Leopard and the Many Macs Left Behind
From Steven Edwards:
Hello Dan:
In my lab, the following five Macs are all left behind due to the
requirements of Mac OS X 10.5: - Tip Jar
Thanks a lot, Apple. What a great way of rewarding faithful
customers by loading the OS with useless eye candy along with other
software bloat in an attempt to motivate new hardware purchases.
Steven,
Thanks for sharing your frustration. It's a shame that
some Macs introduced in 2003 aren't supported by Leopard - and no Mac
introduced prior to 2001 is. I can understand drawing the line at a G4
processor because of all the eye candy, but I can't understand why
Apple isn't supporting older G4 Power Macs with AGP graphics, let alone
800 MHz dual G4s while supporting 867 MHz single CPU models.
We will be following attempts to install OS X
10.5 on unsupported hardware as soon as it officially ships. It's a
shame Apple feels it has to dictate some arbitrary MHz number and
prevent users from deciding for themselves how fast or slow a system
they want to run it on.
Dan
Re: 60 Mac Models Left Behind
From Lee Wilmeth:
Ted,
I really had to laugh when I read this article [60 Mac Models Left Behind:
The Ridiculously High Cost of Leopard
]. My only outrage is the fact that you took
such umbrage with the minimum requirements. Oh my word, only Apple
computers less than 5 years old can run the latest and greatest Apple
OS. Really, this is your greatest complaint?
Basically, your argument could be explained in another way: How dare
you plan to sell an OS upgrade that won't work on my computer.
Apple must spend more time making sure my 5 or 6 or 7 or 8 year
old system can use the latest and greatest OS available. Or, if more
time won't make the software work on my computer, then degrade
the experience for everyone else until it does.
Personally, as a Mac user for over 20 years, I applaud Apple's
decision to limit Leopard to computers that can provide a performance
level appropriate to the task. Let's face facts, one reason Apple can
provide a software upgrade every couple of years is because they have
made a business decision to limit the number of machines that will run
the most current OS. Significant time and effort is saved by excluding
older machines. Not to mention, Apple is a hardware company. And you
can't sell new hardware if Apple's latest and greatest software runs on
very old machines. You do want Apple to remain in business, right? I
don't think it is wrong for Apple to attempt to push hardware upgrades
via compelling software updates.
I'm sure you are aware that there is another software company that
spent 5 years making sure its latest OS would be backward compatible:
It's called Vista. At least Apple's hardware requirements approach some
semblance of reality. As a IT professional for 12 years, this is how
Vista's requirements basically work out: Take whatever Microsoft
recommends and double the recommended suggestions. So, the
1 GHz CPU recommendation becomes 2 GHz, the 1 GB RAM
recommendation becomes 2 GB, and the 15 GB hard drive becomes 30
GB.
Apple spends 2.5 years working on an upgrade to their premier OS,
and every hardware system sold for about the last 5 years will be able
to use it. Microsoft spent 5 years creating their latest OS. There are
few, if any, computers from 5 years ago that can run Vista. There are
very few computers from 2 years ago that can run Vista, and most
computers sold at retail stores today barely meet the realistic
requirements for Vista.
In the end, I find your article specious at best. Faulting Apple for
limiting Leopard's use to computers less than 5 years old seems
childish. The "I want it because somebody else get' to have it"
approach is a fallacy. I won't accept that argument from my daughters
when they demand "fairness". I certainly feel the need to call my
fellow Mac users on it as well.
And yes, I realize this article is on a website dedicated to older
Macs. I've used Low End Mac as a resource for years. I understand and
even empathize with anyone that wants a newer system, but can't, for
whatever reason. I'm also realistic enough to understand the
limitations that exist due to older, slower hardware. But to take Apple
to task for a decision that has significant business and technically
reasoning behind it, deserves an opposing response.
Sincerely,
Lee
Lee,
Thanks for sharing your letter to Ted. While all of us
recognize that Apple has to draw the line somewhere and understand that
with each new version of OS X some more models will be left
behind, the sheer scope of that appalls us.
We can understand cutting off the G3 Macs, as the eye
candy (the interface, not the underlying operating system) requires a
lot more graphic processing power. So the Pismos, G3 iBooks, and G3
iMacs are left behind. That much we grudgingly accepted.
What frustrates us is the sheer arbitrariness of 867
MHz. In the previous developer release, it was 800 MHz. Before that, it
was even lower. And we know that a dual processor Power Mac is a lot
more powerful than a single processor one, so it just annoys us that a
single CPU Power Mac G4 at 867 MHz or a 12" PowerBook G4 at 867 MHz is
supported, but a dual 800 MHz Power Mac G4 isn't.
And why 867 MHz but not 800 MHz? Does an 8% difference
in CPU speed change Leopard performance from acceptable to
unacceptable? And why not support the 733 MHz and dual 533 MHz Digital
Audio Power Macs - the former is just 18% "too slow", and the latter
should offer comparable performance to a single 867 MHz CPU.
It would have made more sense for Apple to draw the
line based on something other than MHz, especially since Apple Inc. has
been selling dual processor Power Macs since
1996 and dual G4 systems since July
2000. Why not draw the line at AGP 2x or AGP 4x graphics? 867 MHz
is just too arbitrary.
We also disagree with you that Apple is a hardware
company. As we've said many times, Apple is a solutions company: They
make the whole widget from hardware through operating system to
software. If Apple could sell 2 million more copies of Leopard by
supporting more G4 systems, that's another billion dollars or two in
their pockets.
Finally, knowing Ted Hodges' appreciation of older
hardware, it's probably not a matter of sour grapes. He's not
complaining that Leopard won't run on his Macs (most of which are
vintage), but that Apple is leaving so many home users, workplaces, and
schools behind unnecessarily. Why should an 800 MHz eMac, iMac, or
iBook G4, or titanium PowerBook be left out in the cold?
We have always supported people running unsupported
installations of Mac OS X (and before that, of things like making
Mac OS 8 run on 68030-based Macs), and we've learned that a lot of
users are very content to have the latest operating system run on
seemingly outdated and underpowered software. We marvel at the lengths
they'll go to to keep their old hardware current.
Dan
Can't Run Leopard? Stick with Tiger!
From Gary Kohl:
I always enjoy Ted Hodges articles, and his latest on the system
requirements for Leopard was another thought provoking piece. However,
I do disagree with some of his arguments.
First, I think the point must be made that there is absolutely
nothing wrong with 10.4 and no reason not to continue to use it. If Ted
finds the price and requirements of Leopard too high, he can continue
to use a proven, solid OS. In fact, my experience is that machines run
best with the OS that came with, or or soon after, the release of the
machine, as Apple optimizes those versions of the OS for recent vintage
machines. Certainly there was a great deal of optimization as OS X
went from 10.0 to 10.4, especially making G4s run noticeably
faster.
It's only if Ted wants the new features and eye candy that he'll
have to buy Leopard - there is no compelling reason to buy it from an
operational point of view (I would not say this for Intel Mac owners,
as I would see Leopard as a must buy simply because Apple almost
certainly is providing optimizations for those machines with it's
release).
My second point is that you can't have it both ways. The more
machines you support, the more complex the code is going to be and the
more the OS is going to cost. Making sure the OS runs on machines more
than 5 years old would require coding and man-hours that would
inevitably drive up the price, while many of the features and
optimizations would not even work on such machines due to hardware
limitations.
I consider about 5 years a reasonable cutoff for Leopard, and I say
this as the owner of a MDD 1.25 machine. I will not be putting Leopard
on that machine, even though it will run Leopard - but I will be
putting it on my Mac Pro. I'm not
interested in the 300 or so new features - I'm interested in that code
optimization. If Apple does as good a job speeding up the Mac Pro even
more with Leopard as it did the G4 with later versions of Tiger, I'll
be very happy to spend the $129.
Finally, if you want a light, fast OS with low system requirements,
there is always Linux. No, wait - have you used a Linux distro lately?
All the eye candy of both Windows and the Mac OS - and a huge increase
in system requirements to go along.
I'm just happy we already have a great OS - Tiger - to run on the
machines which will be left behind and don't need to upgrade. Tiger
still has a lot of life left in it.
Gary Kohl
Gary,
Right you are. Tiger will remain a great operating
system for a lot of G3 and older G4 users, just as some stuck with 10.3
or even 10.2 rather than continue the upgrade cycle. The only drawback
to sticking with Jaguar or Tiger is that they get increasingly left
behind by new versions of QuickTime, iTunes, Safari, etc., not to
mention being unable to touch the new apps that just aren't compatible
with older versions of the Mac OS. (It's the same problem Mac OS 9
users face with regards to a modern Web browser.)
I'm not going to rush into Leopard, as my workflow
remains dependent on Claris Home Page, a classic app, and there is no
Classic Mode in Leopard - and SheepShaver, while it works, is a far cry
from the integration I've been used to with Classic Mode.
As for Linux, I'll have to give it a try when I have
access to an Intel Mac with 2 GB of RAM and some virtualization
software. My experience with Ubuntu Linux on my Acer Aspire showed just
how underpowered a 1.4 GHz computer can be (an it was even worse with
Windows XP). I'll stick with the Mac OS for productivity.
Dan
Caches and DDR Memory
From Scott Cook:
Hey Dan,
I don't pretend to understand all the stuff PowerLogix is talking
about in this paper but I'm pretty sure I'll be buying a processor with
L3 cache when the time comes to upgrade my G4 PowerMac:
http://www.powerlogix.com/downloads/SDRDDR.pdf
"In any instance when data is not in any of the caches
(i.e., a 'cache miss') the processor must access main memory. Main
memory is always much slower than cache memory and has very large
latency. Because of this huge disparity in performance between fetching
data from main memory and getting it from the L3, the actual data
throughput (no matter how fast the L3 may be) is substantially
diminished. To put it another way, it almost doesn't matter how fast
the L3 is as long as it exists. To confirm this hypothesis, we
performed tests which compare two speeds of cache of the same type; for
example, 167 MHz SDR vs. 333 MHz SDR."
"The net result is that there is no advantage in using
DDR L3 cache memory over SDR L3 cache memory on the 7450 family
processors. There is also very little performance difference between an
L3 cache of a given type running twice as fast as another L3 cache of
the same type! Clearly it is important to have some sort of L3 cache,
but the type and speed of the L3 are almost irrelevant. In fact, the
only factor that matters to any degree, is L3 cache size, and even that
is not a huge difference (about 7% maximum difference in our benchmark
tests, which is measurable but not necessarily noticeable in day to day
usage.)"
PowerLogix did their testing in OS 9. I dunno if OS X would be any
different? Based on what I remember from the last time I was able to
locate OWC's processor test page, it looked like G4 upgrades with L3
were generally slower in clock speed, less expensive, and higher
performing in most processor intensive real world applications. I wish
I could locate that page again on OWC's website. They have it well
hidden.
I'm the kind of buyer who would be most influenced by real world
tests running applications. I work with audio/video a lot. The
QuickTime video encoding test on OWC's website was of particular
interest to me. Video encoding is the most demanding task my G4 does.
It runs for weeks sometimes encoding TV shows to H264 video podcast for
a TV network I've been working for. As I recall, video encoding thrives
on large L3 cache in OWC's processor tests. The encoding time
difference could be very large in my case; perhaps days (?)
Scott Cook
Scott,
Apple began using DDR memory because it was a readily
available, affordable commodity - even though the G4 is unable to take
advantage of its double data rate. PowerLogix is right in saying that
SDR memory would work just as well, but it's DDR memory that's
plentiful and cheap.
I searched OWC's website for benchmarks, but they're
missing in action while OWC is working on new reports. Any improvement
from an L3 cache should be the same whether running the classic Mac OS
or a version of OS X.
Here's the current range of G4 upgrades for the
Quicksilver Power Macs available from OWC:
Single CPU
- 1 GHz, 256 KB L2, 2 MB L3, $175
- 1.2 GHz 7455, 256 KB L2, 2 MB L3, $199
- 1.6 GHz 7447A, 512 KB L2, no L3, $220
- 1.6 GHz 7447A, 512 KB L2, no L3, $250
- 1.5 GHz 7455B, 256 KB L2, 2 MB L3, $260
- 1.8 GHz 7447, 512 KB L2, no L3, $299
- 1.8 GHz 7448, 1 MB L2, no L3, $348
- 2.0 GHz 7448, 1 MB L2, no L3, $400
Dual CPU
- 1.4 GHz 7447A, 512 KB L2, no L3, $350
- 1.6 GHz 7447, 512 KB L2, no L3, $388
- 1.6 GHz 7447A, 512 KB L2, no L3, $400
- 1.7 GHz 7447A, 512 KB L2, no L3, $490
- 1.8 GHz 7447A, 512 KB L2, no L3, $550
- 1.8 GHz 7447A, 512 KB L2, no L3, $590
- 1.7 GHz 7448, 1 MB L2, no L3, $599
- 1.8 GHz 7447, 512 KB L2, no L3, $629
- 1.8 GHz 7448, 1 MB L2, no L3, $699
There are a lot of factors involved here: different
generations of G4 processors, different sizes of onboard L2 cache, and
single- vs. dual-processor upgrades. The huge difference between L2
cache and L3 cache for G4 CPUs is that L2 is on the CPU itself, while
L3 (when present) is physically separated and much slower to access.
It's still faster than system memory, but there is a big penalty vs. L2
cache speed.
Additionally, the L3 cache has to be significantly
bigger than the L2 cache. It it's only twice as large, half of the data
it stores may also be in the L2 cache, so you'll typically see an L3
cache four to eight times the size of the L2 cache. With a 1 MB L2
cache, that's 4-8 MB of L3, an unwanted cost.
The benefit of a bigger cache is not linear beyond a
certain point. If the computer is doing most of its work within a
1 MB area of memory, a CPU with a 256 KB L2 cache will benefit
greatly from the presence of an L3 cache. A CPU with a 512 KB L2 cache
will benefit less, and there will be very little difference for a CPU
with a 1 MB L2 cache.
Optimized code is designed to run compactly, so the
benefit of a larger L2 cache generally outweighs the benefit of a
larger L3 cache. The exception is when you're working through huge
quantities of data, such as filtering an image in Photoshop, processing
video, or working with audio files.
So in the end, it's all a matter of "it depends" - for
audio and video work, a slower CPU with a big L3 cache may be a better
value than a faster CPU with no L3 cache. But for general computing,
you'll get better overall throughput from the larger L2 cache found in
the 7447 and 7448 - and pay the price to get it.
Dan
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.
