<strong>thats a Moto upgrade schedule. I think you're going to see .09u sooner than 2005. IBM will move aggressively to .09u manufacturing.</strong><hr></blockquote>
Agreed, East FishKill is set up for .09u. I'm not sure but I think that IBM indicatied that they will move to .09u withing 6 months of the initial release of the 970, but I would imagine no later than 12 months. IBM wants the 970 to be successfull, and they will push it as hard as possible. Weather Apple can get ahold of the fastest chips right away is left to be seen, but I think things look pretty good for 970 PM's shipping within 6 months.
Agreed, East FishKill is set up for .09u. I'm not sure but I think that IBM indicatied that they will move to .09u withing 6 months of the initial release of the 970, but I would imagine no later than 12 months. IBM wants the 970 to be successfull, and they will push it as hard as possible. Weather Apple can get ahold of the fastest chips right away is left to be seen, but I think things look pretty good for 970 PM's shipping within 6 months.</strong><hr></blockquote>
IBM said that they would take the 970 to a .09um in early 04 starting at 2ghz ant that was before the fubar that the German site let slip.
[quote]Should Apple move from 32-bit PPC to 64-bit PPC, Mac users should not expect the same kinds of performance gains that x86 software sees from the jump to x86-64. 64-bit PPC gives you larger integers and more memory, and that's about it. There are no extra registers, no cleaned up addressing scheme, etc. <hr></blockquote>
Anyone else get the impression that Hannibal is just another Wintel shill? I'm not sure I trust those guys over at Ars.. That's why I just sit and lurk... I mean just look at that statement. It's TOO one-sided and someone should call him on it. 64/32-bit was built right into the PPC ISA from the get-go. X86 will just be ANOTHER HACK. Period. Not to mention that BOTH Intel and Micor$oft see no chance of Windows/Intel moving to 64-bit desktops until at least the end of the decade. I've attempted to explain the debacle <a href="http://forums.appleinsider.com/cgi-bin/ultimatebb.cgi?ubb=get_topic&f=1&t=003098"; target="_blank">here.</a>
Another glaring point he CONVENIENTLY forgot to mention was that if all things being *equal* he statement about what Mac users could expect *might* be true, however, as we know, things are far from equal and the PPC 970 is NOT the G4 simply fattened up to 64-bits. It's a completely different design with added functionality and enhancements. I can't imagine what to expect from AltiVec and the additional bandwidth alone.. Or how about that extra FPU ... and so on. Imagine that in an SMP environment etc.... Yeah, he's missed a few points. Anyone care to add anything?
Should Apple move from 32-bit PPC to 64-bit PPC, Mac users should not expect the same kinds of performance gains that x86 software sees from the jump to x86-64. 64-bit PPC gives you larger integers and more memory, and that's about it. There are no extra registers, no cleaned up addressing scheme, etc.<hr></blockquote>
[quote] Anyone else get the impression that Hannibal is just another Wintel shill? <hr></blockquote>
Say what? Hannibal has produced perhaps the most insightful piece on the design of the 970 ever. While he's been stymied in getting more information on it from IBM (still waiting from part 2 of his article), calling him a Wintel shill is ridiculous.
His comment makes a huge amount of sense. The AMD effort will improve the x86 by fixing all sorts of shortcomings by tweaking the 64-bit instructions. The PPC instruction set, having been created by design rather than added to haphazardly, doesn't need and won't get the overhaul.
[quote]64/32-bit was built right into the PPC ISA from the get-go. X86 will just be ANOTHER HACK. Period.<hr></blockquote>
One thing that real life has taught be is that a working hack is how most of the world works. A nice clean design to start out with is a good thing (TM), but the most successful products (hardware, software, human institutions) are merely agglomerations of cruft, each piece of cruft optimized for the time it was released with little thought to the future. Suprisingly enough, the technique works.
As an example, take the fact that OS X was built on NeXT OS rather than BEOS. I know a number of people who were disappointed with the fact that Apple would be adding cruft to an old design rather than using a really fresh, aesthetically pleasing new design.
The reality is that, as many have found out to their horror, rewriting a big crufty software horror from scratch almost always results in a *vastly* more buggy piece of software than simply continuing to patch the old software. The old software might be incredibly painful to look at from the inside, but *it works*, and it probably works a lot better than its newly designed replacement.
(By the way, I *hate* this principle, since as a software designer, I always like the idea of doing things "right". However, it is true, incremental improvement with local optimization does tend to be the way things actually get done in real life.)
Now, this isn't to say that the PPC is buggy. Hardware isn't software. However, the important thing to remember is it doesn't matter what the design looks like, only the results. It's aesthetically painful to see that a collection of nailed together boards, duct tape, and glue can moved like a rocket past our elegantly designed PPC, but it happens (at least for now :-))
[quote]Another glaring point he CONVENIENTLY forgot to mention was ... the PPC 970 is NOT the G4 simply fattened up to 64-bits<hr></blockquote>
Umm, any of the gains that you mention have *nothing* to do with the 64-bitness of the processor. Read the original quote: Moving from 32 to 64 bit, *not* moving to a different processor. The same 970 will not be a significantly better processor in 64-bit mode than in 32-bit mode. The same cannot be said for the Opteron.
Read Hannibal's tech papers like you would read a published paper. (Although they aren't peer reviewed :-)) They mean *exactly* what they say, no more and no less.
"Surprisingly we had to realise that the presentation of the PPC970, as announced before, unfortunately was cancelled. Upon request, we found out that the expected processor (2.3 to 2.5 GHz) is just a preview on the next but one processor generation and that one should not expect it anytime soon. Currently, PPC970 CPUs (still manufactured in 0.13 µm, like the upcoming 2.3/2.5 GHz CPU) are available in speeds up to 1.8 GHz that - according to IBM - might be used by Apple in the future. Exact information on this is still not available."
Surprisingly, we found out that the PPC presentation announced in advance [of CeBIT] was cancelled. When we asked, we learned that the expected processor (2.3 to 2.5 Ghz) was actually the generation after the next [i.e. post initial 970], and that we should thus not expect an introduction [of the 2+ Ghz iron] anytime soon. According to IBM, "current" PPC 970 processors (still produced with a .13 micronc process, which applies to the upcoming 2.3/2.5 Ghz processor as well), clocking up to 1.8 Ghz, could be used by Apple in the future. As always, however, it has been impossible to optain more detailed information.
1.8Ghz initially is pretty darn good if you ask me. It would be expected that before the end of the year IBM will have enough 2.0-2.5GHz 970s for the second revision before moving the processor to 90nm.
Up to 1.8 Is plenty compared to the the G4. SP 970 @ 1.3-1.5-1.8 would be both a clock speed improvement and a substantial improvement over the current line up as well as supposedly bringing the top of the line up on par with the 3 GHz P4.
I would hope that Apple grasp the opportunity to get ahead for a while by having a SP, DP, DP lineup even if that would mean to skip the 1.8 for a Sp1.3-DP1.3-DP 1.5 if the yields are low and IBM want to keep the 1.8 for their servers.
Really to bad about the non event of the demo as the having IBM blade servers running linux side by side using either dual Xeon or dual 970 would tell how well the 970 stacks up agianst the Intel CPUs.
...you do know we have an electoral college, right. and this isn't the first time its happened, right?
Its not like its black art. Its law.
And, BTW, before you draw any conclusions, I greatly dislike Bush. But the Constitution is the Constitution. Pretty straight forward.
So lets sweep that one under the rug, shall we?
Apple does seem to be tightening the noose a little in the ADC. Maybe they have some news coming out at WWDC they don't spread around. Uhm, maybe like 64bit which pretty much lets the cat out of the bag.
I mean, like developers of all kinds gotta know at some point before Apple releases the hardware.
*ahem* He would've won the electoral vote if the votes were counted ah, correctly
I'm not surprised about thw 1.4ghz-1.8ghz Apple has to continue their current speed trend. 1.4ghz as the low end (current top) and 1.8ghz as the new top. That way the 970 can completely replace the line and the G4 won't be missed because the numbers (1.4ghz +) are there. The 2.0ghz+ was wishful thinking for the intro, so this will just put those fires out.
So people putting off upgrading till next year will be happy.
Amorph- I saw that vein in your forehead throbbing, so I didn't want to cause you any further headaches
<strong>Looks good. 2.3-2.5 on 130nm bodes well for the 970 at 90nm.
Now the question is how is the yield rate for the 1.6 and 1.8Ghz parts? I'd love to know that.</strong><hr></blockquote>
Despite the "retraction" of the product "announcement" I suspect that IBM does have 970s on 0.13 running @ 2.5 GHz, but due to the power consumption, heat and yields it won't ever see the light of day in a shipping product. Having some chips clocking really well, however, is very promising for the yields of 1.8 GHz parts.
Comments
<strong>thats a Moto upgrade schedule. I think you're going to see .09u sooner than 2005. IBM will move aggressively to .09u manufacturing.</strong><hr></blockquote>
Agreed, East FishKill is set up for .09u. I'm not sure but I think that IBM indicatied that they will move to .09u withing 6 months of the initial release of the 970, but I would imagine no later than 12 months. IBM wants the 970 to be successfull, and they will push it as hard as possible. Weather Apple can get ahold of the fastest chips right away is left to be seen, but I think things look pretty good for 970 PM's shipping within 6 months.
<strong>
Agreed, East FishKill is set up for .09u. I'm not sure but I think that IBM indicatied that they will move to .09u withing 6 months of the initial release of the 970, but I would imagine no later than 12 months. IBM wants the 970 to be successfull, and they will push it as hard as possible. Weather Apple can get ahold of the fastest chips right away is left to be seen, but I think things look pretty good for 970 PM's shipping within 6 months.</strong><hr></blockquote>
IBM said that they would take the 970 to a .09um in early 04 starting at 2ghz ant that was before the fubar that the German site let slip.
<strong>
Yea, confused about how Gore received the majority of ballots yet Bush got the victory
I can't refute the rest as Amorph will have a hissy.</strong><hr></blockquote>
Hissy? <img src="graemlins/lol.gif" border="0" alt="[Laughing]" />
Go and have at it in AppleOutsider, for all I care. Or in email. Or in a Motel 6.
<a href="http://arstechnica.com/cpu/03q1/x86-64/x86-64-5.html"; target="_blank">X86-64</a>
All hypothetical if you ask me....
Anyone else get the impression that Hannibal is just another Wintel shill? I'm not sure I trust those guys over at Ars.. That's why I just sit and lurk... I mean just look at that statement. It's TOO one-sided and someone should call him on it. 64/32-bit was built right into the PPC ISA from the get-go. X86 will just be ANOTHER HACK. Period. Not to mention that BOTH Intel and Micor$oft see no chance of Windows/Intel moving to 64-bit desktops until at least the end of the decade. I've attempted to explain the debacle <a href="http://forums.appleinsider.com/cgi-bin/ultimatebb.cgi?ubb=get_topic&f=1&t=003098"; target="_blank">here.</a>
Another glaring point he CONVENIENTLY forgot to mention was that if all things being *equal* he statement about what Mac users could expect *might* be true, however, as we know, things are far from equal and the PPC 970 is NOT the G4 simply fattened up to 64-bits. It's a completely different design with added functionality and enhancements. I can't imagine what to expect from AltiVec and the additional bandwidth alone.. Or how about that extra FPU ... and so on. Imagine that in an SMP environment etc.... Yeah, he's missed a few points. Anyone care to add anything?
--
Ed M.
Should Apple move from 32-bit PPC to 64-bit PPC, Mac users should not expect the same kinds of performance gains that x86 software sees from the jump to x86-64. 64-bit PPC gives you larger integers and more memory, and that's about it. There are no extra registers, no cleaned up addressing scheme, etc.<hr></blockquote>
[quote] Anyone else get the impression that Hannibal is just another Wintel shill? <hr></blockquote>
Say what? Hannibal has produced perhaps the most insightful piece on the design of the 970 ever. While he's been stymied in getting more information on it from IBM (still waiting from part 2 of his article), calling him a Wintel shill is ridiculous.
His comment makes a huge amount of sense. The AMD effort will improve the x86 by fixing all sorts of shortcomings by tweaking the 64-bit instructions. The PPC instruction set, having been created by design rather than added to haphazardly, doesn't need and won't get the overhaul.
[quote]64/32-bit was built right into the PPC ISA from the get-go. X86 will just be ANOTHER HACK. Period.<hr></blockquote>
One thing that real life has taught be is that a working hack is how most of the world works. A nice clean design to start out with is a good thing (TM), but the most successful products (hardware, software, human institutions) are merely agglomerations of cruft, each piece of cruft optimized for the time it was released with little thought to the future. Suprisingly enough, the technique works.
As an example, take the fact that OS X was built on NeXT OS rather than BEOS. I know a number of people who were disappointed with the fact that Apple would be adding cruft to an old design rather than using a really fresh, aesthetically pleasing new design.
The reality is that, as many have found out to their horror, rewriting a big crufty software horror from scratch almost always results in a *vastly* more buggy piece of software than simply continuing to patch the old software. The old software might be incredibly painful to look at from the inside, but *it works*, and it probably works a lot better than its newly designed replacement.
(By the way, I *hate* this principle, since as a software designer, I always like the idea of doing things "right". However, it is true, incremental improvement with local optimization does tend to be the way things actually get done in real life.)
Now, this isn't to say that the PPC is buggy. Hardware isn't software. However, the important thing to remember is it doesn't matter what the design looks like, only the results. It's aesthetically painful to see that a collection of nailed together boards, duct tape, and glue can moved like a rocket past our elegantly designed PPC, but it happens (at least for now :-))
[quote]Another glaring point he CONVENIENTLY forgot to mention was ... the PPC 970 is NOT the G4 simply fattened up to 64-bits<hr></blockquote>
Umm, any of the gains that you mention have *nothing* to do with the 64-bitness of the processor. Read the original quote: Moving from 32 to 64 bit, *not* moving to a different processor. The same 970 will not be a significantly better processor in 64-bit mode than in 32-bit mode. The same cannot be said for the Opteron.
Read Hannibal's tech papers like you would read a published paper. (Although they aren't peer reviewed :-)) They mean *exactly* what they say, no more and no less.
[ 03-12-2003: Message edited by: Tom West ]</p>
Isn't IBM suppose to show off the 970 there?
<strong>Doesn't CeBIT start today?
Isn't IBM suppose to show off the 970 there?</strong><hr></blockquote>
Yes, rumors says so, but only to specific people who are under NDA.
We need a German translation please! Sherlock translation makes no sense...
<strong><a href="http://www.macguardians.de/fullstory.php?p=1830&c=1&bereich=2\" target="_blank">IBM at CeBit</a>
We need a German translation please! Sherlock translation makes no sense...
<a href="http://forums.macrumors.com/showthread.php?threadid=22050"; target="_blank">http://forums.macrumors.com/showthread.php?threadid=22050</a>;
"Surprisingly we had to realise that the presentation of the PPC970, as announced before, unfortunately was cancelled. Upon request, we found out that the expected processor (2.3 to 2.5 GHz) is just a preview on the next but one processor generation and that one should not expect it anytime soon. Currently, PPC970 CPUs (still manufactured in 0.13 µm, like the upcoming 2.3/2.5 GHz CPU) are available in speeds up to 1.8 GHz that - according to IBM - might be used by Apple in the future. Exact information on this is still not available."
[ 03-12-2003: Message edited by: joek ]</p>
I could live with a dual processor 1.8 GHz G5.
I just wish we knew the production schedule to see if MWNY is still a realistic release date.
[ 03-12-2003: Message edited by: Fran441 ]</p>
Now the question is how is the yield rate for the 1.6 and 1.8Ghz parts? I'd love to know that.
Surprisingly, we found out that the PPC presentation announced in advance [of CeBIT] was cancelled. When we asked, we learned that the expected processor (2.3 to 2.5 Ghz) was actually the generation after the next [i.e. post initial 970], and that we should thus not expect an introduction [of the 2+ Ghz iron] anytime soon. According to IBM, "current" PPC 970 processors (still produced with a .13 micronc process, which applies to the upcoming 2.3/2.5 Ghz processor as well), clocking up to 1.8 Ghz, could be used by Apple in the future. As always, however, it has been impossible to optain more detailed information.
1.8Ghz initially is pretty darn good if you ask me. It would be expected that before the end of the year IBM will have enough 2.0-2.5GHz 970s for the second revision before moving the processor to 90nm.
I would hope that Apple grasp the opportunity to get ahead for a while by having a SP, DP, DP lineup even if that would mean to skip the 1.8 for a Sp1.3-DP1.3-DP 1.5 if the yields are low and IBM want to keep the 1.8 for their servers.
Really to bad about the non event of the demo as the having IBM blade servers running linux side by side using either dual Xeon or dual 970 would tell how well the 970 stacks up agianst the Intel CPUs.
<strong>
...you do know we have an electoral college, right. and this isn't the first time its happened, right?
Its not like its black art. Its law.
And, BTW, before you draw any conclusions, I greatly dislike Bush. But the Constitution is the Constitution. Pretty straight forward.
So lets sweep that one under the rug, shall we?
Apple does seem to be tightening the noose a little in the ADC. Maybe they have some news coming out at WWDC they don't spread around. Uhm, maybe like 64bit which pretty much lets the cat out of the bag.
I mean, like developers of all kinds gotta know at some point before Apple releases the hardware.
[ 03-11-2003: Message edited by: mooseman ]</strong><hr></blockquote>
*ahem* He would've won the electoral vote if the votes were counted ah, correctly
I'm not surprised about thw 1.4ghz-1.8ghz Apple has to continue their current speed trend. 1.4ghz as the low end (current top) and 1.8ghz as the new top. That way the 970 can completely replace the line and the G4 won't be missed because the numbers (1.4ghz +) are there. The 2.0ghz+ was wishful thinking for the intro, so this will just put those fires out.
So people putting off upgrading till next year will be happy.
Amorph- I saw that vein in your forehead throbbing, so I didn't want to cause you any further headaches
<strong>Looks good. 2.3-2.5 on 130nm bodes well for the 970 at 90nm.
Now the question is how is the yield rate for the 1.6 and 1.8Ghz parts? I'd love to know that.</strong><hr></blockquote>
Despite the "retraction" of the product "announcement" I suspect that IBM does have 970s on 0.13 running @ 2.5 GHz, but due to the power consumption, heat and yields it won't ever see the light of day in a shipping product. Having some chips clocking really well, however, is very promising for the yields of 1.8 GHz parts.
<strong>
Yes, rumors says so, but only to specific people who are under NDA.</strong><hr></blockquote>
I'll be there tomorrow - maybe I'll learn sth. there.
<strong>Reiterating what Amorph has already said, keep the political discussions out of FH. This is the final warning.</strong><hr></blockquote>
Relax, my lengthy paragraph post was on topic except for 1 small sentence. No need to stress out, jesus.
<strong>
Relax, my lengthy paragraph post was on topic except for 1 small sentence. No need to stress out, jesus.</strong><hr></blockquote>
Hey you leave that religious talk in AppleOutsider, you hear?