Power5 based 975 90nm chips to ramp in May?
http://macosrumors.com/42004E2.php
""Taylor Barcroft writes: I am in the "G5 PowerBooks will be announced no sooner than at the January 2005 MacWorld Expo" school of thought.
It doesn't stand to reason that Apple will have a sufficient quantity of cool and cooled G5 processors for a completely new PowerBook cooling system before next year.
From what you wrote, I may even be overly optimistic.
I don't understand anyone who thought they were going to appear this year. That was always pure fantasy to my point of view.
Just the fact that the next G5 lineup is going to be so late in arriving - until Summer '04 - tells me that the next gen 90 nm G5 processors are not as much of an early 2004 mass production "slam dunk" as all of us expected - including IBM and Apple.
And what is the state of the 90 nm Apple G5 CONTROLLER production?? Any word on that front?
From what reliable IBM sources have been telling us, 90-nanometer system controller production has been quite solid. Not quite up to par with optimistic early estimates, but well within the range that wouldn't cause any delays on its own.
The problem is less with IBM's 90nm fabrication process, and more with some specific packaging design on the PPC 970FX.
In moving its new 300mm silicon wafer facilities to 90 nanometer production, IBM also introduced new techniques for CPU packaging and in the case of the 970FX, this backfired badly.
According to sources at IBM's Fishkill, New York facilities many 970FX's from the initial manufacturing run experienced "delamination" at the temperatures experienced during normal operation of chips clocked higher than 2GHz -- thus the 2GHz 970FX-powered Xserve G5.
By "delamination," they mean that the outer shell of the CPU cracked and separated, causing severe damage to the silicon structure inside and rendering most chips useless within only a few hours of testing. As a result, major changes had to be made, delaying the 970FX by several months and greatly decreasing viable yields of higher-clocked chips.
Interestingly enough, these problems did not affect the POWER5-based PowerPC 975 (also a 90 nanometer chip) which is slated to go into full production next month. As a result, the 970FX has almost no head-start on the more powerful 975 and this only adds to Apple's troubles in developing new G5-based systems like the PBG5. ""
This is amazing information, but where is it coming from. I know it is MOSR, but this is very interesting. Does anyone here have a souce that can back any of this up.
I have been talking about the fact that Apple and IBM were developing two different chip lines (Power4 and Power5 based) but so little of Apple's participation has been made public in regards to a PPC based on Power5. It is know and published that IBM is pushing Power5 based systems that will start shipping in June, but the best kept secret might be the fact that the next big thing from Apple might be a Power5 based G5.
ETA the title should be 975, not 970.
[Fixed the title -Amorph]
""Taylor Barcroft writes: I am in the "G5 PowerBooks will be announced no sooner than at the January 2005 MacWorld Expo" school of thought.
It doesn't stand to reason that Apple will have a sufficient quantity of cool and cooled G5 processors for a completely new PowerBook cooling system before next year.
From what you wrote, I may even be overly optimistic.
I don't understand anyone who thought they were going to appear this year. That was always pure fantasy to my point of view.
Just the fact that the next G5 lineup is going to be so late in arriving - until Summer '04 - tells me that the next gen 90 nm G5 processors are not as much of an early 2004 mass production "slam dunk" as all of us expected - including IBM and Apple.
And what is the state of the 90 nm Apple G5 CONTROLLER production?? Any word on that front?
From what reliable IBM sources have been telling us, 90-nanometer system controller production has been quite solid. Not quite up to par with optimistic early estimates, but well within the range that wouldn't cause any delays on its own.
The problem is less with IBM's 90nm fabrication process, and more with some specific packaging design on the PPC 970FX.
In moving its new 300mm silicon wafer facilities to 90 nanometer production, IBM also introduced new techniques for CPU packaging and in the case of the 970FX, this backfired badly.
According to sources at IBM's Fishkill, New York facilities many 970FX's from the initial manufacturing run experienced "delamination" at the temperatures experienced during normal operation of chips clocked higher than 2GHz -- thus the 2GHz 970FX-powered Xserve G5.
By "delamination," they mean that the outer shell of the CPU cracked and separated, causing severe damage to the silicon structure inside and rendering most chips useless within only a few hours of testing. As a result, major changes had to be made, delaying the 970FX by several months and greatly decreasing viable yields of higher-clocked chips.
Interestingly enough, these problems did not affect the POWER5-based PowerPC 975 (also a 90 nanometer chip) which is slated to go into full production next month. As a result, the 970FX has almost no head-start on the more powerful 975 and this only adds to Apple's troubles in developing new G5-based systems like the PBG5. ""
This is amazing information, but where is it coming from. I know it is MOSR, but this is very interesting. Does anyone here have a souce that can back any of this up.
I have been talking about the fact that Apple and IBM were developing two different chip lines (Power4 and Power5 based) but so little of Apple's participation has been made public in regards to a PPC based on Power5. It is know and published that IBM is pushing Power5 based systems that will start shipping in June, but the best kept secret might be the fact that the next big thing from Apple might be a Power5 based G5.
ETA the title should be 975, not 970.
[Fixed the title -Amorph]
Comments
But seriously, what would Apple do if they had the 975 ramping well and ready for release by say September? Does the 970fx go out to pasture, or does it end up being the G4 of today?
(Google translation)
http://croquer.free.fr
Originally posted by Defiant
WTF pissed in your brain for quoting MOSR?
Thank you, when I find out, I will personally let you know. So now that you have flamed me, how about some speculation over my ???'s.
If IBM really invested a lot of time and effort to fix the PPC970fx problem, do you think they'd throw this out the window and start producing the 975 instead? Why fix it in the first place?
If Apple ordered 100.000s of these chips, could they change their order to 975s just like that?
I'm sure Apple intended to introduce new G5s around January-March (just to keep the G5 momentum going), and then perhaps new ones based on the 975 in September. But we all know, that didn't happen. What can they realistically do instead?
We know that the 970fx can be produced at 2GHz (see Xserves), but there's no point offering dual 2GHz G5s with a 970fx chip, they might as well keep selling the current 970 based ones.
They might introduce 2.2GHz G5s (if yields permit) but we don't know whether they have enough 2.4 or 2.6GHz chips as well for a whole new lineup.
It's likely that the new 970fx chip requires a redesign of the motherboard, and with it a redesign of the case (fixing a few 'issues' with the current case design as well). Would Apple bother with this just for a 10% speed increase (to 2.2GHz) without any other higher clock speeds? Unlikely.
So all they can do is wait for enough 2.4GHz chips to introduce at least a 2.0GHz (old design) and 2.2 and 2.4GHz (new design) G5 lineup.
If they can manage to do this by mid May, then not too much is lost and a 975 based G5 can be introduced without much problems 6 months later in November.
Re PowerBooks the question is, whether the PPC975 is suited for notebooks, or whether that would be a 975fx variant, which likely won't ship at the same time as the 975 (my guess being that when a new processor type is introduced a low power variant follows a bit later just to make sure there aren't any big quirks in the original design. Would make sense).
If true, then we really won't see a G5 PowerBook before January 2005, then based on the 975fx.
Now why wouldn't Apple introduce G5 PowerBooks based on the 970fx? I don't know. Beats me. G5 PBs wouldn't need 2GHz so all those poor yield chips of the 970fx at lower than 2GHz speeds would be perfect for PBs. Why not use them? Three reasons I can think of:
- these 970fx chips really aren't useful, their flaw is deeper than just not reaching top speeds.
- the issue is not with the 970fx but with the controller chip which reportedly is manufactured in the same process and plant as the 970fx - running into the same issues.
- they turned out to still be too hot for notebooks and Apple has to wait for yet another revision, the 975fx.
If the first two are true then we might see G5 PowerBooks sooner, with the latter it will be January 2005.
To throw a few other assorted thoughts into this thread:
I'm still not sure whether Nr9's post about two dual core PPC 440s with AltiVec on one chip at 700- 800MHz each couldn't be an alternative for a portable 'G5 Mobile' chip. Especially if such a quad-core chip wouldn't consume more than the promised 7-8W max. In that case we might indeed see G5 PowerBooks at WWDC. Why WWDC? A quad core design would undoubtedly pose challenges for developers so a WWDC intro makes sense. And who knows, maybe it'll be like with the G5s, intro at WWDC, first delivery September...
Another possible hint at PowerBooks at WWDC are the lackluster specs in the 'new' PowerBooks. It's a nice incremental bump, but hardly anything special, particularly for the high-end 17". Yes, nice new ATI 9700 with 128MB VRAM, but still same slow HD, same slow bus, same res screens, about same CPU speed too.
In any case something must have gone wrong somewhere along Apple's schedules, I'm pretty sure that's not what they had intended.
About a quarter ago there was a news blurb about how many G5 chips Apple ordered from IBM and the figure for Q1 2004 was 500.000. Since Apple always targeted the PowerMacs at around 200.000 units per quarter, what did they intend the 300.000 for? I don't think for Xserves. iMac G5 or PowerBook G5? Or a smaller tower perhaps (Cube II anyone)? Personally I think iMacs. But Apple definitely intended to introduce at least one other G5 based Mac in Q1 2004. They haven't. So something must have gone wrong.
If IBM really invested a lot of time and effort to fix the PPC970fx problem, do you think they'd throw this out the window and start producing the 975 instead? Why fix it in the first place?
If Apple ordered 100.000s of these chips, could they change their order to 975s just like that?
I am not suggesting that IBM would through anything out the window, but like the G4, If a PPC based on Power5 comes out then the PPC's based on Power4 would take second fiddle.
Re PowerBooks the question is, whether the PPC975 is suited for notebooks, or whether that would be a 975fx variant, which likely won't ship at the same time as the 975 (my guess being that when a new processor type is introduced a low power variant follows a bit later just to make sure there aren't any big quirks in the original design. Would make sense).
Well, that is just it. We no next to nothing of the Power5 based PPC. For all we know, it could have huge thermal benefits. IBM is going to use the Power5 in blades and other servers starting in June.
I'm still not sure whether Nr9's post about two dual core PPC 440s with AltiVec on one chip at 700- 800MHz each couldn't be an alternative for a portable 'G5 Mobile' chip. Especially if such a quad-core chip wouldn't consume more than the promised 7-8W max. In that case we might indeed see G5 PowerBooks at WWDC. Why WWDC? A quad core design would undoubtedly pose challenges for developers so a WWDC intro makes sense. And who knows, maybe it'll be like with the G5s, intro at WWDC, first delivery September...
Can I have a link to Nr9's post.
Ok, my question is how could they call it a G5? Maybe they could call it a G4, but with out the ability to do anything 64bitish, not that there is a need for a 64 bit laptop, It would be a farce.
Maybe Apple is working on a new mobile chip, maybe instead of PPC, we could have MMC ( Mac Mobile chip) or MPPC (Mobile Power PC).
in any case something must have gone wrong somewhere along Apple's schedules, I'm pretty sure that's not what they had intended. About a quarter ago there was a news blurb about how many G5 chips Apple ordered from IBM and the figure for Q1 2004 was 500.000. Since Apple always targeted the PowerMacs at around 200.000 units per quarter, what did they intend the 300.000 for? I don't think for Xserves. iMac G5 or PowerBook G5? Or a smaller tower perhaps (Cube II anyone)? Personally I think iMacs. But Apple definitely intended to introduce at least one other G5 based Mac in Q1 2004. They haven't. So something must have gone wrong.
And two eggs in the basket are worth one in the hand. Now is this order broken down by part, If some are PPC 970's and some are PPC 970fx's , this number doesn't seem that out of line. Does anyone know how many XServe's were ordered since it's introduction?
Remember that just because Apple orderd that many, does not mean IBM shipped that many. Do we have any knowledge of what IBM actually shipped to Apple. Remember, most of the G5 units Apple has sold are of the dual processor variety. This also reduces the number of units that can be produced for that 500,000 cpu's.
Originally posted by oldmacfan
I am not suggesting that IBM would through anything out the window, but like the G4, If a PPC based on Power5 comes out then the PPC's based on Power4 would take second fiddle.
The point was that the 970 was always a transitional chip. It was never intended for a long production run, even as second fiddle. The real goal was always a chip based on the Power5 design, which was also intended to be available in a well-designed low power version for notebooks.
Well, that is just it. We no next to nothing of the Power5 based PPC. For all we know, it could have huge thermal benefits. IBM is going to use the Power5 in blades and other servers starting in June.
Yes, and which is why I think that the PowerBooks will have to wait for the 975fx, as this was the chip intended for them all along.
Can I have a link to Nr9's post.
Here you go. Very long thread, but makes some good reading. I truly hope that Apple is giving this design concept some trials.
Ok, my question is how could they call it a G5?
No they wouldn't. Nr9 speculated on 'Portable G5' or it could be 'G5 mobile' to differentiate that it is not in fact a true 64bit chip.
But the beauty of this design would be that a quad-core chip would make the PowerBook a lot more responsive in feel - if applications make good use of effectively 4 CPUs.
A design which involves an OS that works well multi-threaded, a new hardware platform and applications which demo this power - all 3 Apple can provide, as almost the only mainstream computer manufacturer in the world. A unique position, something Apple could (and should) make best use of.
Does anyone know how many XServe's were ordered since it's introduction?
From the conference call last quarter, if I remember correctly, it was somewhere between 5,000 and 10,000 shipped during that quarter. Hardly worth mentioning. No way near 100,000. Not even close. Let alone 300,000.
Remember that just because Apple orderd that many, does not mean IBM shipped that many. Do we have any knowledge of what IBM actually shipped to Apple. Remember, most of the G5 units Apple has sold are of the dual processor variety. This also reduces the number of units that can be produced for that 500,000 cpu's.
We don't know how many of the ca. 190.000 PowerMacs Apple shipped last quarter are duals. Maybe half? That'll make ca. 285,000 G5 chips. Plus ca. 10,000 xServes makes for ca. 300,000 G5 CPUs.
And that's exactly my argument. Apple had ordered 500,000 for shipment in that quarter. I doubt that Apple intended to hord 200,000 chips for a few months. Better to ship them ASAP in Macs. Since we know that Apple never forcast to sell more than 200,000 PowerMacs a quarter, my guess was that they intended to sell 200,000 G5 iMacs on top (single G5 CPUs) - which they couldn't since IBM didn't deliver. Which is why the iMac G5 was not introduced during the last quarter.
There really aren't many Xserves sold. Probably not even 10,000 in half a year!
Let's do a thinking experiment:
Even if all of their 190,000 PowerMac G5s sold in Q2 were duals (and they weren't), plus 10,000 Xserves (which Apple didn't sell and which probably are already counted as part of those 190,000) we would end up with 200,000 'PowerMacs' sold totalling 400,000 G5 processors. If so, what are the remaining 100,000 Apple ordered for?
Would Apple really put that many G5 processors in storage "just in case"?
Is Apple's forecast that poor that they have to buy enough chips just in case their forecast is 25% off (i.e. they sell 250,000 duals instead of 200,000)?
Or did they perhaps intend to sell some other G5 product too, which didn't happen?
Just a speculative question.
Originally posted by aHobbit
According to this news report Apple shipped 6,000 Xserves in Q1 2004 and "very, very few" in Q2 2004.
There really aren't many Xserves sold. Probably not even 10,000 in half a year!
Let's do a thinking experiment:
Even if all of their 190,000 PowerMac G5s sold in Q2 were duals (and they weren't), plus 10,000 Xserves (which Apple didn't sell and which probably are already counted as part of those 190,000) we would end up with 200,000 'PowerMacs' sold totalling 400,000 G5 processors. If so, what are the remaining 100,000 Apple ordered for?
Would Apple really put that many G5 processors in storage "just in case"?
Is Apple's forecast that poor that they have to buy enough chips just in case their forecast is 25% off (i.e. they sell 250,000 duals instead of 200,000)?
Or did they perhaps intend to sell some other G5 product too, which didn't happen?
Just a speculative question.
The #'s for XServe you are quoting are for the original Xserve, not the G5 Xserve. Like I said, how many were ordered? Also, what is the date of the order? Maybe Apple thought it would need that many for the Xserve.
1) Unlike many others, I don't think a "975" reaching production Real Soon Now is unreasonable. It may not be "POWER5-based", but it could share certain features with the POWER5. If it does exist, it is certainly higher power, higher performance than the 970FX.
2) The 970FX's problems affected all speeds and resulted in catastrophically bad yields which prevented even the very low sales numbers for the Xserve to be fullfilled. This has been corrected, and Apple is probably rapidly accumulating stock at this point.
3) Apple was almost certainly planning to introduce a PowerMac using the 970FX at speeds around 2.5 GHz a couple of months ago. This went out the window when the 970FX problems turned up early this year or late last year. Perhaps they figured that IBM would correct the problem quicker ... it is quite common to fall victim to the "its just about fixed, any day now" syndrome which lasts for months.
4) It is debateable whether the 970FX is going to reach 3 GHz, and it is possible that Steve's promise of 3 GHz was based on the 975 from the beginning. Since the 975 is only hitting production next month (supposedly) it will never suffer from the initial troubles the 970FX had. The 90nm problems basically "soaked up" the 970FX's lead time.
5) The 970FX will eventually be Apple's consumer, and possibly portable. I don't see the portables getting it for at least a year since even at 1.8-2.0 GHz the new G4 is argueably more appropriate. And if Motorola (Freescale now) is actually getting their act together a 90nm G4 w/ SoC capabilities may be a better portable option entirely. See Intel's chip lineup for details. Nonetheless, the 970FX's Apple has ordered will have plenty of uses for Xserves and non-highend desktops.
But unfortunately Apple didn't give any Xserve figures in their Q1 2004 figures.
They only gave totals, which was 829,000 Macs sold. And I don't even know whether the Xserve would be counted as a 'Mac', but I would assume so. In which case the Xserves would have to be counted as part of the 206,000 PowerMacs sold (since iMacs, iBooks or PowerBooks would seem even less appropriate). Which means that the 190,000 'PowerMacs' sold in this quarter also already includes all G5 Xserves sold.
And which also means that Apple's goal of selling of 200,000 PowerMacs each quarter would also include the Xserves too. So no, they didn't think of selling some extra 50,000 Xserves last quarter.
Well, perhaps we will never know what really happened, but I think Apple did forecast to sell at least a high-end iMac G5 model - which didn't come to pass due to a lack of 970fx CPUs.
If I remember the original roadmap for the G5 correctly then a Power5 derived desktop chip was always slated for around September 2004, which could very well be the 975 chip. And there were always rumors from the beginning, that Apple might in fact call it 'G6' since the Power4 derived chip was called 'G5'.
So personally I would not even be surprised to see a G6 in September.
The 100mio$ question is just whether Apple will release another G5 until then. If they wait until May/June (let alone WWDC end of June, actually July) to introduce a PowerMac speed bump, it becomes very unlikely they introduce the next bump to the 975 chip two months later. Would Apple then delay their introduction until next year? Assuming that this chip would be the one fulfilling the 3GHz promise, I don't think so.
Therefore I think it would be much more likely if Apple cannot introduce a 2.5GHz speed bump as planned, that they go for a 2.3 or 2.4GHz speed bump sooner (e.g. beginning of May, hoping that 2.3GHz yields will catch up faster than 2.5GHz yields now that IBM fixed the 970fx problems) and delaying the 975 based next version a month or two until October/November.
I can literally see Steve at the press event mid October apologizing for not making the promised 3GHz 'by the end of summer' while adding in the same sentence 'and in order to make up for it I would like to introduce our brand new 3.1GHz PowerMac G6'.
Originally posted by Programmer
There seems to be a lot of odd reasoning in this thread...
That's the understatement of the century. This whole thread is really baseless and just doesn't offer any kind of realistic speculation.
Even worse, it references Nr9's blab about dual-core 440's going into PowerBooks. When Apple releases a router called "PowerBook", I'll buy it...until then this is all just a bunch of crap!
Power5's have been shipping for a few months, but how long were the Power4's in products before a 970 was released? Couple of years? No, IBM isn't that good/ignorant.
Originally posted by aHobbit
If I remember the original roadmap for the G5 correctly then a Power5 derived desktop chip was always slated for around September 2004, which could very well be the 975 chip. And there were always rumors from the beginning, that Apple might in fact call it 'G6' since the Power4 derived chip was called 'G5'.
What roadmap? IBM and Apple have not issued a processor roadmap, all you are remembering is the speculation generated by MOSR and other rumour sites. As for what Apple will call the G6, that's anybody's guess. I doubt they'll switch from G5 to G6 after 1 hardware revision though, even if they have a new chip (975, or whatever they call it).
As for what speeds they can achieve, don't fall into the trap of thinking that since they've delivered no 970FXes that they will have to ramp through the slow speeds first. Imagine this is the Ferrari manufacturing facility but somebody lost the key to the front gate... when they finally get the gate open the first Ferraris aren't limited to 50 mph.
Originally posted by Programmer
2) The 970FX's problems affected all speeds and resulted in catastrophically bad yields which prevented even the very low sales numbers for the Xserve to be fullfilled. This has been corrected, and Apple is probably rapidly accumulating stock at this point.
How do we know that this has been corrected? There has been some announcement or do you know something not publicly available?
Originally posted by Rhumgod
Power5's have been shipping for a few months, but how long were the Power4's in products before a 970 was released? Couple of years? No, IBM isn't that good/ignorant.
I don't understand well what you are trying to say here. Maybe that a POWER5 derivative (PPC975 or whatever) would take a couple of years to appear? If so, I don't think that the initial POWER4 -> PPC970 transition time frame is a good indicator.
Originally posted by aHobbit
According to this news report Apple shipped 6,000 Xserves in Q1 2004 and "very, very few" in Q2 2004.
There really aren't many Xserves sold. Probably not even 10,000 in half a year!
Apple haven't said how many Xserves they managed to ship last quarter but they did say that there was around 16,000 on back order.
Originally posted by Programmer
What roadmap? IBM and Apple have not issued a processor roadmap, all you are remembering is the speculation generated by MOSR and other rumour sites.
Yes, of course it's all rumors since neither Apple nor IBM would comment on future products.
The article I was thinking of was from MacRumors. Maybe not trustworthy.
They quoted a "PowerPC 980" name and a release date of Q3 2004.
And thanks for the Ferrari comparison.
According to IBM's own Roadmap they classify the 970FX as "up to 2.0+ GHz", which I'm not quite sure what it means.
"up to 2GHz and more" doesn't really sound like a lot more than 2.0GHz to me. Sounds more like "let's try it and see what we can get". And perhaps they found that they can't get much more than 2GHz...