la saga du G5 cntinue ( assez long )

[jpblcm]

( pardon, mais c'est encore de l'ingliche) article issu de la geeknewsletter - geek.com

"0) INSIDE THE CHIPS - Will Macs say "IBM Inside"?

Since 1984, when the first Macintosh arrived from Apple, Apple has
used Motorola processors. After 18 years, it may be time for a change.
Apple stuck with the 680x0 processors from Motorola for nearly 10
years and then switched to the PowerPC architecture in 1994, beginning
with the PowerPC 601. At that time, IBM, Motorola, and Apple were all
working together on the PowerPC.

Over the years, Motorola and IBM cooled off, but both still kept
producing PowerPC chips. IBM has recently concentrated on the Power3
and now the Power4 line of server processors, and speedy low-power G3
PowerPC chips for Apple laptops. Motorola has ruled the Apple Mac
performance roost with its respected G4 PowerPC and its AltiVec
instruction set since 1999. The first G4 PowerMac showed up at a
decent 400MHz top speed. (There's more Apple history here:
http://www.apple-history.com/history.html)

Now that three years have passed, you'd expect a big performance boost
above those early G4 chips. Unfortunately, Motorola has run into many
delays and missteps in its attempts to speed up the G4. The 500MHz G4
was released on February 16th, 2000. There were no faster G4 chips
until January 9, 2001. That was almost 11 months' worth of no new
speed grades or improvements in performance. The last set of delays
left the G4 chip at a 1GHz top speed for over 7 months before a 1.25
GHz G4 reached the market just this week.

The next-generation PowerPC chip from Motorola, code-named the 8500
series, seems stuck at 1GHz. The 8540 has been on the market since
April 2001, but Motorola hasn't seen fit to update the clock speeds,
even though it has added new features with the 8560 model. So far,
Apple has steered clear of these Motorola 85xx chips, confusing many
who were expecting a move to 64-bits and a new "G5" code-name.

Several months ago, I received an e-mail from an anonymous source
stating that the Motorola G5 desktop program was canceled. Before
that, I got an e-mail with roadmaps and various details on the G5,
G5+, G6, and G7 chips from Motorola. You can still see the G5+, G6,
and G7 details here:
http://www.geek.com/procspec/apple/g5+.htm - G5+
http://www.geek.com/procspec/apple/g6.htm - G6
http://www.geek.com/procspec/apple/g7.htm - G7

The G5 chart has been updated to include information on Motorola's
shipping the 85xx series of chips:
http://www.geek.com/procspec/apple/g5.htm

The e-mail was strange, as Apple had simply been using Motorola's
"embedded" chips right in the Macs (not special "desktop" versions).
Was that not possible with the 8500 series chips, or was something
else going on? Another anonymous e-mail arrived soon after stating
that although Motorola G4 chips would be improved and used in Apple
computers in an ongoing basis, the Motorola G5 wouldn't make the cut.
Furthermore, the e-mail stated that Apple would use IBM chips for its
top-end machines, and continue to use Motorola chips only for low-end
and low-power machines (iMacs, iBooks).

Last week, MDR Online put up its agenda for the October 15
Microprocessor Forum 2002. The second presentation highlighted after
the keynote will feature Peter Sandon of IBM's PowerPC Organization.
Mr. Sandon will be discussing a 64-bit PowerPC chip based on the
Power4 server chip design, but, unlike earlier Power4 chips, this new
chip is "... designed for desktops and entry-level servers." As well,
it will feature a vector processing unit with over 160 vector
instructions. The 64-bit chip will also feature an unnamed system
interface capable of 6.4GB/s:
http://www.mdronline.com/mpf/conf.html#day1_2

The 6.4 GB/s external interface on the new IBM chip could be using
Motorola's RapidIO, but perhaps more likely is the use of AMD's
HyperTransport that also will just happen to run at 6.4GB/s in AMD's
Opteron chip. Another big question about the new IBM chip, assuming
that it will be aimed at the Macintosh, is how that vector processing
unit will operate. Will it be compatible with AltiVec in some way, or
a whole new architecture?

More details on IBM's chip come from Silicon.com. Writers there say
that the new chip can execute 8 instructions per clock cycle and
support symmetric multiprocessing. As well, IBM seems to be trying for
a 2GHz clock speed.
http://www.silicon.com/public/door?6004REQEVENT=&REQINT1=
55033&REQSTR1=silicon.com

Another piece of potential "evidence," which may just be an unrelated
coincidence, is that the next version of OS X will no longer boot into
Mac OS 9. The next release of OS X, a.k.a. "Pinot," is expected to
arrive at Macworld 2003, the same date I've seen it said that IBM's
new 64-bit desktop chip will make its debut in a Mac. It does seem
possible that OS X could add support for another chip in the PowerPC
family, and that Apple may not want to bother spending resources to
get OS 9 to work with it:
http://www.geek.com/news/geeknews/2002Aug/bma20020805015732.htm

Based on the earlier leaks and the new information, as well as
Motorola's seeming continued inability to boost the clockspeed and
performance of its chips to a competitive level (aside from making
Apple sell systems with two processors), it seems the logical choice
that this new chip from IBM will be included in Apple Macintosh
computers someday soon.

Perhaps all of those "G5" roadmaps that have been leaked have been for
IBM's chips? Reports did appear now and then about Motorola G5 chips
ramping up to 2GHz, but maybe it was no longer a priority for the
company:
http://www.geek.com/news/geeknews/2001apr/bch20010410005320.htm
http://www.geek.com/news/geeknews/2001sep/bma20010917007901.htm

Those reports of speedy Motorola chips are one thing that gives me
pause about fully believing that IBM has simply taken over the high-
end PowerPC market and will automatically win Apple's business.
However, there has been plenty of talk about how unprofitable
Motorola's chip division is. It may not make sense for the company to
spend time and money developing a faster G5. Hopefully, on October
15th we'll get some clues as to how IBM plans to use the 64-bit
PowerPC desktop chips--or at least a customer list so that we can see
if Apple is in there. I think if IBM's new chip does make it into
Apple computers, that will be a good thing for Apple. It will make the
Power Mac single processor configuration a force to be reckoned with
once again.
"
déjà que je ne savais pas quoi faire, alors là