+1 for the tweak being useful only in (very) limited scenarios. Increasing paged pool memory does increase the size of the kernel paged pool memory pool, which is useful as it stores information about PTEs created for disk access, the registry, sections for memory mapped files, etc. Increasing this value means that the paged pool size is larger, so on a non-busy system it will take longer for these pages to be written out of RAM and to the paging file on disk (just because something is in paged pool doesn't mean it's automatically paged to disk - this only happens if RAM is needed for other apps that paged pool currently consumes).
So, if you have a very underutilized system (check email, browse the web, open *small* documents, etc.) this *might* help you depending on the drivers you have installed, the apps you run and how large the data in those applications are, etc. However, if you game, open large applications or open large numbers of files at any one time, this sort of tweak is actually going to be a *bad* thing in the long term (the busier your system gets, the worse performance will be when paging occurs as there will be more data to page), especially if you're still using 32bit Windows; the default and the potential max size of kernel paged pool on x86 is much smaller (XP, 470MB; 2003, 650MB; Vista/7, sized at 75% of RAM by default up to 2GB, whichever is smaller) than the default size on x64 (XP/2003, 400KB/MB of RAM up to 128GB; Vista/Win7, sized at 75% of RAM by default up to 128GB, whichever is smaller)- another area where switching to x64 will benefit you.
Also, given that on Vista and newer the kernel pools (both paged and nonpaged) are dynamic (on x86 as well as x64), a tweak like this is almost useless at this point simply for that fact - especially in Windows 7, Windows already tries to make the size of the pools the correct size for load, as well as making sure that as much other oft-used data is precached as well. My guess is that folks using XP or 2003, or using x86 Vista or Windows 7 would *maybe* see something useful from this, but that would be a big maybe (and you'd have to show me data to back it up, not just the placebo "it feels faster now!" effect.
