Old G4s Not Discontinued
In response to The January 2003
Power Mac G4 Value Equation, Ken Watanabe notes:
To add to the confusion, the former Dual 1.25 GHz G4 is technically
not even "discontinued." It's currently still offered by Apple as the
"Mac OS 9 bootable" system for $2,499. Just go to the store, first
click on the Power Mac G4, and then look for the small link that says
"Mac OS 9 Systems."
Good point. His Jobsness promised OS 9 booting machines for
those who need them through June 2003. What many don't seem to realize
is that the two "Mac OS 9 systems" Apple has at the Apple Store are the
old dual processor G4s, not 9-booting versions of the new Power
Macs.
The CPU Race
After reading The Longer Apple
Sticks with Motorola, the Behinder They Get, Thomas Eberhard
writes:
The 68030/40 in the IIfx in
1990 was the last time Motorola made a leading edge CPU. Then came the
debacle with the '040 CPU that could not keep up with the 486, forcing
Apple to migrate to PPC.
The NuBus generation was not good, but the PCI generation fared
better. The 9600/350 did
beat the crap out of most PCs, and when the B&W G3/450 was out, it was
leading edge again. However, the fast 604e and G3 were and are
manufactured by IBM, not Motorola.
With no IBM in the G4, the G4 has tormented us for 4 years and then
some. The two main benefits of the G4 are the SMP [Symmetric Multi
Processing] the OS nor applications have supported for most of the
time. The second is the AltiVec that also has had weak support and
obviously is not that important anyhow, as Apple sell G3 computers
4-1/2 years after its introduction.
Yes the G4 truly was Giga Flop CPU.
My hopes are
- That the 970 arrives ASAP (and is not called G something)
- That long overdue changes in the towers comes along. Like a sound
support better than the plain stereo we have had since the Quadra
- That the 970 quickly migrate to the iMac and eMac, and then later
to the portables. An iMac with a 1.5 GHz G4 is money invested in a dead
end desktop CPU. A equally performing iMac with a 1.x GHz 970 is an
investment in the future.
With the 1 GHz G4 dissipates 30W and the 1.8 GHz 970 is estimated to
be at 42W, there is reason to believe that all CPUs from the G3s to
high end G4 can be replaced by 970s at a rapid pace. PowerBooks with
970s in the 1.2-1.5 GHz range (about 20W) would be very nice.
I have understood that SMP support that is going from a single to
dual CPU is quite substantial works in the software. But how about
adding software support for quads? My guess would be that the step from
dual to quads is less than from single from dual, but that is just my
guess.
I have to disagree on the 68040. It was a cutting edge processor
that offered approximately 2.5 times the performance of the 68030/68882
at the same clock speed. It was definitely a cutting edge design, and
even a 20 MHz Centris 610 with
the math-challenged 68LC040 outperforms the "wicked fast" 40 MHz
68030-based IIfx.
Motorola managed to push the '040 to 50 MHz, as evidenced by several
third-party processor upgrades, but Apple didn't go there. Motorola
designed and abandoned a 68050, then went on to produce a 68060 that
from all reports would have outperformed the PowerPC 601 at the same
clock speed.
But Motorola was also working on RISC processors (their 88000). And IBM
had already implemented RISC in the Series/6000 computers. The whole
computing industry was rabid about the potential of RISC CPUs, and even
Intel made a pair of phenomenal ones, the 860 and 960. Of course, none
of these were backward compatible with the x86, so the industry stayed
in that path right through today's Pentium 4.
Apple somehow joined forces with IBM and Motorola to leverage IBM's
Power design, Motorola's fledgling 88000 effort, and Apple's need to
move from the dated 680x0 architecture to something new and modern.
Thus was born the AIM consortium and the PowerPC architecture.
The PowerPC 601 was a bit of a hack, but it worked well and allowed
excellent emulated performance for old 680x0 software. The original 603
was hobbled with too small a cache to emulate the 680x0 efficiently,
but the 603e fixed that. And the industrial strength 604 and 604e could
crunch numbers with the best - and was even used in some quad processor
systems from DayStar Digital.
Of course, the G3 put an end to all that. By designing the CPU to work
efficiently with existing PowerPC code, the G3 provided about 50% more
power per cycle than the 603e or 604e, allowing Apple to claim a
processor "up to twice as powerful" as a Pentium at the same
speed.
I'd definitely say things were cutting edge up to that point, which is
also the last point at which IBM and Motorola agreed on the future
direction of the PowerPC. IBM wanted to push CPU speed. Motorola wanted
to add the complex AltiVec velocity engine. And we all know which route
Apple chose.
As for the efficiency of additional processors, because of the overhead
involved in managing the CPUs themselves, the data cache, and memory
writes to system RAM, there is generally about a 10% decrease per
processor added. That means a dual G4 will have about 1.8 times the
power of a single G4 at the same clock speed (90% x 2), and a quad G4
system would provide roughly 2.4 times the power of a single CPU (60% x
4).
This is a rough figure, and the more that overhead can be reduced, the
more viable it is to increase computing power by adding more and more
processors. IBM's Power architecture, on which the PowerPC 970 is
based, is explicitly designed for two CPUs on a single core and up to
32 CPUs in a computer. To make that at all practical, IBM had to find a
way to reduce MP overhead to well under 3% per processor added, so it's
conceivable that a dual PPC 970 system would offer 190% as much power
as a single CPU (95% x 2), and a quad processor system would likewise
provide 360% as much power as one CPU (90% x 4).
For those who need ultimate power, the step to quad processors could be
just what the doctor ordered.
In a follow-up email, Eberhard writes:
Perhaps I phrased it poorly. I agree that the '040 was better than
the '030 (a 25 MHz LC 475, a
great machine for its time, did beat the 25 MHz IIci by quite margin I remember when
trying out Marathon :-) I just got the impression that the '040 had
problem keeping up with the 486.
And I totally agree about the 604E and the G3s.
The G4 was questionable from the start. The main advantages SMP was
not supported by either OS nor application and as such useless for
several years. AltiVec does give a boost of some things in some
applications, but it has largely been a disappointment. The real sad
thing, though, is not the initial manufacturing problems but the fact
that there has been no sign of reducing the 2x MHz gap during more than
3 years. In fact, the trend seem to be closer to sliding back to a 3x
MHz gap.
I have understood that rewriting application from one to two CPUs is
a substantial work, but I have not found any information about the step
to quads from duals.
The efficiency is of course very important in multi CPUs. With a 10%
penalty per CPU the 1.8 to 2.4 really makes the dual setup the useful
limit ( doubling the CPU cost for a 60% gain) and having 8 CPUs would
be slower than one! With a 5% penalty the dual score of 1.9 instead of
1.8 is marginal but for the quad 3.6 instead of 2.4 would be a very
substantial boost with a 90% gain by doubling the CPU cost
I will save my money for the next generation non-G4. (That is, I
will not buy a box were the CPU has been grafted in what basically is
the previous generation like the NuBus PPC line and the Beige G3s. Both of these lines
was far less impressive than the following PCI PPC and B&W G3 even
if they had the same CPUs
Thanks for the extensive and informative reply.
For most of the past year, the fastest G4 has been roughly 46% as
fast as the fastest Pentium 4 in clock speed. Now quite as bad as a 3x
ratio, and a bit better than the 42% figure it was a bit further back,
but Motorola is still lagging.
I don't understand your antipathy toward the beige G3, which has a
whole new motherboard in the same basic case as the earlier Power Mac 7500. The 66 MHz bus was
one-third faster than in earlier Power Macs and the onboard video was
better as well. There were some issues with slave drives, and the IDE
bus was awfully slow (at least by today's standards), but it was a nice
step up from earlier models.
Still, the blue & white G3 with its 100 MHz system bus, faster IDE
bus, and support for both USB and FireWire was a much nicer computer.
Waiting for the second generation of hardware generally gives you a
better machine with fewer teething problems.
Back to WYSIWYG
Responding to Fulfilling the Promise
of Aqua and the Quartz Rendering Engine, Bruce McLaughlin
writes:
Nice article, although I'd like to add one thing and take the whole
concept of fluidity one step farther.
Aqua really is beautiful, but do these icons have to be so large?
Why can I only choose between this size of icon and no icons at all -
why not a smaller set of icons?
Control-clicking on the toolbar gives you quite a few display
choices, although not small icons as you request.
Overall, Mac OS X cries out for a much larger display than the 512
x 384 or 640 x 480 screens of early Macs. Under OS X, an 800 x 600
display seems crowded, and 1024 x 768 seems to be the least resolution
you want to really work comfortably.
I completely agree.
How about the same kind of thing in AppleWorks instead of the
discrete 50%, 66.7%, 100%, 200%, and other settings? Sure, leave the
standards in place, but why not add a slider that lets us zoom in or
pull back fluidly in real time?
The dock does it. The Finder does it. The technology is already there
in the Quartz rendering engine. Apple ought to tap into it in other
areas as well.
Why not return the Mac to WYSIWYG? One inch on the screen equals one
inch in real life. I want one to one text size as the system wide
default, with a preference panel for setting it up. My ideal system
would keep the text size the same at any display resolution. Those who
want everything on their screen smaller or larger can adjust the
WYSIWYG preference accordingly, independent of display resolution. As
you say, the technology is there.
I think Apple will wait until Classic mode disappears before trying
anything like this. Another two years is my guess.
Thanks for the interesting article.
Although I think Apple should offer 1:1 correspondence between
screen size and printout as an option, I don't think it should be the
default behavior. Think about it - the 12" and 14" iBooks both offer
1024 x 768 displays, and those who have older eyes or need to share the
screen in a presentation choose the larger iBook so everything displays
a bit bigger.
They can do that because both screens show exactly the same amount of
information.
Now imagine going into an Apple dealer and seeing the two iBooks side
by side if their displays corresponded to the 72 points per inch world
of the typesetter. Instead of the 14" iBook displaying things bigger
than the 12", it would display everything the same size and simply
display more information. Conversely, the 12" would be a harder sell
because it would show so much less if there were a 1:1 correspondence
with physical screen size.
That's what some people expect of a different display size, but many
pick it because they want or need things displayed quite a bit larger
than on the 12" iBook.
Still, Apple should provide some system options where the screen would
default to current behavior, have a 1:1 physical correspondence as a
standard option, and also offer behaviors that would make the make more
palatable to the visually challenged.
The underlying technology exists. I hope they will make it
accessible.
Affordable Comm Slot II Ethernet
Card
Scott Solin writes:
I been reading your site a ton and getting a lot of great info off
it.
I wanted to add DSL service to my Power PC Performa 6400/200, but with a video card
and a USB card in there. That took up all my PCI slots. I read on your
site that a Com Slot II ethernet adapter card is available. I did a
search around and found most around the $70-90 range! Then I did a
search out of the blue and found a great store! It's in Hayward, CA, so
shipping is kinda steep with tax and all. But they have the card in
stock for $14.95!
I thought this was an exceptional deal and just wanted to tell you
and maybe let others know on the site. Here is the link to it in case
you wanna check it out. http://store.yahoo.com/justdeals/faret10basco.html
I enjoy the site and the emails everyday. Keep up the good work.
Gotta keep the old Macs current and up to date!
Thanks again!
Scott, thanks for the link. I think you'll make a lot of people
happy with this source for inexpensive Comm Slot II ethernet
cards.
UPDATE: Several readers have noted that the above card is marked as a
Comm Slot card, not a Comm Slot II card. Although the two slots are
different and some cards only work in one or the other, some cards are
designed to be compatible with both slots. Based on the writer's
success, I have to assume that this is one of those cards.
dk
Will OS X Speed Up My Power Mac?
I have a Power Mac 8500
that's been upgraded to a G3/300 with 206 MB of RAM. It's currently
running Mac OS 9.1. Would it run faster if I used XPostFacto to install
OS 10.1.x on it? Please let me know.
Anything but. Your Power Mac will run slower with OS X
installed. Mac OS X is a very different, very powerful, very nice
operating system, but because of all the work going on in the
background, it is not as fast as the classic Mac OS.
For more on the pros and cons of upgrading, check out 10 Forward.
Dan Knight has been publishing Low
End Mac since April 1997. Mailbag columns come from email responses to his Mac Musings, Mac Daniel, Online Tech Journal, and other columns on the site.
