Optimum page file size for 1 GB?

[Jim Garrison]

I just upgraded my Athlon 1800 512 MB to 1 GB. Is there any general
consensus on the 'best' setting I should use for page file please? I
recall a few years ago much debate/controversy over this, but wonder
if a consensus has now emerged? My CPU is now slow by today's
standards (runs at 1533 MHz), so I naturally want to get the most out
of this extra RAM.

My recommendation based on experience: Upgrade to 2GB and disable
paging altogether (set pagefile size to zero). Windows is aggressive
about paging out storage that hasn't been 'recently' touched, when
it doesn't really need to. I did this about 2 months ago and have
seen significant improvements, especially when switching applications.
With a paging file, if you start up, say, MSWord and Firefox, and then
use one of these exclusively for half an hour, the other one will be
mostly swapped to disk. When you switch back, there's a long delay
while a flurry of page faults occurs and the app is loaded back in.
If you can spring for the extra GB you can dispense with paging
altogether. You will probably see even more significant relative
improvement on a slower CPU.

Note that none of this applies to Linux. Linux in contrast is
frugal with paging and swaps out only when it really needs to,
i.e. when the total working-set exceeds real memory size. Adding
memory to a Linux system that isn't already swapping won't
improve performance much.

 

[Eric Gisin]

My recommendation based on experience: Upgrade to 2GB and disable
paging altogether (set pagefile size to zero). Windows is aggressive
about paging out storage that hasn't been 'recently' touched, when
it doesn't really need to.

Absolute bullshit! The pages still reside in RAM.

I did this about 2 months ago and have
seen significant improvements, especially when switching applications.
With a paging file, if you start up, say, MSWord and Firefox, and then
use one of these exclusively for half an hour, the other one will be
mostly swapped to disk. When you switch back, there's a long delay
while a flurry of page faults occurs and the app is loaded back in.

That's exactly what a paging OS is supposed to do, moron.

If you can spring for the extra GB you can dispense with paging
altogether. You will probably see even more significant relative
improvement on a slower CPU.

Ignore all pagefile tweaking advice. Leave it alone if you have a single disk.

 

[Arno Wagner]

My recommendation based on experience: Upgrade to 2GB and disable
paging altogether (set pagefile size to zero). Windows is aggressive
about paging out storage that hasn't been 'recently' touched, when
it doesn't really need to. I did this about 2 months ago and have
seen significant improvements, especially when switching applications.
With a paging file, if you start up, say, MSWord and Firefox, and then
use one of these exclusively for half an hour, the other one will be
mostly swapped to disk. When you switch back, there's a long delay
while a flurry of page faults occurs and the app is loaded back in.
If you can spring for the extra GB you can dispense with paging
altogether. You will probably see even more significant relative
improvement on a slower CPU.

Interesting! I will try this.

Note that none of this applies to Linux. Linux in contrast is
frugal with paging and swaps out only when it really needs to,
i.e. when the total working-set exceeds real memory size. Adding
memory to a Linux system that isn't already swapping won't
improve performance much.

This is not entirely true. If Linux thinks it needs more buffer/cache
for disk operations, it might start to swap out some memory.
I think that the default ''swappiness'' parameter today is set in
a way that this typically does not happen. But this behaviour can be
changed if desired, e.g. if you have some processes that allocate
a lot of memory and only very seldomly need it.

Still, you can disable swap on Linux and it works pretty well, at
least with 2GB. The reasons for this are not performance increases,
but to avoid confidential styff ending up in swap. A different
approach to this is using swap encrypted with a random password chosen
at boot-up.

Arno