People are starting to blow thi sout of proportion. According to what we know, the yield isn't due to bad design, it was a fault in one of the materials used. Big deal, it is 'easily' fixed and we'll have our faster 0.09nm G5s, don't go around prophesying that Apple will fall due to another curse (i.e. like Apple did with motorola).
I am dissapointed, yes, but Apple isn't doomed at all. The G5 is moving to a new micron process for crying out loud, how long did it take with G4s? EXACTLY.
Is this not a dead topic with IBM saying it is their fault?
There are no know "problems with adapting 90nm G5's", except for the fact that IBM cannot produce them correctly.
Actually that is not correct -- they are currently claiming that they are producing them as they have already fixed the yield problem they were having. Now it is just a matter of delivering enough for Apple to launch their product(s) based on the chip (beyond just the Xserve G5).
It doesn't matter how old PPC970 is, the thing is the 90nm process itself is so immature that not only IBM, but Intel and AMD too, are having problems with it.
IBM need the 970 to run fast for their own Linux servers so they have a big interest to make the 970 run as fast as possible. I got the impression that Motorola lacked not only talent but also reasons to get over the 500 MHz barrier.
It is bad with this problem with the delayed speed bump of the towers as well as the delayed delivery of the G5 servers. It appears that the 3 GHz this summer will streach the definition of "summer" and "shipping".
It might be time to grumble but not yeat so scream blody murder
I'm still confused about SSDOI. I searched IBM's site and found only references to SSDOI and nothing concerning them using the Intel nitrate process. Please note here that my search skills ain't that proficient.
I only found 2 articles both written by Tony Krazit that mentioned Intel's process.
I'm still confused about SSDOI. I searched IBM's site and found only references to SSDOI and nothing concerning them using the Intel nitrate process. Please note here that my search skills ain't that proficient.
I only found 2 articles both written by Tony Krazit that mentioned Intel's process.
Here is April 20 news from Taiwan on IBM yield issues. My question is why does Apple stick with losing IBM?
http:/2004/04/20 Taiwan: TSMC (2330 TT, NT$59.5), UMC (2303 TT, NT$31.5): TSMC and UMC benefit from poor yield of IBM's 12" fab
IBM facing poor yield of 0.13um and 90nm
Main customers such as nVidia and Qualcomm are switching back to TSMC and UMC
Maintain BUY recommendations on TSMC and UMC
According to the "Economic Daily," nVidia, Qualcomm, Broadcom, Cisco and Xilinx are switching back from IBM foundry to TSMC and UMC due to the lower yield of IBM's 12" fab. IBM's lower yield of 0.13um and 90nm, and reducing IP revenue are causing IBM's 1Q losses in the semiconductor division.
Analyst comment:
1) We believe that IBM's lower 12" yield in 1Q04 will benefit TSMC and UMC going forward. The order rotation from IBM back to TSMC and UMC is expected to begin in 2Q04. We expect TSMC to show 8-10% and UMC 10% QoQ sales growth in 2Q04.
2) We expect UMC to benefit from capacity support from SiS and He-Jian beginning in 2Q04, which to show stronger revenue growth momentum than TSMC. However, we retain our BUY on both TSMC and UMC, due to their leading position in the technology universe, and capacity ramp-up, respectively.
Main customers such as nVidia and Qualcomm are switching back to TSMC and UMC
Maintain BUY recommendations on TSMC and UMC
.
.
.
Analyst comment:
1) We believe that IBM's lower 12" yield in 1Q04 will benefit TSMC and UMC going forward. The order rotation from IBM back to TSMC and UMC is expected to begin in 2Q04. We expect TSMC to show 8-10% and UMC 10% QoQ sales growth in 2Q04.
Ouch, bad news indeed. And there was some time ago Nr9 (I think) who talked about the superior (to IBM) fab capabilities of TSMC and UMC . I am curious to see how things will evolve for IBM and Apple this year.
Ouch, bad news indeed. And there was some time ago Nr9 (I think) who talked about the superior (to IBM) fab capabilities of TSMC and UMC . I am curious to see how things will evolve for IBM and Apple this year.
This is old bad news. IBM claims to have cracked their 90nm problems and their 970FX production is ramping. This will allow them to draw customers back from the competition. The thing that Nr9 never seemed to "get" was that running into a few problems is not a good reason to throw in the towel. There is value to making the investment and toughening it out to be able to compete in this market with relatively few players. At the very least its not a good idea to have the world's silicon chip production all centered on a contested small island off the coast of China, sitting right on the "Ring of Fire".
Here is April 20 news from Taiwan on IBM yield issues. My question is why does Apple stick with losing IBM?
If you're asking that question you don't understand the services IBM is providing to Apple. IBM isn't simply providing fab services to Apple, but also a complete PowerPC chip design(s) and future.
Your question should have been "Why isn't IBM outsourcing their fab process?". And I think someone already answered that question.
While i think apple will strive for 3GHz, it is absolutely necessary for them as a company to update by WWDC, no matter what the speeds are. Powermac sales have slipped every quarter, and extending their drought to longer than a year won't sit well with investors and power users. If there aren't 3GHz at WWDC, then I think we'll see them late October/early November.
You know what else is interesting though? Mac people are b**ching about processor speeds! Yes, as a power user myself I know that every hertz matters, but I bet a lot people griping about not having 3GHz are among the 90% of computer users that never make full use of the power of their system.
While i think apple will strive for 3GHz, it is absolutely necessary for them as a company to update by WWDC, no matter what the speeds are. Powermac sales have slipped every quarter, and extending their drought to longer than a year won't sit well with investors and power users. If there aren't 3GHz at WWDC, then I think we'll see them late October/early November.
I agree with you... if Apple was to release a PowerMac update with a new Superdrive, Dual 2.4Ghz G5 and ATI Radeon 9800 I'd be happy to buy it. I just don't want to purchase the current Dual 2.0Ghz which is over 7 months old now. I also don't really care about the promised 3Ghz... it would be nice, but it's not needed.
dual 3GHz G5's with on die memory controllers will destroy the competition until 2006. Here's to hoping it comes out before then.
I don't think you'll see the on-die memory controllers, but you may see improvements to things which mitigate the latency problems. Imagine dual 3+ GHz G5's that run two hardware threads each at full speed. That is largely equivalent to the quad-core machines people have been lusting after for a decade, only cheaper. This I think you might see by early 2005 (conservatively).
Comments
Originally posted by msantti
I think IBM should stop yaking about future chips and get a handle on what they are producing right now.
And they seem to be having a BIG problem with that.
Which is a big suprise... being it got them into HUGE trouble in the 80's.
Originally posted by mattyj
People are starting to blow thi sout of proportion. According to what we know, the yield isn't due to bad design, it was a fault in one of the materials used. Big deal, it is 'easily' fixed and we'll have our faster 0.09nm G5s, don't go around prophesying that Apple will fall due to another curse (i.e. like Apple did with motorola).
I am dissapointed, yes, but Apple isn't doomed at all. The G5 is moving to a new micron process for crying out loud, how long did it take with G4s? EXACTLY.
How long did it take?
There are no know "problems with adapting 90nm G5's", except for the fact that IBM cannot produce them correctly.
Originally posted by oldmacfan
Is this not a dead topic with IBM saying it is their fault?
There are no know "problems with adapting 90nm G5's", except for the fact that IBM cannot produce them correctly.
Actually that is not correct -- they are currently claiming that they are producing them as they have already fixed the yield problem they were having. Now it is just a matter of delivering enough for Apple to launch their product(s) based on the chip (beyond just the Xserve G5).
Originally posted by mattyj
Years for the G4. The G5 has't even been out for a year. A rev.b Powermac is getting a micron shrink, the g4 went through several revisions.
Actually IBM announced the 970 in January of '03. This makes it about 16 months old.
It is bad with this problem with the delayed speed bump of the towers as well as the delayed delivery of the G5 servers. It appears that the 3 GHz this summer will streach the definition of "summer" and "shipping".
It might be time to grumble but not yeat so scream blody murder
Originally posted by DrBoar
It appears that the 3 GHz this summer will streach the definition of "summer" and "shipping".
Now come on Summer doesn't even start till late June and ends in late September. No need to redefine.
I only found 2 articles both written by Tony Krazit that mentioned Intel's process.
PowerPC 970FX chip can be used in PCs, laptops and networking equipment
IBM blends three manufacturing techniques in 970FX
and 2 articles claiming the 970 uses SSDOI
IBM Combines CPU Design's Greatest Hits in One Chip
AMD to offer strained silicon chips
Who to believe?
Tony Krazit published in TECHWORLD and COMPUTERWORLD
or
Eric Grevstad published in CPU Planet and
Tony Smith published in the venerable publication The Register
Originally posted by rickag
I'm still confused about SSDOI. I searched IBM's site and found only references to SSDOI and nothing concerning them using the Intel nitrate process. Please note here that my search skills ain't that proficient.
I only found 2 articles both written by Tony Krazit that mentioned Intel's process.
PowerPC 970FX chip can be used in PCs, laptops and networking equipment
http://forums.appleinsider.com/showt...460#post601460
http:/2004/04/20 Taiwan: TSMC (2330 TT, NT$59.5), UMC (2303 TT, NT$31.5): TSMC and UMC benefit from poor yield of IBM's 12" fab
IBM facing poor yield of 0.13um and 90nm
Main customers such as nVidia and Qualcomm are switching back to TSMC and UMC
Maintain BUY recommendations on TSMC and UMC
According to the "Economic Daily," nVidia, Qualcomm, Broadcom, Cisco and Xilinx are switching back from IBM foundry to TSMC and UMC due to the lower yield of IBM's 12" fab. IBM's lower yield of 0.13um and 90nm, and reducing IP revenue are causing IBM's 1Q losses in the semiconductor division.
Analyst comment:
1) We believe that IBM's lower 12" yield in 1Q04 will benefit TSMC and UMC going forward. The order rotation from IBM back to TSMC and UMC is expected to begin in 2Q04. We expect TSMC to show 8-10% and UMC 10% QoQ sales growth in 2Q04.
2) We expect UMC to benefit from capacity support from SiS and He-Jian beginning in 2Q04, which to show stronger revenue growth momentum than TSMC. However, we retain our BUY on both TSMC and UMC, due to their leading position in the technology universe, and capacity ramp-up, respectively.
/www.taiwanresearch.com/breaking/breaking_03.asp
Originally posted by gensor
IBM facing poor yield of 0.13um and 90nm
Main customers such as nVidia and Qualcomm are switching back to TSMC and UMC
Maintain BUY recommendations on TSMC and UMC
.
.
.
Analyst comment:
1) We believe that IBM's lower 12" yield in 1Q04 will benefit TSMC and UMC going forward. The order rotation from IBM back to TSMC and UMC is expected to begin in 2Q04. We expect TSMC to show 8-10% and UMC 10% QoQ sales growth in 2Q04.
Ouch, bad news indeed. And there was some time ago Nr9 (I think) who talked about the superior (to IBM) fab capabilities of TSMC and UMC
Originally posted by PB
Ouch, bad news indeed. And there was some time ago Nr9 (I think) who talked about the superior (to IBM) fab capabilities of TSMC and UMC
This is old bad news. IBM claims to have cracked their 90nm problems and their 970FX production is ramping. This will allow them to draw customers back from the competition. The thing that Nr9 never seemed to "get" was that running into a few problems is not a good reason to throw in the towel. There is value to making the investment and toughening it out to be able to compete in this market with relatively few players. At the very least its not a good idea to have the world's silicon chip production all centered on a contested small island off the coast of China, sitting right on the "Ring of Fire".
Originally posted by gensor
Here is April 20 news from Taiwan on IBM yield issues. My question is why does Apple stick with losing IBM?
If you're asking that question you don't understand the services IBM is providing to Apple. IBM isn't simply providing fab services to Apple, but also a complete PowerPC chip design(s) and future.
Your question should have been "Why isn't IBM outsourcing their fab process?". And I think someone already answered that question.
You know what else is interesting though? Mac people are b**ching about processor speeds! Yes, as a power user myself I know that every hertz matters, but I bet a lot people griping about not having 3GHz are among the 90% of computer users that never make full use of the power of their system.
-ipod
</$0.02>
Originally posted by ipodandimac
While i think apple will strive for 3GHz, it is absolutely necessary for them as a company to update by WWDC, no matter what the speeds are. Powermac sales have slipped every quarter, and extending their drought to longer than a year won't sit well with investors and power users. If there aren't 3GHz at WWDC, then I think we'll see them late October/early November.
I agree with you... if Apple was to release a PowerMac update with a new Superdrive, Dual 2.4Ghz G5 and ATI Radeon 9800 I'd be happy to buy it. I just don't want to purchase the current Dual 2.0Ghz which is over 7 months old now. I also don't really care about the promised 3Ghz... it would be nice, but it's not needed.
Originally posted by Imergingenious
dual 3GHz G5's with on die memory controllers will destroy the competition until 2006. Here's to hoping it comes out before then.
I don't think you'll see the on-die memory controllers, but you may see improvements to things which mitigate the latency problems. Imagine dual 3+ GHz G5's that run two hardware threads each at full speed. That is largely equivalent to the quad-core machines people have been lusting after for a decade, only cheaper. This I think you might see by early 2005 (conservatively).