Is this true? Defragging with virtual memory off is better/enhances performance?

So I hear this rumor from this kid from E.B. (electronics boutique) that if you disable virtual memory in windoze before you defrag, this is actually going to improve performance in gaming, benchmarks, etc....

What, if I may be so naive, is this statement based on? Is there any evidence that a defragmented swap file is better than a fragmented one? Also, this has to be RAM dependent, right? That is all just a load of B.S. , isn't it.

Also, this kid actually said to disable V.M., defrag, install the game, defrag again, re-enable V.M. in order to get the best performance...

Is this insane, or am I?

I don't think you are insane, but that is not to say that this won't work. The kid you talked to has probably tried it and can attest that it does. That doesn't make it the best practice by any means.

If you are ever concerned about a fragmented swapfile, in 9x you can boot to DOS and delete the c:\windows\win386.swp file, then reboot, and perform the defrag with a newly created (by the OS) swapfile. In 2k under group policy there is an option to "clear virtual memory pagefile when system shuts down". If you enable that, then reboot, you would get a new c:\pagefile.sys file, which should be contiguous. I might have to take a minute to figure NT4 out, but it probably could be done from system properties, setting a 0-0 for pagefile, reboot, then create one again.

I am sure this topic will generate a very interesting discussion. I look forward to reading on...

I failed to mention, yes, a fragmented swapfile would be (in theory) slower than a file with minimal fragmentation. Honestly, I don’t understand how installing any app with VM disabled can make the game run faster as soon as you turn VM back on. It sounds like the kid is on to something, but he doesn’t know what he is actually doing.

It should work considering you had something over 300mb or RAM or something. Turning the swapfile off when using 64mb or less is just dumb.

The speed increase from a contiguous swapfile will be in the miliseconds and not overly noticable unless the drive was really fragmented from the get go. Also, I think that windows just pulls everything to RAM and puts it where it wants to on the HDD anyway with VM enabled or disabled (I think it is set up to use only RAM during the defreag process and not the HDD at all so it can move the swapfile - I think). Personally I think the kid is full of it though to be sure a defraged drive is better than a fragmented drive. I can't see how there would be a good deal of difference between defraging with the swapfile turned on or off.

I'll give you a better idea.

Set up a partition on a secondary drive, say about 512 MB, and use it specificly for VM.

Or you can go one step farther, like I did and install an old 1GB drive just for VM. Once set I never worry about it again.

Typical games nowadays are big, so all that defraging can't help but speed things up.

I know why he thinks defragging with virtual memory off will help, because when you are defragging what is windows using to do it with? the swapfile. Why do you think it says drive contents changed restarting, the swapfile is resizing, Doh! . If this is 98/me(95 uses physical ram first by default) then you want windows using all of its physical ram first so read this thread :-

<a href="http://forums.windrivers.com/cgi-bin/forum3/ultimatebb.cgi?ubb=get_topic&f=17&t=007239&p=" target="_blank">http://forums.windrivers.com/cgi-bin/forum3/ultimatebb.cgi?ubb=get_topic&f=17&t=007239&p=</a>

Ever thought how often windows resizes that swap file, milliseconds or not it adds up. Given the price of memory just fill it up and then turn virtual memory off, what's the worst you are gonna get? out of memory errors?

Can you get windows 9x to display how many bits the swap file is in? You can in NT. If the page file is in say 10 pieces then you are going to have 10 x 10 milliseconds seek (best guess) every time you read the swapfile, admittedly only one hundreth of a second, but on every disk access? If you bought a drive with a seek time like that you'd be really pissed.

I know we are going to potentially run into issues regarding the size of any AGP gart and max cache settings and so on, but you can fix most of these in system ini once you know the exact system specs.

[quote]Originally posted by *MAYHEM*:
<strong>I'll give you a better idea.

Set up a partition on a secondary drive, say about 512 MB, and use it specificly for VM.

Or you can go one step farther, like I did and install an old 1GB drive just for VM. Once set I never worry about it again.</strong><hr></blockquote>

Only problem with that is older harddisks are generally slower than newer hardisks. One of my favorite methods s a separate 512Mb FAT (not FAT32, this is slower) partition.

[quote]Originally posted by Stanley_Kubrick:
[QB]

Is there any evidence that a defragmented swap file is better than a fragmented one?
[QB]<hr></blockquote>

I know with my WinNT4.0 machine at work, when my swap file (500 MB) got fragmentated, it usually meant the next time I rebooted, I would see a BSOD....thanks to my IS dept. :mad:

Originally posted by confus-ed

I know why he thinks defragging with virtual memory off will help, because when you are defragging what is windows using to do it with? the swapfile.
---------------------

I'll disagree to some extent, if you will take the time to move your swapfile to any other drive/partition and defrag c: in expanded veiw, you can pause the process, move to the tail end of the partition and then resume defragging, you will see data being read and written with extreme activity. I believe that defrag uses the tail end of the partition to store clusters it wants to arrange in a more ordered fashion just as soon as the opportunity presents itself. IMO this also means that defrag keeps a running log file of some sort logging its own progress so that in the event of a power failure (or outright quiting) it can resume the job or at least keep files/clusters valid even if they are now located at the end of the partition.

Also, if you will defrag C:, with paging on C: drive, you will see a honker block of read only clusters that always dissapears from C: when you move the swap file to another partition. The huge read only blocks ARE the swapfile. They are consistantly right in the middle of things and almost never contiguous. The swapfile as well as the partition itself is then by definition, fragmented. But defrag won't move the read only clusters of the swapfile on C: - SO - I always move my swap file to another partition, then defrag, then move it back if I want.

Just perhaps that's what the misguided gammer is actually trying to achive in the first place? A defraged swap file AND partition in the least amount of moves? It would even work - IF he had enough ram such that a swap file was not used in the first place. But if that was true then he wouldn't need to do anything other than defrag in the first place. Kind of cart before the horse.

PS - defrag.exe from Windows ME is about 20 to 50 times faster than win9x versions and will even run under 95 and my win98se, the rumor IS true! I've been using ME defrag on 98se machine for about a month now and no problems so far.