Can a pagefile be too big?

[Shakir_Akbari]

not to beat a dead horse, but there is a tip i'd like to share with ya'll

if you create a separate partition on your drive, and stick the PF over there (set c: to 0 mb, and partition to #mb), it wont get fragmented, and you can forget about the 3rd party apps and whatnot.

also i've heard people say that the PF should be 3x the ram size. this has never made sense to me, and i'm glad to hear some other opposing views on that

 

[place]

if you create a separate partition on your drive, and stick the PF over there (set c: to 0 mb, and partition to #mb), it wont get fragmented, and you can forget about the 3rd party apps and whatnot.

how can you specify where the page file gets written to?

 

[orion456]

how can you specify where the page file gets written to?

my computer/properties/advanced/performance settings/advanced/virtual memory

You can pick your drive and the size of the page file. The best thing to do is set both the min and max to the same size, then windows automatically sets a contiguous region of hard drive space for the swap file. It also stops windows from contantly "managing" the space, making it bigger and smaller. 4096 mb is more than enough for most people.

If you have a gig or more of memory you should also set the the following flag in the registry to 1, so that windows keeps the its kernal in memory instead of swapping part out to disk.

Click on Start => Run => type regedit and enter
- Go to HKEY_LOCAL_MACHINE\ SYSTEM\ CurrentControlSet\ Control\ Session Manager\ Memory Management\ DisablePagingExecutive
- Set the value to be 1
- Reboot the computer

 

[Shakir_Akbari]

If you have a gig or more of memory you should also set the the following flag in the registry to 1, so that windows keeps all its files in memory instead of swapping them out to disk.

i dont think a gig will be sufficient if either of the following is true: you run vista or you use an integrated graphics card

i am currently getting really interested in optimizing my page file, i think that i am going to end up getting a compact flash -> SATA adapter and run my page file off that.

yes my pc might go though a CF card every 2 years but that is ok with me - seek times of .2ms make me happy. there are a few other tweaks to make XP lighter, see the FAQ here:

http://mason.gmu.edu/~sfiorito/eXPinstall.htm

this doc talks about using some of the features in XPe to make XP less memory hungry

 

[BrutalDrew]

f you create a separate partition on your drive, and stick the PF over there (set c: to 0 mb, and partition to #mb), it wont get fragmented, and you can forget about the 3rd party apps and whatnot.

All this would is is increase the average seeking distance, thus hurting performance in the end. For optimal performance the pagefile should be set to the most-used partition and on the last-used drive. If you only have one drive, leave it with applications an the OS, don't seperate it. There's also no point in setting a static size page file. Windows only resizes the pagefile if it needs to, so if the initial is set high enough, fragmentation won't be an issue. I'd definitely rather leave that "safety net" there and allow it to resize when it needs to. Also, with a dynamically sized pagefile, if more is needed, it is added somewhere else on the disk, and upon reboot will go back to its initial non-fragmented self. Thus, the pagefile does not get fragmented anyway.

I suggest just leaving the pagefile system managed (and no, a large pagefile won't hurt you), or if you really want to, monitor your PF usage using the "%usage" counter in perfmon and use that as to determine the optimal size for your pagefile. Just run your most intensive applications for a while and see how much of the PF is actually getting used. Then set the initial size to 4x that number. Then the max should be set to at least 2x the initial size. This insures the initial size is large enough to where it doesn't fragment, but still leaves that "safety net." So you get the best of both worlds.

Not that all this fragmentation talk really matters though. The only way a fragmented pagefile would really effect performance were if it fragmented files around it that are read in a sequential fashion. The pagefile itself is not read in such a fashion. In fact, Windows doesn't write more than 64K per buffer to the pagefile at a time anyway (pages are 4K, but Windows writes them out in "chunks" of 64K at a time). Also, very rarely are they even in 64K chunks. The majority of the time paging to the pagefile is inturputed by paging to many other files on disk. So even if the pagefile is one large contigious chunk, it doesn't even matter. The head is hopping all over the place anyway.