dongmin said:
You're missing my point. The Power 5 is targetted for servers. The 975, or whatever the Power 5 derivative will be called, is intended for the desktop.
The Power4 was rated at 1.3ghz at the end of its life cycle, and yet the 970 was introduced on at 2.0ghz and with the addition of a SIMD unit. The respective die sizes were 414mm^2 and 267mm^2, with more than a
halving of the transistors from 170 million approximately 58 million. The first chip is a 180nm process, the second a 130nm process.
By comparison, the Power5 is 2 389mm^2 cores on a single chip, manufactured at 130nm CuSOI (copper-process silicon-on-insulator) with
276 million transistors, an on-chip DDR/DDR2 memory controller, interconnects between processors, and other necessary aspects of the multi-core design. If IBM can coax 2.0ghz+ (the 970 has been rumored to reach clocks of at least 3.2ghz in engineering samples, but not quanitity) out of the aging Power4 line, imagine what a well-designed Power5 companion chip could do. Drop the transistors by an equal amount (276 x 0.34 = 93.8 million) and you're in the ballpark, especially with the improved manufacturing process (copper and SOI).
Also, the 970 is intended for servers, thought not the kind of Big Iron and mid-level designs that the Power series chips are aimed at.
These systems have VERY DIFFERENT tolerances for heat.
Which is why you can easily build an Opteron workstation with two or four processors, provided you can find the motherboard? I'm not sure that you're right on this point, especially since AMD and Intel are licensing manufacturing processes that IBM pioneered (SOI, SSOI, and so on).
The point of producing the 975 at 90 nm is to reduce the heat to manageable levels for a desktop. Yes, the 130 nm is a much more reliable process than 90 nm, but if a 130 nm 975 doesn't work in a desktop (and 1U servers) heat-wise, it's useless.
Wait, wait, wait... People are practically pissing themselves thinking that the PowerBook G5 is around the corner, with all the heat issues of the 970 over the G4, and yet you think that a 975 couldn't be put into a desktop? For one thing, the 970s that are currently in the towers are 51 watts
each at 2.0ghz and there's not really been a single problem cooling them efficiently and quietly because of Apple's case design. The 975 is supposedly around 65 watts at 3.0ghz, and if so, that's still barely hotter than a single Prescott 3.4ghz. That also assumes that PowerTune hasn't been refined even further, and that power management couldn't keep the heat down.
I found at least two rumors that says the Power 5 derivative will be starting at 90 nm (note that the Power 5 derivative was originally called the 980 by rumor-mongers):
What it's called is largely marketing, you know. It could just as easily be called the PowerPC 10 and it wouldn't make any difference to the features or capabilities of the chip in question.
"Following the microprocessor forum, IBM presented Apple with a handful of PPC 980 alpha samples to begin work on the next generation Powermac due out in 9-12 months. ... The reason why the 980 is appearing only 12 -16 months after the 970 is that Apple chose to engage in parallel development with the Power 5, rather than wait 12-18 months after the fact. The 980 samples that were given to Apple were 90nm chips, as opposed to 130nm chips for the PPC 970 and the Power 5.
Interestingly, they got the die size on the Power5 right, but the thing that struck me about this rumor is that the date it's posted is October 29th. Let's call that November for the sake of argument... What's 9 months after November? August. If on-schedule, neither behind nor ahead, that could be demo at WWDC with one of the machines that will go on sale, and then delivery in August.
Sadly, the latter half of that rumor was obviously false. The 90nm problems were more extensive than thought, and the revisions didn't come in February. At least Fishkill is running almost on-target now, though, according to IBM's recent press.
Apple and IBM have been working on parallel development of the Power5 and PPC 980. The PPC 980 is a single core version of the Power 5. While prototype forms of this chip exist, it is almost a year away from shipping in Macs.
Once again, this points at an August release of 975/980 chips, as the rumor is from July of last year. It's intersting to note that, even at that date, they were claiming that 90nm was necessary for heat, but the server chips are running at dual-core, dual-processor without any issues.
Steve's comment of 3GHz in 1 year will not be accomplished with the G5 (970) - which will top out at 2.6-2.8GHz. The PPC 980 will start at speeds of 2.6-3GHz and top out around 4.5-5GHz. The G5 will make its way into PowerBook lines in Jan/Feb, Xserve's later this year, and iMacs in approximately one year.
This also goes along with what I've been saying, since I've been suggesting for a while now that Steve never once said that the PowerPC 970 would be at 3.0ghz this summer, merely that Apple and IBM would be there. A release of a new chip, one built from the ground up for the performance Apple would like, is much more palatable and desirable than the relatively stopgap-feeling 970.
As always, just as Arn cautioned, take it with a grain of salt. Just be like me and hope for the best, but don't let yourself be too disappointed if it doesn't happen. Dreaming is fun.
