Dell Mini 9 Discussion Discussion on the Dell Inspiron Mini 9.

Thread Tools Display Modes
  (#21) Old
jefito jefito is offline
Posts: 93
Join Date: Oct 2008
Default Re: I just experienced a significant speed increase! - 12-30-2008, 10:48 PM

Originally Posted by glenrich
Selecting no pagefile at all can make things faster, though -- you're not swapping memory pages from fast RAM into slow SSD and vice-versa. That seems clear to me.
This is yet another misconception about the memory management system.
It is understandable since it is a man-made construct and it is not completely intuitive. I was fooled by this as well for many years before I finally learned enough to understand the true behavior. There is a video by David Solomon regarding the memory manager that contains the best and easiest to understand explanation of this that I have heard to date.

You should check it out:
That's a fantastic video -- thanks for posting it. I've been able to work my way through about half of it during work today.

The main performance hit from paging is when you have to read data pages back in from the page file.
Well, writing them out in the first place isn't very fast either...

Now when data is written to the page file, it is not simply written out to the page file and left out there. Actually, a copy of the data in the page file is cached in the memory manager standby list so that if the application using the data tries to access it, it reads in straight from the memory via the standby list rather than having to go to the page file to get it.
Sure, that's fine. It's also cached in the modified page list -- this list is for modified pages that have not been written to disk yet, right? The standby list is for modified pages that have been written to the page file. Both of these lists service soft page faults.

So if you use so much memory that the cached copy has to be overwritten to make room for more memory hogs then you might see some paging in from the page file and slower performance of apps waiting for it.

The important point here is that by not having a page file you remove the ability to swap these pages between RAM and your disk but on a system that does not have enough memory pressure to require a pagefile, you are not swapping pages in from the pagefile anyway so the ability to use one does not effect performance.
This is the bit that I don't fully understand yet -- maybe it's in the second half of the video that I haven't seen. Since pages that wind up being actually paged from disk must reside first on the modified pages list, what is the criterion for actually writing those modified pages out? I believe that he said that the size of this list is small (approximately 2MB?), so it could be exceeding that limit regardless of how whether they actually need to be written out (if there is actually a page file there to receive them). I'll watch further and see if I can understand that better.

By not having a pagefile, you are not allowing pages to be written to the disk.
Hard to argue with that.

You should understand that the memory manager pages things to the disk for a reason. It realizes that the memory has not been in use for a long period of time or there is an app that is using way too much memory and needs to be trimmed down. These pages that are trimmed are the ones that are candidates for being paged to the disk.
[nitpick]That's not strictly true, is it? Only *modified* pages are paged out -- unmodified pages (like those containing executable code) are just discarded, since they can be read back in from their original source. But still, the general idea is right.[/nitpick]

If you do not have a pagefile at all, then these un-used items cannot be set aside to make room for other apps that are currently active. By taking up memory with apps that are not really using it, you are potentially preventing apps that can really use it from having access. This can slow down other applications.
Not sure I wholly agree with this -- if there's enough memory to go around, nobody gets hurt -- the modified page list maybe grow larger than its usual size, or maybe those pages just stay in their respective working sets. I guess those working sets might need to be smaller because of the amount of modified pages, that might cause more some thrashing as the OS tries to keep working set pages around. I'll think about this as I watch the video some more.

Of course, if all of the apps you are running at the momemt are not taxing your memory at all and you have memory to spare, then you will not notice any difference by not having a pagefile.
I agree with this, but it's contrary to what some folks here are saying, and possibly experiencing.

thanks again for a great link,

Reply With Quote
  (#22) Old
3c3 3c3 is offline
Junior Member
Posts: 3
Join Date: Dec 2008
Default Re: I just experienced a significant speed increase! - 12-31-2008, 12:20 AM

Virtual memory is nothing new and has been around for many tens of years. If you have sufficient RAM to run your programs, there is no need to have a pagefile. If you don't have enough RAM, that's another matter. Turning pagefile off when you don't need it avoids useless writes to the SSD, which has limited write cycles. System, user, and Internet temporary files should also be directed to a RAM disk if you have enough RAM.
Reply With Quote

« Trick xp into recognizing more than 1gb of ram | Can I delete old software updates? »
Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Copyright © 2008-2016