Turn off page file in Windows XP for DAWs?

[Guest]

Archived from groups: microsoft.public.windowsxp.perform_maintain (More info?)

There is a persistent recommendation on the net to turn of the page file when
optimizing XP for work as a DAW (Digital Audio Workstation, or computer used
for multi-track music recording). This goes against everything I know about
how paging works, but I wanted to know if anyone had more in-depth knowledge
about the subject.

I suspect the rumor persists because people think that turning off the page
file will prevent page faults. I know this is not the case. But is there
any way that turning off the page file can help improve performance?

Keep in mind that this is for a computer used for a very specific
application. The DAW software (something like Cakewalk's SONAR) will be
running, along with all companion programs (effects, synthesizers, etc).
During a recording session, the CPU and disc drive can get constant, heavy
use, as does the port communicating with the audio interface (PCI bus, or
integrated Firewire or USB).

Obviously, if there is only a single hard drive, paging can cause problems.
Since disc usage can be extremely high (even the limiting factor), page
faults can be a killer. This can be easily solved by adding a new physical
drive dedicated for the audio data being recorded. Then page faults do not
pull the disc head away from the audio data, and the problem goes away.

But is there any other issue? Say the computer has a lot of RAM, way more
than ever gets used, so the computer never runs low on physical RAM. Will
turning off the page file in this instance cause any increase in performance,
either in the CPU, or in disc access? I've run into references that say that
Windows will copy ALL RAM to the pagefile (see
http://support.microsoft.com/default.aspx?scid=kb%3Ben-us%3B99768). What's
the deal with this? Can this cause a problem when there is a constant, heavy
stream of data being channeled from Audio Interface to CPU to disc?

Finally (if that isn't enough, what happens when the computer runs out of
ram and the pagefile is disabled? Will the computer swap out executables to
get more space for data, or will it just stop working?

 

[Guest]

Archived from groups: microsoft.public.windowsxp.perform_maintain (More info?)

My comments are interspersed.

"Sinc" <Sinc@discussions.microsoft.com> wrote:

>There is a persistent recommendation on the net to turn of the page file when
>optimizing XP for work as a DAW (Digital Audio Workstation, or computer used
>for multi-track music recording). This goes against everything I know about
>how paging works, but I wanted to know if anyone had more in-depth knowledge
>about the subject.
>
>I suspect the rumor persists because people think that turning off the page
>file will prevent page faults. I know this is not the case. But is there
>any way that turning off the page file can help improve performance?

No. In fact the opposite is true. Turning off the page file will
very often result in poorer performance. This is because Windows XP
uses address space in the page file to satisfy the memory requirements
of the *unused* portions of memory allocation requests. By design
Windows must identify memory addresses for all memory requests that
are issued by Windows components, device drivers, and application
programs. Normally Windows will only allocate RAM to those portions
of the requests that are actually used and will map the unused
portions to locations in the page file.

In the absence of a page file then all requested memory must be
allocated to RAM. For example, on my own system with 768 mb of RAM
right now Windows Task Manager is reporting 450 mb of PF usage.
However a check with another utility shows that there is only 124 mb
of active memory content in the page file. So that means that there
is 450 - 124 = 326 mb of memory that has been requested but never
used.

>
>Keep in mind that this is for a computer used for a very specific
>application. The DAW software (something like Cakewalk's SONAR) will be
>running, along with all companion programs (effects, synthesizers, etc).
>During a recording session, the CPU and disc drive can get constant, heavy
>use, as does the port communicating with the audio interface (PCI bus, or
>integrated Firewire or USB).
>
>Obviously, if there is only a single hard drive, paging can cause problems.
>Since disc usage can be extremely high (even the limiting factor), page
>faults can be a killer. This can be easily solved by adding a new physical
>drive dedicated for the audio data being recorded. Then page faults do not
>pull the disc head away from the audio data, and the problem goes away.
>
>But is there any other issue? Say the computer has a lot of RAM, way more
>than ever gets used, so the computer never runs low on physical RAM. Will
>turning off the page file in this instance cause any increase in performance,
>either in the CPU, or in disc access? I've run into references that say that
>Windows will copy ALL RAM to the pagefile (see
>http://support.microsoft.com/default.aspx?scid=kb%3Ben-us%3B99768). What's
>the deal with this? Can this cause a problem when there is a constant, heavy
>stream of data being channeled from Audio Interface to CPU to disc?

If you look at that article you will see that it applies to, and only
applies to, Microsoft Windows NT Advanced Server 3.1 and Microsoft
Windows NT Workstation 3.1. It has been at least 5 years since I
last encountered anyone still using those long-outdated Windows
versions.


>
>Finally (if that isn't enough, what happens when the computer runs out of
>ram and the pagefile is disabled? Will the computer swap out executables to
>get more space for data, or will it just stop working?

Crash city. Bang. Boom. Thud.

Good luck


Ron Martell Duncan B.C. Canada
--
Microsoft MVP
On-Line Help Computer Service
http://onlinehelp.bc.ca

In memory of a dear friend Alex Nichol MVP
http://aumha.org/alex.htm

 

[Guest]

Archived from groups: microsoft.public.windowsxp.perform_maintain (More info?)

Thanks for the reply - it jibes with everything else I've discovered.

I actually did a bunch of further research on the subject, as well as a
bunch of actual experimentation, and the results are here:

http://www.well.com/~sinc/DAW_Paging.html

Let me know what you think. I made some assumptions about the internal
workings of NT that may not be entirely correct.