What I understood of the conference call it was 90nm 1.8-2.5 that were in tight supply.
My understanding is that the 1.8 and 2.0 chips are .13 and not .90. And yet in the conference call (starting around 3:30) they said this:
Quote:
Unfortunately, IBM's manufacturing problems will impact our September quarter as well. We expect to have shortages of our dual 1.8 and dual 2.0 Ghz models in July, though supply of these models should catch up in August, and we expect to have shortages of our high end dual 2.5 Ghz model throughout the quarter.
I wasn't aware of any shortages of the 1.8 or 2.0 Ghz machines before they were re-packaged at the low and middle. So why would they start having problems now? Maybe they really are .90. But why would they use those new, delayed chips, when we know the old 1.8 and 2.0 chips work just fine?
They are probably using the 130nm in the G5 Towers wince they have room for cooling and they are going to use the 90nm ones (in tight supply) in the new iMac G5. Hence the reason for the delay.
My understanding is that the 1.8 and 2.0 chips are .13 and not .90. And yet in the conference call (starting around 3:30) they said this: I wasn't aware of any shortages of the 1.8 or 2.0 Ghz machines before they were re-packaged at the low and middle. So why would they start having problems now? Maybe they really are .90. But why would they use those new, delayed chips, when we know the old 1.8 and 2.0 chips work just fine?
perhaps they both are in the same facility and in order to manufacture 90 nm they had to reduce capacity of 130nm.
who knows, but the dual 1.8 and dual 2 currently use 130nm processors
I'll add yet another interpretation of the 970fx situation. It is evidently being offered at three clock rates, 1.8, 2.0 and 2.5 GHz. The 2.0 GHz 970fx is going into the Xserve, the 1.8 GHz 970fx is being reserved for the new iMac, and the 2.5 GHz for the top Power Mac. The supply of these processors is constrained, but the supply of 1.8 and 2.0 GHz 970 chips is plentiful. These 130 nm 970 chips are used in the two lower speed Power Macs, at least for now.
The G5 processor, both 970 and 970fx, can operate at more than one supply voltage. The 90 nm 970fx no doubt can operate at 1.8 and 2.0 GHz with lower supply voltage than the 970. Lower supply voltage plus lower capacitance of the smaller chip both contribute to lower power for the 970fx, over a 130 nm 970 operating at the same clock rate. The 970fx may not have achieved all the power reduction that IBM and Apple hoped for, but it should still be lower than an equivalent 970 chip, especially at a lower supply voltages.
The yield problem for the 970fx is due to higher than anticipated leakage current, if I am remembering what I read correctly. Leakage current contributes to a chip's power dissipation, and increases dramatically with increased supply voltage. Since higher supply voltage is needed to achieve the highest clock rate, it is most serious for the 2.5 GHz 970fx. (Power dissipation due to switching increases with clock rate too, but only linearly.) As a point of interest, it is conceivable that the same 970fx chip could be used at 2.0 GHz in an Xserve at one supply voltage, or a 2.5 GHz Power Mac at a higher voltage. It would be interesting to get the actual CPU supply voltages for various G5 Macs.
If you are talking about those coloured cube shaped machines, they were a mock-up. If you are not then which pic's are you talking about? Link please.
I wish I could link to the picture, but they've been taken down due to a c&d by Apple. I'm not talking about mockups, I'm talking about the pics that Apple posted in a tech manual showing a smaller mobo in the current PMac case. Everyone was like, "What will go in that big empty spot?" and "Why is there only one heatsink?".
I'm thinking they were testing the new mobo. If that's true, we may finally get an upgradeable iMac.
The most likely reson for Apple ot move to a complete line up of PPC 970FX chips is to avoid any errata in the old chips. There MAY and I stress the word may be fixes in the 970FX that are significnat to Apple.
The other reason would revolve around economics, the 90nm device does use less power. It is also possible that the chip will be significantly cheaper due to the smaller die size. This smaller cooler chip could result in savings else where.
What I see happening with the 970 and 970FX, is that it is exposing a very significant weakness in IBM's ability to deliver low power devices. It is one thing for the orignal 970 to be a bit hot as it was rushed to market. The 970FX is another matter and I would say a technical failing on the part of IBM.
Dave
Quote:
Originally posted by BRussell
My understanding is that the 1.8 and 2.0 chips are .13 and not .90. And yet in the conference call (starting around 3:30) they said this: I wasn't aware of any shortages of the 1.8 or 2.0 Ghz machines before they were re-packaged at the low and middle. So why would they start having problems now? Maybe they really are .90. But why would they use those new, delayed chips, when we know the old 1.8 and 2.0 chips work just fine?
Another thing to remember is that G5 chips are not just for Macs. IBM puts 970 chips in servers too, so they are doing a little juggling act between their needs and those of Apple.
Could it be that the 130nm 970 was installed only in some early models of the new 1.8 and 2.0 GHz Power Macs? I sort of alluded to that possibility, and I agree with the comment about the 970fx being much cheaper. It seems unlikely that IBM and Apple would continue shipping 130nm 970 chips once the 970fx process is running smoothly. The 970fx will be the standard low-end G5 for the next year. Someone is even building an ATX motherboard for dual 1.4 GHz 970fx chips.
What I see happening with the 970 and 970FX, is that it is exposing a very significant weakness in IBM's ability to deliver low power devices. It is one thing for the orignal 970 to be a bit hot as it was rushed to market. The 970FX is another matter and I would say a technical failing on the part of IBM.
Dave
BINGO! That's my interpretation. A great design would've have given a 970FX 2.5GHz chip producing about as much heat as the 130 nm 2.0GHz 970 chip. Alas, this isn't the case which explains the incorporation of liquid cooling (at higher cost for Apple) on the dual 2.5's.
Now, what has Motorola been up to in the background?
Liquid cooling might not be because of higher heat, just higher heat *density*. 4 cm^2 dissipating 40W puts off more heat than 1cm^2 dissipating 20W, but that 20W has to be moved off the chip faster to avoid damage.
Well I haven't reviewed solid data on the 90nm chips power output at 2.5 GHz. At 2 GHz the unit did have a significant lowering of power disapated over the 130nm device. Which is nice until you realize that the 130nm device is EXTREMELY hot for what it does and its transistor count. In the end IBM's 90nm 970FX can't compete with other processors running on old 130nm processes.
That is in a nut shell is the problem with IBM's processes, they produce chips that are to hot for the performance they deliver. Now IBM's manufacturing issues are another thing entirely as they affect Apples capability to deliver any hardware. Hot chips only effect the desgn of the hardware.
There is a great deal of potential at 90nm but it does appear that other companies will be the ones taking advantage of that potential. There is one possibility though that I've yet to here alot on, and that is the possiblity that it isn't the 970FX that Apple is waiting on but a lower power variant. That would possibly explain some of the comment seen lately. As to Motorola maybe Apples problems with IBM would tend to soften the apparrent seperation of Apple and Motorola as partners. I believe that Motorola has a far greater chance of producing real low power processors once its produciton line comes on board. Part of Freescales advantage is that they will start out with a low power design and put in on a lower power process - that could be huge for the portables. In the end though Freescale will also need a 64 bit solution as the 32 bit world will quickly die off.
Thanks
Dave
Quote:
Originally posted by DVD_Junkie
BINGO! That's my interpretation. A great design would've have given a 970FX 2.5GHz chip producing about as much heat as the 130 nm 2.0GHz 970 chip. Alas, this isn't the case which explains the incorporation of liquid cooling (at higher cost for Apple) on the dual 2.5's.
Now, what has Motorola been up to in the background?
Well I haven't reviewed solid data on the 90nm chips power output at 2.5 GHz. At 2 GHz the unit did have a significant lowering of power disapated over the 130nm device. Which is nice until you realize that the 130nm device is EXTREMELY hot for what it does and its transistor count. In the end IBM's 90nm 970FX can't compete with other processors running on old 130nm processes.
That is in a nut shell is the problem with IBM's processes, they produce chips that are to hot for the performance they deliver. . . .
It would be interesting to know how you were able to check the power data. Did you actually measure input voltage and current to the CPU? If you did, does the 2.0 GHz 970fx run at lower voltage, like 1.1 Volts rather than 1.3 Volts for the 970?
Regarding the 90nm process, I only know what I read, which is that Intel is in the same boat. The 90nm Intel chips are showing very unimpressive gains over the 130nm chips. Now, if we compare the Intel chips with the IBM 970fx, I have trouble believing your evaluation of IBM. The 2.0 GHz dual Xserve was chosen for a cluster to do research for the Army and it will be the second most powerful computer in the world. Apple, with IBM's 970fx, beat out the competition on performance, price and ease of administration. Maybe you could elaborate. Also, I am pretty sure that Intel's fast processors run very hot too.
It would be interesting to know how you were able to check the power data. Did you actually measure input voltage and current to the CPU? If you did, does the 2.0 GHz 970fx run at lower voltage, like 1.1 Volts rather than 1.3 Volts for the 970?
All my information comes either from Apple IBM or the net. I do believe that it is accurate. Now if you make comaprisons with some of Intels product the argument may take a different slant but if you compare with AMD hardware I think everyone will see the light.
Quote:
Regarding the 90nm process, I only know what I read, which is that Intel is in the same boat. The 90nm Intel chips are showing very unimpressive gains over the 130nm chips. Now, if we compare the Intel chips with the IBM 970fx, I have trouble believing your evaluation of IBM. The 2.0 GHz dual Xserve was chosen for a cluster to do research for the Army and it will be the second most powerful computer in the world. Apple, with IBM's 970fx, beat out the competition on performance, price and ease of administration. Maybe you could elaborate. Also, I am pretty sure that Intel's fast processors run very hot too.
I'm not so sure Intel is in the same boat. They are doing very well with Dothan. Many of the problems with Prescott go beyond the 90nm issue.
Comments
Originally posted by oldmacfan
What I understood of the conference call it was 90nm 1.8-2.5 that were in tight supply.
My understanding is that the 1.8 and 2.0 chips are .13 and not .90. And yet in the conference call (starting around 3:30) they said this:
Unfortunately, IBM's manufacturing problems will impact our September quarter as well. We expect to have shortages of our dual 1.8 and dual 2.0 Ghz models in July, though supply of these models should catch up in August, and we expect to have shortages of our high end dual 2.5 Ghz model throughout the quarter.
I wasn't aware of any shortages of the 1.8 or 2.0 Ghz machines before they were re-packaged at the low and middle. So why would they start having problems now? Maybe they really are .90.
Macaddict16
Originally posted by BRussell
My understanding is that the 1.8 and 2.0 chips are .13 and not .90. And yet in the conference call (starting around 3:30) they said this: I wasn't aware of any shortages of the 1.8 or 2.0 Ghz machines before they were re-packaged at the low and middle. So why would they start having problems now? Maybe they really are .90.
perhaps they both are in the same facility and in order to manufacture 90 nm they had to reduce capacity of 130nm.
who knows, but the dual 1.8 and dual 2 currently use 130nm processors
The G5 processor, both 970 and 970fx, can operate at more than one supply voltage. The 90 nm 970fx no doubt can operate at 1.8 and 2.0 GHz with lower supply voltage than the 970. Lower supply voltage plus lower capacitance of the smaller chip both contribute to lower power for the 970fx, over a 130 nm 970 operating at the same clock rate. The 970fx may not have achieved all the power reduction that IBM and Apple hoped for, but it should still be lower than an equivalent 970 chip, especially at a lower supply voltages.
The yield problem for the 970fx is due to higher than anticipated leakage current, if I am remembering what I read correctly. Leakage current contributes to a chip's power dissipation, and increases dramatically with increased supply voltage. Since higher supply voltage is needed to achieve the highest clock rate, it is most serious for the 2.5 GHz 970fx. (Power dissipation due to switching increases with clock rate too, but only linearly.) As a point of interest, it is conceivable that the same 970fx chip could be used at 2.0 GHz in an Xserve at one supply voltage, or a 2.5 GHz Power Mac at a higher voltage. It would be interesting to get the actual CPU supply voltages for various G5 Macs.
I suppose that means that all G5's use the 90nm. That's taken directly out of the PowerMac Spec PDF. Page 9.
Originally posted by Myst
"The PowerPC G5 is fabricated in one of IBM?s world-class semiconductor manufacturing facilities. It uses 90-nanometer circuitry..."
I suppose that means that all G5's use the 90nm. That's taken directly out of the PowerMac Spec PDF. Page 9.
it doesn't
Originally posted by Addison
If you are talking about those coloured cube shaped machines, they were a mock-up. If you are not then which pic's are you talking about? Link please.
I wish I could link to the picture, but they've been taken down due to a c&d by Apple. I'm not talking about mockups, I'm talking about the pics that Apple posted in a tech manual showing a smaller mobo in the current PMac case. Everyone was like, "What will go in that big empty spot?" and "Why is there only one heatsink?".
I'm thinking they were testing the new mobo. If that's true, we may finally get an upgradeable iMac.
This is the thread the pics used to be on:
http://forums.appleinsider.com/showt...threadid=42679
Here is an image pscates put together after the original pics were removed:
Note the empty spot.
pscates mockup thread:
http://forums.appleinsider.com/showt...threadid=42736
The other reason would revolve around economics, the 90nm device does use less power. It is also possible that the chip will be significantly cheaper due to the smaller die size. This smaller cooler chip could result in savings else where.
What I see happening with the 970 and 970FX, is that it is exposing a very significant weakness in IBM's ability to deliver low power devices. It is one thing for the orignal 970 to be a bit hot as it was rushed to market. The 970FX is another matter and I would say a technical failing on the part of IBM.
Dave
Originally posted by BRussell
My understanding is that the 1.8 and 2.0 chips are .13 and not .90. And yet in the conference call (starting around 3:30) they said this: I wasn't aware of any shortages of the 1.8 or 2.0 Ghz machines before they were re-packaged at the low and middle. So why would they start having problems now? Maybe they really are .90.
http://www.970eval.com/
Originally posted by tak1108
Maybe the new montherboard is for the FX, and that's why no one has seen it yet in person. Because the 2.5 hasn't shipped yet?
More from the service manual showing the logic board of the liquid cooled PowerMacs here.
Looks like a long logic board to me, not the short one that AI posted a pic of and can be viewed in my above post.
Originally posted by wizard69
What I see happening with the 970 and 970FX, is that it is exposing a very significant weakness in IBM's ability to deliver low power devices. It is one thing for the orignal 970 to be a bit hot as it was rushed to market. The 970FX is another matter and I would say a technical failing on the part of IBM.
Dave
BINGO! That's my interpretation. A great design would've have given a 970FX 2.5GHz chip producing about as much heat as the 130 nm 2.0GHz 970 chip. Alas, this isn't the case which explains the incorporation of liquid cooling (at higher cost for Apple) on the dual 2.5's.
Now, what has Motorola been up to in the background?
That is in a nut shell is the problem with IBM's processes, they produce chips that are to hot for the performance they deliver. Now IBM's manufacturing issues are another thing entirely as they affect Apples capability to deliver any hardware. Hot chips only effect the desgn of the hardware.
There is a great deal of potential at 90nm but it does appear that other companies will be the ones taking advantage of that potential. There is one possibility though that I've yet to here alot on, and that is the possiblity that it isn't the 970FX that Apple is waiting on but a lower power variant. That would possibly explain some of the comment seen lately. As to Motorola maybe Apples problems with IBM would tend to soften the apparrent seperation of Apple and Motorola as partners. I believe that Motorola has a far greater chance of producing real low power processors once its produciton line comes on board. Part of Freescales advantage is that they will start out with a low power design and put in on a lower power process - that could be huge for the portables. In the end though Freescale will also need a 64 bit solution as the 32 bit world will quickly die off.
Thanks
Dave
Originally posted by DVD_Junkie
BINGO! That's my interpretation. A great design would've have given a 970FX 2.5GHz chip producing about as much heat as the 130 nm 2.0GHz 970 chip. Alas, this isn't the case which explains the incorporation of liquid cooling (at higher cost for Apple) on the dual 2.5's.
Now, what has Motorola been up to in the background?
Originally posted by wizard69
Well I haven't reviewed solid data on the 90nm chips power output at 2.5 GHz. At 2 GHz the unit did have a significant lowering of power disapated over the 130nm device. Which is nice until you realize that the 130nm device is EXTREMELY hot for what it does and its transistor count. In the end IBM's 90nm 970FX can't compete with other processors running on old 130nm processes.
That is in a nut shell is the problem with IBM's processes, they produce chips that are to hot for the performance they deliver. . . .
It would be interesting to know how you were able to check the power data. Did you actually measure input voltage and current to the CPU? If you did, does the 2.0 GHz 970fx run at lower voltage, like 1.1 Volts rather than 1.3 Volts for the 970?
Regarding the 90nm process, I only know what I read, which is that Intel is in the same boat. The 90nm Intel chips are showing very unimpressive gains over the 130nm chips. Now, if we compare the Intel chips with the IBM 970fx, I have trouble believing your evaluation of IBM. The 2.0 GHz dual Xserve was chosen for a cluster to do research for the Army and it will be the second most powerful computer in the world. Apple, with IBM's 970fx, beat out the competition on performance, price and ease of administration. Maybe you could elaborate. Also, I am pretty sure that Intel's fast processors run very hot too.
Originally posted by snoopy
It would be interesting to know how you were able to check the power data. Did you actually measure input voltage and current to the CPU? If you did, does the 2.0 GHz 970fx run at lower voltage, like 1.1 Volts rather than 1.3 Volts for the 970?
All my information comes either from Apple IBM or the net. I do believe that it is accurate. Now if you make comaprisons with some of Intels product the argument may take a different slant but if you compare with AMD hardware I think everyone will see the light.
Regarding the 90nm process, I only know what I read, which is that Intel is in the same boat. The 90nm Intel chips are showing very unimpressive gains over the 130nm chips. Now, if we compare the Intel chips with the IBM 970fx, I have trouble believing your evaluation of IBM. The 2.0 GHz dual Xserve was chosen for a cluster to do research for the Army and it will be the second most powerful computer in the world. Apple, with IBM's 970fx, beat out the competition on performance, price and ease of administration. Maybe you could elaborate. Also, I am pretty sure that Intel's fast processors run very hot too.
I'm not so sure Intel is in the same boat. They are doing very well with Dothan. Many of the problems with Prescott go beyond the 90nm issue.
Thanks
Dave