Just rad an article over at http://www.macosrumors.com, and it was pretty interesting:
-----------
At Apple, we are quite pleased with the way the G5 has progressed. As of noon, we received version 0.7 of the G5. Altivec performance is now at par with equivalent clock speed 7460's. We spent a late night Friday night fitting prototypes with the new revision, and spent the day saturday doing various tests.
Yields are now just at the commercially acceptable level. Good news is that clock speeds have been improved to the point where 1.6 Ghz chips will be in adequate quantities. Another clock speed record was also set: 1 chip tested at 2.8 Ghz, 2 tested at 2.6 Ghz, 13 tested at 2.4 Ghz, 13 tested at 2.2Ghz, and 54 tested at 2Ghz. This shows that the G5 has tremendous potential at reaching high frequencies, being this early in its life. This is in sharp contrast to Intel's Itanium, which when I spoke to an Intel engineer at the semiconductor forum, they still are not getting sufficient yields above 800Mhz, an yields on current processors are very, very poor, hence the steep price of the Itanium. Mckinley is not faring too well either, progress has not been very good on increasing its clock speed for release sometime next year. The aim is to speed bump the G5 to between 2Ghz and 2.4 Ghz for Macworld New York. Above 1.6 Ghz, the G5 will be produced in 400Mhz increments.
Apple could theoretically sign off now, but Jon Rubinstein wants to go through one more revision. All the critical bugs have now been worked out, but there are a couple of minor optimizations that will go into revision 0.8, which is due within two weeks. Likely, these slight optimizations will result in version 0.8 being declared 1.0, and mass production will go on throughout December to get a critical volume of chips for a Mid-December production run of Power Mac G5's.
Anyone considering buying a G5 better be forwarned: the chip price may mean that Apple may not be able to offer G5 Power Macs for the same price as current G4 models. There has been talk between Steve Jobs, Jon Rubinstein, and Phil Schiller about possibly offering 7460 G4's at the low end in the professional models in two configurations, which would also appease Motorola. Apple would have five models of pro desktops until G5 prices fall low enough to warrant having them in the low-end pro models. There is talk of two 7460 G4 models, and 3 G5 models. Talk is that the low end G5 model will sell for slightly more than the current 867Mhz G4. The G5 towers will also sport the quicksilver enclosure initially, which will be changed at Macworld New York. People should understand that even though the G5 is considerably more expensive than the G4, it is a steal considering that we are getting at least 60% overall instructions per cycle than Intel's Itanium, and that it is a 64-bit processor. The 32 bit version of the G5 will be solely targeted towards embedded applications, as 32-bit addressing is no longer adequate for desktop applications.
The long awaited LCD iMac will also make its debut at Macworld San Francisco. It will be available at up to 1Ghz, 900Mhz being the scenario should yields of IBM's next generation G3's not be sufficient enough at 1Ghz.
Steve Jobs has very ambitious plans for Apple's processor strategy. He recently said "We've been stuck with the G4 for over two years, that's too long". His intentions are that the G5 have a life of 18 months in the Pro models. He wants the G6 to hit initial silicon between next December, and February 2003, and release it in mid-2003. Initially, the G6 will be fabbed with a 0.1 micron process moving to .07(.065) micron. It will be built upon the HIP 8.0 process, which is still not quite finalized. It will feature Altivec II, which promises to at least double performance of the current Altivec. Early estimates are that it will contain over 100 million gates. The G6 will be introduced at between 4.5 and 5Ghz and scale up to 10 Ghz.
This week Apple has committed itself to going beyond the G6 to build a G7, and maybe beyond. Apple is looking at Motorola's recently announced Gallium Arsenide technology to give this chip insanely high clock speeds. Talk is the G7 could go as high as 20 Ghz. The G7 would debut in early to mid 2005. This renewed hope with the PowerPc architecture is in light of the fact that Cisco Systems has committed to being a significant customer of G5's for their high-end routers, and Silicon Graphics being in the last stages of abandoning development on its R16000 and R18000 processors as a cost-cutting measure. It looks very likely it will sign a commitment with Apple, IBM and Motorola within a month. It has prototype G5 chips in a prototype workstation of theirs, and is hard at work developing Irix 7.0 for the G5.
In terms of future G5 development, work is well underway on the 8510, which is a low-power SOI LoK dielectric version of the G5. It is due out in late Q4 2002, and it will be an IBM product fabbed with its 0.1 micron process. Work is also progressing on the 8550, which is due out Q1 2003. It will be a 0.1 micron chip built upon SOI LO-K dielectric. It is a candidate to receive Altivec II if it is completed in time.
Relationships between Apple and Motorola as of late could very well be described as Jeckyll and Hyde. Just three weeks ago, Steve Jobs said "I am going to sue the ass of those guys at Motorola. At the last minute, they f*cked us up. They told us we would have 1Ghz G4's, and days before Macworld, the ****ers told us there was a defect which would cause them to fail above 900Mhz." Days later, the relationship becomes cordial again when Motorola shows renewed interest when SGI and Cisco Systems start looking into the G5.
-------------
What do you guys think?
-----------
At Apple, we are quite pleased with the way the G5 has progressed. As of noon, we received version 0.7 of the G5. Altivec performance is now at par with equivalent clock speed 7460's. We spent a late night Friday night fitting prototypes with the new revision, and spent the day saturday doing various tests.
Yields are now just at the commercially acceptable level. Good news is that clock speeds have been improved to the point where 1.6 Ghz chips will be in adequate quantities. Another clock speed record was also set: 1 chip tested at 2.8 Ghz, 2 tested at 2.6 Ghz, 13 tested at 2.4 Ghz, 13 tested at 2.2Ghz, and 54 tested at 2Ghz. This shows that the G5 has tremendous potential at reaching high frequencies, being this early in its life. This is in sharp contrast to Intel's Itanium, which when I spoke to an Intel engineer at the semiconductor forum, they still are not getting sufficient yields above 800Mhz, an yields on current processors are very, very poor, hence the steep price of the Itanium. Mckinley is not faring too well either, progress has not been very good on increasing its clock speed for release sometime next year. The aim is to speed bump the G5 to between 2Ghz and 2.4 Ghz for Macworld New York. Above 1.6 Ghz, the G5 will be produced in 400Mhz increments.
Apple could theoretically sign off now, but Jon Rubinstein wants to go through one more revision. All the critical bugs have now been worked out, but there are a couple of minor optimizations that will go into revision 0.8, which is due within two weeks. Likely, these slight optimizations will result in version 0.8 being declared 1.0, and mass production will go on throughout December to get a critical volume of chips for a Mid-December production run of Power Mac G5's.
Anyone considering buying a G5 better be forwarned: the chip price may mean that Apple may not be able to offer G5 Power Macs for the same price as current G4 models. There has been talk between Steve Jobs, Jon Rubinstein, and Phil Schiller about possibly offering 7460 G4's at the low end in the professional models in two configurations, which would also appease Motorola. Apple would have five models of pro desktops until G5 prices fall low enough to warrant having them in the low-end pro models. There is talk of two 7460 G4 models, and 3 G5 models. Talk is that the low end G5 model will sell for slightly more than the current 867Mhz G4. The G5 towers will also sport the quicksilver enclosure initially, which will be changed at Macworld New York. People should understand that even though the G5 is considerably more expensive than the G4, it is a steal considering that we are getting at least 60% overall instructions per cycle than Intel's Itanium, and that it is a 64-bit processor. The 32 bit version of the G5 will be solely targeted towards embedded applications, as 32-bit addressing is no longer adequate for desktop applications.
The long awaited LCD iMac will also make its debut at Macworld San Francisco. It will be available at up to 1Ghz, 900Mhz being the scenario should yields of IBM's next generation G3's not be sufficient enough at 1Ghz.
Steve Jobs has very ambitious plans for Apple's processor strategy. He recently said "We've been stuck with the G4 for over two years, that's too long". His intentions are that the G5 have a life of 18 months in the Pro models. He wants the G6 to hit initial silicon between next December, and February 2003, and release it in mid-2003. Initially, the G6 will be fabbed with a 0.1 micron process moving to .07(.065) micron. It will be built upon the HIP 8.0 process, which is still not quite finalized. It will feature Altivec II, which promises to at least double performance of the current Altivec. Early estimates are that it will contain over 100 million gates. The G6 will be introduced at between 4.5 and 5Ghz and scale up to 10 Ghz.
This week Apple has committed itself to going beyond the G6 to build a G7, and maybe beyond. Apple is looking at Motorola's recently announced Gallium Arsenide technology to give this chip insanely high clock speeds. Talk is the G7 could go as high as 20 Ghz. The G7 would debut in early to mid 2005. This renewed hope with the PowerPc architecture is in light of the fact that Cisco Systems has committed to being a significant customer of G5's for their high-end routers, and Silicon Graphics being in the last stages of abandoning development on its R16000 and R18000 processors as a cost-cutting measure. It looks very likely it will sign a commitment with Apple, IBM and Motorola within a month. It has prototype G5 chips in a prototype workstation of theirs, and is hard at work developing Irix 7.0 for the G5.
In terms of future G5 development, work is well underway on the 8510, which is a low-power SOI LoK dielectric version of the G5. It is due out in late Q4 2002, and it will be an IBM product fabbed with its 0.1 micron process. Work is also progressing on the 8550, which is due out Q1 2003. It will be a 0.1 micron chip built upon SOI LO-K dielectric. It is a candidate to receive Altivec II if it is completed in time.
Relationships between Apple and Motorola as of late could very well be described as Jeckyll and Hyde. Just three weeks ago, Steve Jobs said "I am going to sue the ass of those guys at Motorola. At the last minute, they f*cked us up. They told us we would have 1Ghz G4's, and days before Macworld, the ****ers told us there was a defect which would cause them to fail above 900Mhz." Days later, the relationship becomes cordial again when Motorola shows renewed interest when SGI and Cisco Systems start looking into the G5.
-------------
What do you guys think?
). Now R&D for a DP G5 has to be