favour of other ways to make it faster. The 2->3 GHz improvement in the G5 may result in better performance than you'd expect from a straight 50% jump in clock rate (which is quite a bit less than a 50% improvement in performance, by the way). If they increase the transistor count they can improve caches, internal tables, add more execution units, add more rename registers, add more dispatch groups, etc etc etc. Or they'll just use the improved process to reduce power consumption & heat generation.
Get it out of your head that clock rate == performance. This is marketing nonsense intended to deceive the consumer into buying a machine with a bigger number. Getting better performance is a far more complex issue than just looking at the clock rate.
I noticed that too. I tend to believe that "next generation" means something substantially different from the 970. And in the current IBM offerings, this should correspond to a POWER5 derivative (980?).
On the other hand, this does not prove that we will see something like the 980 next year nor that the 970 will have so short life. If I remember well, Jobs said that this chip "has legs".
It's also important to remember that the introduction of the 980 does not spell curtains for the 970. It can drop down to second best and continue to get smaller and cheaper for a couple more process generations.
Completely untrue -- Moore's Law, as you say, relates to transistor density which does not correlate to clock rate. I repeat, it does not correlate to clock rate. Shall I repeat it again?
[..]The clock rate gains of Intel CPUs are an anomoly anyhow -- Intel's Pentium4 has been pushing clock rate to get performance,
No more questions, thanks.
So intel *has* been able to more than double the clockspeed on its chips in the last two 18-month periods, right?
And I was referring specially to clockspeed (as opposed to spec-performance) going from 2 to 3Ghz. Shall I repeat? C L O C K S P E E D - I was talking clockspeed. The speed of the quartz system clock generator. Clockspeed right?
Just spare me your lectures on system performance vs. clockspeed, I know that.
The Apple-apologists have been reiterating the "clockspeed is irrelevant"-mantra for years now while intel has happily built wasteful designs that blew the socks off the carefully constructed but low-in-clockspeed chips Moto got. I sincerely hope IBM does better, but the "50% in 12 month" doesn't sound like it.
We just got the 970, so waiting expectantly for the 980 at this point is ridiculously premature.
Do note however that IBM coming out with the 980 this time next year would not make the 970 obsolete. Not any more than the G4 coming out 3 years ago made the G3 obsolete. Five years later, we're still using it in the iBook.
The 980 coming out next year would allow Apple to put it in the PowerMacs and have 970's across the board in every machine, including even the eMac and iBook.
The 970 isn't going to stop being used in Apple's machines for some time to come...980 or not. But at least now I think the "Apple using Intel" rumors should be put to rest. At least not as a replacement to their PPC line of machines. If at all, some sort of separate branch.
What a devilishly loaded statement. It seems he meant "All the options IBM is offering." The 970 with its branch of advances, Gobi lurking in the wings and the 980 in 12 to 18 months (maybe).
So intel *has* been able to more than double the clockspeed on its chips in the last two 18-month periods, right?
No, 18 months ago Intel released a 2.2GHz P4 (January 2002). Are they at 4.4GHz yet? They've not even reached a 50% increase yet.
Quote:
Originally posted by Smircle
And I was referring specially to clockspeed (as opposed to spec-performance) going from 2 to 3Ghz. Shall I repeat? C L O C K S P E E D - I was talking clockspeed. The speed of the quartz system clock generator. Clockspeed right?
Then why did you mention Moore's Law then?
Quote:
Originally posted by Smircle
I sincerely hope IBM does better, but the "50% in 12 month" doesn't sound like it.
What a devilishly loaded statement. It seems he meant "All the options IBM is offering." The 970 with its branch of advances, Gobi lurking in the wings and the 980 in 12 to 18 months (maybe).
Yes, options all around.
Screed
I'm sure that he ment more than that, and it was a "warning" call to both IBM and Moto. The portability of OS X does give them options, and more it is adopted in the Mac market the more powerfull that bargining chip becomes. I'm not saying there is a plan to move to Intel, AMD or some other chip, but if Apple didnt have a contigincy plan in place for it I would be suprised.
And I was referring specially to clockspeed (as opposed to spec-performance) going from 2 to 3Ghz. Shall I repeat? C L O C K S P E E D - I was talking clockspeed. The speed of the quartz system clock generator. Clockspeed right?
Please! Please! This is supposed to be a happy occasion! Let's not bicker and argue about 'oo outclocked 'oo!
Comments
Feb: 2.0, 2.4, 2.6.
August: 2.6, 3.0, 3.2...
Originally posted by Programmer
Did I miss something and IBM has even mentioned the number "980"? Last I heard it was entirely rumour and speculation, but people seem to be treating it an awful lot like fact. The POWER5 is fact, but that is entirely a different animal and it isn't even out yet -- and won't be until next year so assuming that the 980 will arrive shortly thereafter seems aggressive.
Or did I miss an official roadmap from IBM?
John Kelly said something about IBM having already built the prototypes for the "next generation" of PPC chips (does this mean the G6??) in the G5 intro video - I guess it depends on your reading of his comment - is he referring to (1) a boost in speed for the 970; or (2) a 90NM 970; or (3) the 980?
Originally posted by Telomar
Moore's law is is a statement about the economics of integrated circuit manufacture and states absolutely nothing about performance. Transistors do not necessarily directly relate to performance or clock speed.
I did not say different.
However, in the x86 camp, the advance in clockspeed has been faster than 2x in 18 month for some time now.
Originally posted by Smircle
I did not say different.
However, in the x86 camp, the advance in clockspeed has been faster than 2x in 18 month for some time now.
Advance in clockspeed /= faster machines.
I disagree. The 970 isn't interim anything, and it should scale quite well, especially with the move to 90nm. I don't expect to see the 980 in anything (ie, a shipping Apple box) until at least 2005, likely 2006.
Once again, I find moki and me in complete agreement. How nice for a change
G-News
Originally posted by G-News
Once again, I find moki and me in complete agreement. How nice for a change
G-News
The 980 was laid out in some "roadmap" PDFs from IBM that were on the web at one point. I do not know if there are still there or not. In any event, if the roadmaps are to be believed, the 980 is a real project and is underway with planned introduction in '04. Whether Apple will adopt it is open to question, but IBM has stated their intention to do so.
Although the 970+ provides some of the process benefits that were initially thought to be first incorporated in the 980, the architectual changes in the 980 core are still significant, according to various reports.
Time will tell.
Originally posted by McCrab
John Kelly said something about IBM having already built the prototypes for the "next generation" of PPC chips (does this mean the G6??) in the G5 intro video
I noticed that too. I tend to believe that "next generation" means something substantially different from the 970. And in the current IBM offerings, this should correspond to a POWER5 derivative (980?).
On the other hand, this does not prove that we will see something like the 980 next year nor that the 970 will have so short life. If I remember well, Jobs said that this chip "has legs".
But it is good anyway to hear that IBM already works on the next generation chip.
- Power5/PowerPC980 work is on target.
- Power5 due later this year, PowerPC 980 in Q3 2004 @ 90nm process.
-PowerPC 980 (4.5GHz-5GHz top speeds)
- In 2006: Power 6, PowerPC 990. 990 is targeted at 8GHz.
OSNews Forums:
If IBM is really commited to it's new 9x0 line, expect a single core PowerPC 980 chip a few months after the intoduction of the POWER5 (rumored around April/May 2004).
The POWER5 will have 2 new main features compared to the POWER4+:
- SMT (Simultaneous MultiThreading) and new caches (more associative)
- FastPath, hardware assist to do some high level tasks (TCP/IP processing, table walks...).
Geek.com
Future IBM PowerPC chips include the Power5-based PowerPC 980 at around 3GHz arriving in Q3 2004, and in 2006 the Power6-based PowerPC 990.
Screed
Originally posted by Smircle
Going from 2 to 3 Ghz in 12 month sounds impressive - unless you consider that Moores law predicts a doubling of transistor density (translating in a doubling of chip speed) every 18 month.
Completely untrue -- Moore's Law, as you say, relates to transistor density which does not correlate to clock rate. I repeat, it does not correlate to clock rate. Shall I repeat it again?
Quote:
IBM does not seem to be able to achieve this. With their predicted rate, they only gain 75% over the next 18 month. As far as I know, intel cpus had seen an even faster gain in clockspeed over the last years.
Well you don't know very much then. The clock rate gains of Intel CPUs are an anomoly anyhow -- Intel's Pentium4 has been pushing clock rate to get performance, and that is not necessarily the best way to go about it. Even Intel has been backing off the advance of clock rate in favour of other ways to make it faster. The 2->3 GHz improvement in the G5 may result in better performance than you'd expect from a straight 50% jump in clock rate (which is quite a bit less than a 50% improvement in performance, by the way). If they increase the transistor count they can improve caches, internal tables, add more execution units, add more rename registers, add more dispatch groups, etc etc etc. Or they'll just use the improved process to reduce power consumption & heat generation.
Get it out of your head that clock rate == performance. This is marketing nonsense intended to deceive the consumer into buying a machine with a bigger number. Getting better performance is a far more complex issue than just looking at the clock rate.
Originally posted by PB
I noticed that too. I tend to believe that "next generation" means something substantially different from the 970. And in the current IBM offerings, this should correspond to a POWER5 derivative (980?).
On the other hand, this does not prove that we will see something like the 980 next year nor that the 970 will have so short life. If I remember well, Jobs said that this chip "has legs".
It's also important to remember that the introduction of the 980 does not spell curtains for the 970. It can drop down to second best and continue to get smaller and cheaper for a couple more process generations.
IIRC, IBM still sells POWER3s.
Originally posted by Programmer
Completely untrue -- Moore's Law, as you say, relates to transistor density which does not correlate to clock rate. I repeat, it does not correlate to clock rate. Shall I repeat it again?
[..]The clock rate gains of Intel CPUs are an anomoly anyhow -- Intel's Pentium4 has been pushing clock rate to get performance,
No more questions, thanks.
So intel *has* been able to more than double the clockspeed on its chips in the last two 18-month periods, right?
And I was referring specially to clockspeed (as opposed to spec-performance) going from 2 to 3Ghz. Shall I repeat? C L O C K S P E E D - I was talking clockspeed. The speed of the quartz system clock generator. Clockspeed right?
Just spare me your lectures on system performance vs. clockspeed, I know that.
The Apple-apologists have been reiterating the "clockspeed is irrelevant"-mantra for years now while intel has happily built wasteful designs that blew the socks off the carefully constructed but low-in-clockspeed chips Moto got. I sincerely hope IBM does better, but the "50% in 12 month" doesn't sound like it.
Thanks for reading my postings carfully.
Do note however that IBM coming out with the 980 this time next year would not make the 970 obsolete. Not any more than the G4 coming out 3 years ago made the G3 obsolete. Five years later, we're still using it in the iBook.
The 980 coming out next year would allow Apple to put it in the PowerMacs and have 970's across the board in every machine, including even the eMac and iBook.
The 970 isn't going to stop being used in Apple's machines for some time to come...980 or not. But at least now I think the "Apple using Intel" rumors should be put to rest. At least not as a replacement to their PPC line of machines. If at all, some sort of separate branch.
-- Ensoniq
What a devilishly loaded statement. It seems he meant "All the options IBM is offering." The 970 with its branch of advances, Gobi lurking in the wings and the 980 in 12 to 18 months (maybe).
Yes, options all around.
Screed
Buy at Dual 2GHz now, up it to dual 3Ghz next year? That'd be good.
Originally posted by Smircle
So intel *has* been able to more than double the clockspeed on its chips in the last two 18-month periods, right?
No, 18 months ago Intel released a 2.2GHz P4 (January 2002). Are they at 4.4GHz yet? They've not even reached a 50% increase yet.
Originally posted by Smircle
And I was referring specially to clockspeed (as opposed to spec-performance) going from 2 to 3Ghz. Shall I repeat? C L O C K S P E E D - I was talking clockspeed. The speed of the quartz system clock generator. Clockspeed right?
Then why did you mention Moore's Law then?
Originally posted by Smircle
I sincerely hope IBM does better, but the "50% in 12 month" doesn't sound like it.
It's better than what Intel have done.
Originally posted by sCreeD
"We like to have options." --Steve Jobs
What a devilishly loaded statement. It seems he meant "All the options IBM is offering." The 970 with its branch of advances, Gobi lurking in the wings and the 980 in 12 to 18 months (maybe).
Yes, options all around.
Screed
I'm sure that he ment more than that, and it was a "warning" call to both IBM and Moto. The portability of OS X does give them options, and more it is adopted in the Mac market the more powerfull that bargining chip becomes. I'm not saying there is a plan to move to Intel, AMD or some other chip, but if Apple didnt have a contigincy plan in place for it I would be suprised.
Originally posted by Smircle
And I was referring specially to clockspeed (as opposed to spec-performance) going from 2 to 3Ghz. Shall I repeat? C L O C K S P E E D - I was talking clockspeed. The speed of the quartz system clock generator. Clockspeed right?
Please! Please! This is supposed to be a happy occasion! Let's not bicker and argue about 'oo outclocked 'oo!