OSX with Excel X speed problems

[The Ledge]

I have an AppleMac network with a G3 Server running OSX, a mix of G3s
(running OS 9.1, 9.2 and OSX) and a PowerPC G4 (867 MHz 640Mb) running
OSX 10.2.6.

We are running Excel v.2001 on the OS9 machines, Excel X for Mac
Service Release 1 on the OSX machines.

Excel runs considerably faster on the OS9 machines than the OSX
machines.

For instance, take an Excel spreadsheet with 7,500 rows by 15 columns
of data located on our server.

Using the vertical scrollbar, from a network 400 MHz G3 running OS9.1
and Excel 2001, it takes less than six seconds to scroll from top to
bottom.

Using the vertical scrollbar, from a network 687 MHz G4 running OSX
Excel v.X, it takes over a minute to scroll from top to bottom.

Can anyone explain or help me to rectify this?

The Ledge
London

 

[Bob Greenblatt]

I have an AppleMac network with a G3 Server running OSX, a mix of G3s
(running OS 9.1, 9.2 and OSX) and a PowerPC G4 (867 MHz 640Mb) running
OSX 10.2.6.

We are running Excel v.2001 on the OS9 machines, Excel X for Mac
Service Release 1 on the OSX machines.

Excel runs considerably faster on the OS9 machines than the OSX
machines.

For instance, take an Excel spreadsheet with 7,500 rows by 15 columns
of data located on our server.

Using the vertical scrollbar, from a network 400 MHz G3 running OS9.1
and Excel 2001, it takes less than six seconds to scroll from top to
bottom.

Using the vertical scrollbar, from a network 687 MHz G4 running OSX
Excel v.X, it takes over a minute to scroll from top to bottom.

Can anyone explain or help me to rectify this?

The Ledge
London

I can't explain it, nor do I believe there is a way to rectify it. Excel X
is just plain slower in some operations than Excel 2001. I am confident that
Microsoft is aware of the problem and will have a solution in some future
version.

 

[Dave Ring]

If Ford or GM brought out a new model with 70% greater engine
displacement but ten fold lower top speed, I wonder how well it would
sell? I guess Microsoft feels it's a favor to Mac users to provide them
with anything, no matter how poorly it performs.

Has anyone done a speed comparison in OS X between Excel X and Open
Office/X11?

Dave Ring

 

[Dave Ring]

I haven't noticed slow scrolling as a general problem in OS X, but I'll
take your word that it's a documented problem. If it happens because OS
X is giving time to other programs, then I'd expect it to be sensitive
to how many are running. Maybe John Moorhouse can comment on whether
that's the case in his situation.

I haven't compared Excel X to Excel 2001, because too many posted
comments on the slowness of Excel X have discouraged me from buying it.
I have compared Excel 2001 running in OS X Classic mode versus MacOS
9.2, and it isn't noticeably slower. If the slowdown is due to
preemptive multitasking, it seems to me that should also affect Excel
2001 running in Classic while I simultaneously run other OS X programs.
I haven't noticed such an effect, but I have noticed others posting
that Excel 2001 in Classic is faster than Excel X running native in OS X.

My question about Open Office was simply a question. I've looked
briefly at Open Office (which seems to include a replica of VB), but
haven't timed it against the Office X I don't have. Again, I'll take
your word that Open Office lacks many features present in MS Office.
However, it's not obvious to me why the presence or absence of
AppleScript support, VB, Org Chart, Query, Equation Editor, Clip &
Project Gallery, Graph or templates should have much effect on something
as basic as scrolling speed. The listed features all seem like
peripheral stuff that should not need (or be allowed) to affect core
performance.

Maybe OS X deserves part of the blame for the slowdown, but a ten fold
slowdown in a substantially faster machine is severe. It wouldn't upset
me in a beta version, but I don't want to pay full price for it. If I
seem prone to snipe at Microsoft, it's because of past experience in
buying Mac versions of MS programs and then finding out that they are
missing key features or performance. The hard feelings arise because I
don't find these things out until after I've put my money down.

Dave

 

[The Ledge]

Thanks to you all for your responses to my inquiry.
I have run my speed test over our network, i.e from workstations
running local based Excel programs and data on OSX Server. Also From
local workstations with data on the local workstation. The results are
the same. In both cases, no other applications being run.

When running macros in OS9 Excel 2001, screens just fly by. In OSX
Excel v.X they just plod by at a fraction of the speed.

It is beginning to look as if I shall have to revert to OS9 Excel 2001
which as an Apple user of some 25 years, will seriously dent my
confidence in Apple and Apple based products.

In the meantime if I make any progress or solve the problem, I will be
back in touch.

Once again thanks to you all.

The Ledge

 

[J.E. McGimpsey]

Thanks to you all for your responses to my inquiry.
I have run my speed test over our network, i.e from workstations
running local based Excel programs and data on OSX Server. Also From
local workstations with data on the local workstation. The results are
the same. In both cases, no other applications being run.

When running macros in OS9 Excel 2001, screens just fly by. In OSX
Excel v.X they just plod by at a fraction of the speed.

It is beginning to look as if I shall have to revert to OS9 Excel 2001
which as an Apple user of some 25 years, will seriously dent my
confidence in Apple and Apple based products.

In the meantime if I make any progress or solve the problem, I will be
back in touch.

Once again thanks to you all.

One more check you may want to do before you switch.

I don't know what your speed test entailed, but from the description
"screens just fly by", it sounds like you're doing a lot of
selections/activations.

I don't know, but I strongly suspect that screen refresh is slower
in XLv.X. There are two primary ways to speed things up - one treats
the symptoms, the other is a real improvement (note: these
techniques will speed up XL01, too - but for me, they seem to
provide a bigger benefit for XLv.X).

First, to treat the symptoms, always start any code that will be
making selections/activations with

Application.ScreenUpdating = False

It does what it sounds like - keeps XL from updating the screen
every time a selection is made. While the ScreenUpdating property
will be automatically reset to True when your code ends, I like to
include an explicit Application.ScreenUpdating = True at the end of
my main routine.

If you use that technique with subroutines, you should save the
Screenupdating property and restore it at the end, so that you don't
inadvertently turn it on when exiting the subroutine:

Public Sub MySubroutine()
Dim oldScrnUpdate As Integer
With Application
oldScrnUpdate = .ScreenUpdating
.ScreenUpdating = False
End With
<your code here>
Application.ScreenUpdating = oldScrnUpdate
End Sub

The second technique can make things up to an order of magnitude
faster. You almost *never* have to select a cell, range, worksheet
or workbook in order to use it. Instead of

Worksheets("Sheet1").Select
Range("A1:J10").Select
Selection.Copy
Worksheets("Sheet2").Select
Range("B20").Select
ActiveSheet.Paste

use the one line:

Worksheets("Sheet1").Range("A1:J10").Copy _
Destination:=Worksheets("Sheet2").Range("B20")

In addition to avoiding all the overheads associated with
Selections, it avoids cycling the data through the Clipboard. As an
added benefit, since no changes in Selection were made, you don't
have to "return" to the selected cell after a macro.

 

[Dave Ring]

What happens if you run Excel 2001 in OSX Classic mode? Is that as fast
as Excel 2001 / OS9 (in which case I'd blame Excel X), or as slow as
Excel X / OSX (in which case I'd blame OSX).

Dave Ring

 

[Bob Greenblatt]

What happens if you run Excel 2001 in OSX Classic mode? Is that as fast
as Excel 2001 / OS9 (in which case I'd blame Excel X), or as slow as
Excel X / OSX (in which case I'd blame OSX).

Dave Ring

In my experience, Excel 2001 in classic mode is faster than Excel X.

 

[Dave Ring]

How does Excel 2001 in classic mode compare in speed to Excel 2001 in
MacOS 9.2.2? If it is equally fast or faster, then it's hard for me to
blame the slow speed of Excel X on OS X.

Dave Ring

 

[Dave Ring]

I went ahead and did a head to head comparison of Excel 2001 running in
MacOS 10.2.6 Classic mode versus Excel 2001 running in MacOS 9.2.2 on
the same PowerBook. I timed six different tasks chosen to cover various
kinds of calculations, display updates and I/O. The results:

Task OS 10.2.6 OS 9.2.2

Scroll to bottom of sheet 42 50
Find unique words in Huck Finn 21 20
Save worksheets & modules 21 17
Rescore sequence alignments 66 59
MergeSort 65336 strings 36 32
Repaint structure/activity map 20 19

Total 206 197

Overall, there was only 3-4% difference between running in OS X Classic
mode and OS9.

My conclusion -- I don't think OS X deserves the blame for the sluggish
performance of Excel X. If OS X can run Excel 2001 as fast as OS 9, I
think the ball is in Microsoft's court to explain why Excel X falls down
so badly.

Dave Ring

 

[J.E. McGimpsey]

Overall, there was only 3-4% difference between running in OS X Classic
mode and OS9.

My conclusion -- I don't think OS X deserves the blame for the sluggish
performance of Excel X. If OS X can run Excel 2001 as fast as OS 9, I
think the ball is in Microsoft's court to explain why Excel X falls down
so badly.

well, I don't know whether it would make much of a difference, but
there's a couple other factors involved here:

1) The reason that the overall results are so close is that you're
comparing two different types of activities - screen activity where
Classic has the edge vs. calculation-intense processes, where OS9
has a >10% edge. This would be typical of OS X/Classic behavior -
higher priority activity such as a screen refresh done promptly,
interruptable calculations somewhat at a lower priority.

2) Office X and Classic undoubtedly have different nice settings,
meaning that they get different amounts of processor time slices. If
I've reniced XLv.X to -10 (as root) it's pretty snappy. If i renice
it to 10, it's noticeably more sluggish. I don't know what Classic's
default nice setting is (and in certain configurations of my machine
I've renice it and TruBlueEnvironment to 15, in any case, so that
when it's idle it doesn't interrupt OSX much more than background
activity), but it may be grabbing more of the processor than XL can.

As a good multitasking citizen, Office X should have a default nice
value no higher priority than 0, and any background processes should
be significantly lower priority, say 15-20.

 

[Dave Ring]

Per John Moorhouse's reply on 7/14, his slow scrolling problem in Excel
X occurs with no other programs running. If the problem is nevertheless
caused by multitasking taking a lot of time away from Excel X, then
renicing should solve it.

But I doubt it. It's a lot easier for me to believe that Microsoft
rushed out another Macintosh version with minimal development effort,
has more interest in selling it than fixing it, and has no interest in
ever seeing it run as well as the Windows version.

If it's truly hard to make programs run as fast in OS X as OS 9, why
don't I notice slow-downs in the OS X versions of other programs?

Dave Ring

 

[Jim Gordon]

----------

The observation that Excel 2001 runs about as fast in the Classic
emulator in OS X as it does in MacOS 9.2.2 suggests that it is not some
fundamental feature of OS X (like the shift to preemptive multitasking)
that causes the slowdown in Excel X. If it were so, why would it not
also slow down the Excel 2001 / Classic combo?

I don't know why classic and OSX can both talk to the processor
simultaneously with classic taking a barely perceptible hit in speed. It
makes me wonder if Microsoft Windows could also be run natively like that on
a Mac.

That leaves two possibilities -- either the native routines provided by
OS X are a lot slower than their emulated counterparts in Classic, or
Excel X is not efficiently using OS X. The latter seems more likely to
me. My experience with a number of other programs is that the OS X
versions are not noticeably slower than the OS 9 versions, and if Apple
could make their emulator run rings around native OS X, I think they
would have applied that information to speed up OS X.

In regard to Excel X, as I said, some things are not slower at all and every
once in a while some things are actually faster in X than in 2001. But many
things are slower. Most applications that I run seem much slower in OSX, for
example, Norton Utilities goes at less than half the speed in OSX than it
did in OS9. Norton isn't doing much different except running on OSX. I sure
wish Apple would apply what they know and speed up OSX. Maybe Panther will
be better, but it didn't seem like it to me in the demos I've seen so far.

I'd love to do the comparison you suggest, but I don't have Excel X. I
downloaded the test drive at one point, but lost interest when I found
out it didn't include the VBA editor. I'm not impressed by any major
new features in Excel X, and I see no reason to buy it until it's faster
than Excel 2001 in Classic.

I do lots of stuff with graphs, so that alone makes Excel X worth it to me.
It's impossible to make a completely fair benchmark, but some day I might
have time to come up with something that's reasonable. What really puzzles
me is that Excel XP running in virtual PC is sometimes faster than Excel X.
I wish I had some answers for you, Dave, but all I have is more stuff to
ponder.

-Jim