Better performance: Vista w/4gb RAM or XP w/3GB RAM

[jg70124]

I use Excel for operations and financial modeling; my spreadsheets are
large and have many named ranges, lookups and array functions. I'm
currently using a Pentium M 1.2 machine with XP and 2GB of RAM, and
it's barely adequate for the work.

So I'm getting ready to buy a Lenovo with a Core 2 Duo T7500, which
will be a big improvement. I'm also planning on switching from XL 2003
to XL 2007 (for bigger worksheets). But would I be better off with XP
(which maxes out at 3GB of RAM) or Vista (which can go to 4GB)?

Anyone here know?

Thanks

 

[Don Guillett]

I have an HP with a 6300 and Vista Home Premium & 2GB and it works well for
me. This particular machine makes it very simple to add 2 more as it has 2
open slots. So, it's just plug em in. No need so far. I am still using
xl2003 but will load 2007 soon. I suggest keeping both versions in separate
folders.

I have problems running the xl95 & xl97 on the Vista machine even tho I
installed in order.
Although I'm not quite sure how to do this, I may partition my 320 hd into
dual boot xp & vista, putting the older versions on the xp partition.

 

[Bob I]

I use Excel for operations and financial modeling; my spreadsheets are
large and have many named ranges, lookups and array functions. I'm
currently using a Pentium M 1.2 machine with XP and 2GB of RAM, and
it's barely adequate for the work.

So I'm getting ready to buy a Lenovo with a Core 2 Duo T7500, which
will be a big improvement. I'm also planning on switching from XL 2003
to XL 2007 (for bigger worksheets). But would I be better off with XP
(which maxes out at 3GB of RAM) or Vista (which can go to 4GB)?

Anyone here know?

Thanks

Vista AND XP are both limited to 4 GB RAM.

 

[Roger Govier]

Hi Don

I asked a question about whether there were any problems dual booting XP and
Vista in the UK MVP newsgroup recently, and Mike Maltby posted this
response .....

I've set up several systems to dual boot Vista and XP including a new Sony
Vaio SZ4MN/B laptop that came with Vista Business pre-installed. In my
case I use BING (www.bootitng.com) as the boot manager thus ensuring that
I don't have any part of another OS installed on another OS's partition.

Due to XP's system restore conflicting with shadow copy in Vista it is
essential to hide each OS partition from each other sharing only data
partitions and to have system restore only enabled on the system drives.
I go even further and using BING remove the second OS partition info from
the partition table when booting the first OS (and vice versa).

and in a follow up posting .....
BING is far more than just a boot manager and is an excellent partition
manager including partition imaging and highly recommended by many of the
Windows desktop MVPs

By the way, congratulations on the MVP award.

 

[Guest]

I assume you do not use other non-MS Office apps.

My company uses a Proj Mgmt pkg that runs through IE 6 or 7. However, while
it runs with either version of IE, it will not run on Vista due to the fact
that it needs MDAC. MS now uses "WDAC" in Vista.

 

[Guest]

Check out these links on performance. Excel can only use so much RAM...

http://msdn2.microsoft.com/en-us/library/aa730921.aspx
http://www.decisionmodels.com/memlimits.htm

Decision Models also has some interesting discussions on optimising speed
and calculations...

My 2 Cents. If you are running your spreadsheets to the max then upgrading
will only get you marginal improvements. I can easily grind XL to a stand
still with 65,536 rows. Adding more rows will just make things worse. Look at
Databases and Cubes (Pivot Tables) for substantial performance increases. I
regularily attach XL pivot tables to Access and DB2 queries of a million plus
records with very acceptable performance (never more than a couple of seconds
wait).

 

[jg70124]

Vista AND XP are both limited to 4 GB RAM.

XP can have 4gb, but it can only address 3gb - the addresses used by
the last 1gb are reserved by the OS for addressing ACPI drivers. I've
heard that 64 bit XP can address all 4gb, but I can't seem to find
anyplace to buy it anymore, and I read on the net that many XP drivers
won't work in the 64 bit version.

 

[jg70124]

My 2 Cents. If you are running your spreadsheets to the max then upgrading
will only get you marginal improvements. I can easily grind XL to a stand
still with 65,536 rows. Adding more rows will just make things worse. Look at
Databases and Cubes (Pivot Tables) for substantial performance increases. I
regularily attach XL pivot tables to Access and DB2 queries of a million plus
records with very acceptable performance (never more than a couple of seconds
wait).

Yes, I'm trying to learn Access for the big jobs. But the end product
of each analysis is a series of ~50-100 bar and scatter charts
presented in Powerpoint, so it means I have to go from Access to XL to
PPT. Plus I'm not very good with Access, so it adds significant
overhead to my personal workload.

What do you mean by "cubes"?

Thanks.

 

[Bob I]

XP can have 4gb, but it can only address 3gb - the addresses used by
the last 1gb are reserved by the OS for addressing ACPI drivers. I've
heard that 64 bit XP can address all 4gb, but I can't seem to find
anyplace to buy it anymore, and I read on the net that many XP drivers
won't work in the 64 bit version.

Yes 64 bit OSes CAN address much more than 4 GB ram. The amount of RAM
available in that last gig of RAM in 32 bit OSes is completely system
dependant.

 

[Harlan Grove]

(e-mail address removed) wrote...

I use Excel for operations and financial modeling; my spreadsheets
are large and have many named ranges, lookups and array functions.
I'm currently using a Pentium M 1.2 machine with XP and 2GB of RAM,
and it's barely adequate for the work.

....

No matter how much usable RAM you have, if your spreadsheets are
larger than 100MB, you're begging and pleading for trouble. If your
spreadsheets are smaller than that, going from 2GB RAM to 3 or 4GB RAM
wouldn't likely lead to any performance improvement.

I'd bet there are many optimizations you could use that would both
reduce file size and speed up your calculations.

And there are better tools for massive calculations than Excel. I need
to perform periodic (roughly once a quarter) large-scale simulation
problems. I use R, which is the GNU version of the S language. I can
perform billions of calculations, including thousands of conditional
sums on datasets of tens of thousands of entries in seconds. The disk
output is the real drag. It can take up to a minute. But I can run
this along with a browser, e-mail and even Excel also running at the
same time.

You're not using the right tools. Even a titanium hammer with a tooled
leather grip won't drive screws as well as a $1.99 screwdriver.

 

[jg70124]

(e-mail address removed) wrote...

You're not using the right tools. Even a titanium hammer with a tooled
leather grip won't drive screws as well as a $1.99 screwdriver.

So, what would you suggest.

I run my analyses on invoice data I get from clients, mostly small and
mid-size manufacturers and distributors. Most do *not* have modern
ERP systems; they typically use old home-brewed DOS or SQL based
systems. The most sophisticated ones might have a 10 year old ERP
system with limited reporting capabilities. So the best I can hope
for, usually, is to get CSV files with raw invoice data (one row for
every line item on every invoice), and for costs (one row for every
SKU). My job is to calculate margins by SKU, customer, SKU-customer,
product type, and so on.

My typical process is:
1) Use an Excel pivot table to get sales data by SKU and customer from
the raw invoice data
2) Join that to the cost data with an Access query, and export the
result into a new Excel file.
3) Use pivot tables against the new Excel file to get the margin
information I need.

This is iterative; many times during the analysis my team will decide
we need to change the segmentation, which might require that we start
again from step 1 or step 2.

Also, note that then end product *must* be a simple Excel spreadsheet
that the client can use to review and change prices for every SKU,
based on the outcome of our analyses. (It has to be very simple
because most do not even know how to use pivot tables, and they almost
never have access to business or IT analysts to help them).

I'm open to any suggestions you might have for better methods.