Proper XP Virtual Memory Settings for Two Physical Hard Drives

[TrickyDick]

I have SP/SP3, with 4 GB of RAM, and two separate physical drives. 
C-Drive is NTFS, 75 GB total, 19 GB free.
D-Drive is NTFS, 233 GB, 146 GB free.

I have received conflicting advise about my virtual memory settings.
Can anyone please suggest appropriate Paging File Sizes, or whether
or not to use "System managed size".   Per one suggestion, current
settings are:

C-drive = Initial Size = 100 MB
C-drive = Max  Size =  500 MB
D-drive = Initial Size = 500 MB
D-drive = Initial Size = 6000 MB

Thanks!   

rick

[gregflowers]

From:
http://support.microsoft.com/kb/889654


When you set up a 32-bit version or a 64-bit version of Windows XP, a page file is created that is one and a half times the amount of RAM that is installed in the computer provided there is sufficient free space on the system hard disk. However, as more RAM is added to a computer, the need for a page file decreases. If you have enough RAM installed in your computer, you may not require a page file at all, unless one is required by a specific application.


There is no specific recommendation for page file size. Your requirements will be based on the hardware and software that you use and the load that you put on the computer. To monitor page file usage and requirements, run System Monitor, and gather a log during typical usage conditions.

[Allan]

Trust me, with the speed of today's hd's and the amount of installed ram it just doesn't matter. On my systems with multiple drives I put the pagefile on the "second" drive (the one without the OS) and leave it as "system managed". In fact, for no particular reason at all I always give the pagefile its own partition.

I don't put a pagefile on the system partition because I disable crash dumps.

[BC_Programmer]

I don't put a pagefile on the system partition because I disable crash dumps.

But however can you use Windbg to investigate mysterious bluescreen errors?   

[Allan]

But however can you use Windbg to investigate mysterious bluescreen errors?  

I suppose that is a drawback, isn't it? 

[BC_Programmer]

there's nothing quite as exhilarating as walking a kernel stack trace, with your nose pushed right up against the screen. It's refreshing, like being smacked in the face with a cast iron frying pan- your nose never quite feels the same. In fact, kick things up a notch and disable symbols!   

[Geek-9pm]

there's nothing quite as exhilarating as walking a kernel stack trace, with your nose pushed right up against the screen. It's refreshing, like being smacked in the face with a cast iron frying pan- your nose never quite feels the same. In fact, kick things up a notch and disable symbols!   

When at a computer fair,
I look for the people with flat noses. 

[Salmon Trout]

If you have enough RAM installed in your computer, you may not require a page file at all, unless one is required by a specific application.

Xp can get very weird indeed without a page file. I always have at least a very small page file. I would suggest a minimum of 512 MB.

[patio]

Why any other drive other than the one that contains the OS would need a pagefile is beyond me...

Perhaps i'm missing something though.

[Allan]

Xp can get very weird indeed without a page file. I always have at least a very small page file. I would suggest a minimum of 512 MB.

Agreed. Not only that, but you cannot disable the pagefile in XP, only your ability to manage it. XP was designed by MS developers to page. If you set the pagefile to "disabled", XP will page where and when it wants to and you'll never know it.

[Allan]

Why any other drive other than the one that contains the OS would need a pagefile is beyond me...

Perhaps i'm missing something though.

It's a bit more efficient if the system pages to a second drive - especially if that drive is faster than the one containing the OS. But in the end, any difference is negligible.

[patio]

I've yet to see proof of this...
And if it's a 2nd physical drive the constraints would be on data transfer speeds involving the board itself...

However i have milliseconds to spare for a test run...

[Allan]

Exactly. Look, clearly it's faster to read from one disc and write to another - but again, when it's paging we're discussing the difference isn't worth discussing.

[Salmon Trout]

clearly it's faster to read from one disc and write to another

Can be quicker because of seek time. The difference would be most marked when reading/writing lots of small files. Not so much with one big file.

[BC_Programmer]

it's only faster if that drive is on a separate channel altogether; a different SATA drive, or a separate IDE controller for example.

The reason is simple; The system drive is where all the system dlls and files are loaded from, and program files, and so forth. the theory holds that using a separate drive will, as ST said, reduce seek times, as the pagefile is accessed sequentially, rather then, for example, accessing a bit of the PF, seeking over to say, comctl32.dll to load a resource for some program, etc; basically, moving the read/write heads around a lot, constantly back to the pagefile.

having partitions just for the pagefile are helpful solely because they prevent the pagefile from getting fragmented, which is unlikely in the worst scenarion anyway.


The only time one would get a good benefit is if the other hard drive is a LOT faster then the system drive- I encountered this scenario with my old PC, whereby my C: drive was both too slow and too small to hold the pagefile (A mere 2GB) I was therefore forced to move the pagefile to my second hard drive, an 8GB drive that was far newer and faster (still old though). originally, this actually caused my PC to freeze every boot. I had to do a chkdsk, because the pagefile had been places where a sector had gone bad (I was poor and had to work with old hardware ). After that things worked fine for me, and it was far faster then it was when I had the pagefile on my system drive, likely because the system drive was old and slow.