WinXP SP2 Virtual memory-> Let the OS manage it or set it yourself?

[JEDI]

back in the old win98 days, it was best to set the virtual memory to 1.5x system memory in 1 contiguous block. ie: 512megs ram = 768megs virtual

you didnt want the os to dynamically allocate virtual memory because it might be created in multiple blocks, thus a performance hit because the hd seeks for the data.

What's the general consensus for setting WinXP sp2 virtual memory?

and what if my virtual memory is 768megs, but i have 1gig ram? (ie: upgrade memory at a later date.)

 

[stash]

1. That's not virtual memory.
2. Let Windows manage the paging file.

 

[azel]

General consensus is VRAM = 1.5 times the physical memory on a drive which isn't your OS drive.

Having less than 1.5x physical memory does not really speed up paging (except initialization and hibernation) but it does save space.

Windows NT does have pretty good paging algorithms in terms of efficiency so you won't see performance improvements over a system with no VM and with VM if the total memory usage is under the total physical memory capacity.

Setting the pagefile manually to 1.5x and letting system choose something won't really make a difference in performance either.

 

[casper114]

The only thing that I ever do with the page file is move it to another drive on my machine. Although the Microsoft curriculum states that the page file being on a different partition from the system partition can prevent many server problems ex. server with 2 processors, one is being maxed out by applications while the other is 98% system idle process.

 

[Phoenix86]

The virtual memory size should be set based on your memory requirements. This is essently what the OS does when it's managing it. If you have 2 physical drives, move it to the non-OS drive.

The first questions is, what is your memory requirements? Check commit charge peak in task manager after your have used the system a bit (to build a good peak).

If your RAM is significantly over your commit charge, you aren't using your page file much at all, so any tweak to the PF isn't going to yield much.

 

[JEDI]

Originally posted by: STaSh
1. That's not virtual memory.
2. Let Windows manage the paging file.

what do you mean that's not virtual memory?

if windows manages paging file, will it be fragmented into multiple blocks if i have the machine on for long enough? (ie: opening/closing word/excel/internet explorer)

 

[Phoenix86]

Virtual memory is the whole memory subsystem. The page file is one piece of VM. You're confusing the terms like so many do because 1/2 the people writting up published articles don't know the difference either. PF!=VM. What most people reffer to as VM is in fact, the PF. There's another thread asking "can I disable VM" and they're doing the same thing, confusing the terms...

Yes, windows will fragment the PF if you let it manage it, this is why I recommend setting the min=max after you have evaluated your system's requirements. Alternatively you can buy a degragmenter that will touch the PF and let windows manage it.

 

[MrChad]

You're not a n00b.

SEARCH :roll:

 

[Nothinman]

what do you mean that's not virtual memory?

It's just not, if you want the full explanation grab the Inside Windows book and read the chapter on memory management.

if windows manages paging file, will it be fragmented into multiple blocks if i have the machine on for long enough? (ie: opening/closing word/excel/internet explorer)

No, the pagefile is only expanded if you run out of pagefile space not on random whims.

 

[Phoenix86]

Originally posted by: Nothinman

No, the pagefile is only expanded if you run out of pagefile space not on random whims.

True, but it is in it's nature to expand, so fragmentation is possible if not likely over time. If you do use your PF a lot (you don't have enough RAM) a fragmented PF can cause additional slowness.

Really the answer is to get enough RAM to fill your commit charge requirements.

 

[bwnv]

Originally posted by: Phoenix86
Virtual memory is the whole memory subsystem. The page file is one piece of VM. You're confusing the terms like so many do because 1/2 the people writting up published articles don't know the difference either. PF!=VM. What most people reffer to as VM is in fact, the PF. There's another thread asking "can I disable VM" and they're doing the same thing, confusing the terms...

Yes, windows will fragment the PF if you let it manage it, this is why I recommend setting the min=max after you have evaluated your system's requirements. Alternatively you can buy a degragmenter that will touch the PF and let windows manage it.

I would add that 90% of windows users should probably just leave it alone, they'll never see any diffrence. Here on AT however there are alot of admins, performance users and extreme gamers so pretty much anything goes. If you want to try it go ahead, for me it's a waste of time.

Just my .02

 

[Phoenix86]

Originally posted by: bwnv

Originally posted by: Phoenix86
Virtual memory is the whole memory subsystem. The page file is one piece of VM. You're confusing the terms like so many do because 1/2 the people writting up published articles don't know the difference either. PF!=VM. What most people reffer to as VM is in fact, the PF. There's another thread asking "can I disable VM" and they're doing the same thing, confusing the terms...

Yes, windows will fragment the PF if you let it manage it, this is why I recommend setting the min=max after you have evaluated your system's requirements. Alternatively you can buy a degragmenter that will touch the PF and let windows manage it.

I would add that 90% of windows users should probably just leave it alone, they'll never see any diffrence. Here on AT however there are alot of admins, performance users and extreme gamers so pretty much anything goes. If you want to try it go ahead, for me it's a waste of time.

Just my .02

So true. Geek site=YMMV on tweaking.

This is more of a preventative maint. setting than performance. If you want more performance, get enough RAM to fill your needs.

 

[bwnv]

Originally posted by: Phoenix86

Originally posted by: bwnv

Originally posted by: Phoenix86
Virtual memory is the whole memory subsystem. The page file is one piece of VM. You're confusing the terms like so many do because 1/2 the people writting up published articles don't know the difference either. PF!=VM. What most people reffer to as VM is in fact, the PF. There's another thread asking "can I disable VM" and they're doing the same thing, confusing the terms...

Yes, windows will fragment the PF if you let it manage it, this is why I recommend setting the min=max after you have evaluated your system's requirements. Alternatively you can buy a degragmenter that will touch the PF and let windows manage it.

I would add that 90% of windows users should probably just leave it alone, they'll never see any diffrence. Here on AT however there are alot of admins, performance users and extreme gamers so pretty much anything goes. If you want to try it go ahead, for me it's a waste of time.

Just my .02

So true. Geek site=YMMV on tweaking.

This is more of a preventative maint. setting than performance. If you want more performance, get enough RAM to fill your needs.

LMAO, just thinking that too But I'd be willing to bet that most of the folks ANSWERING questions already know that.

Cheers :beer: for you

 

[Nothinman]

True, but it is in it's nature to expand, so fragmentation is possible if not likely over time. If you do use your PF a lot (you don't have enough RAM) a fragmented PF can cause additional slowness.

Sure, but the fact that you're using the pagefile at all will kill performance, having the pagefile be fragmented will be the least of your worries.

 

[dawks]

Let Windows mangage the pagefile. If your worried about fragmentation, but it on its own dedicated partition and set large cluster sizes (if you want to get fancy). That way the PF won't get mixed in with other files.

I suggest we just make a thread on this topic and sticky it. This is asked more often the the NTFS encryption question was.

 

[Nothinman]

Let Windows mangage the pagefile. If your worried about fragmentation, but it on its own dedicated partition and set large cluster sizes (if you want to get fancy). That way the PF won't get mixed in with other files.

Only put it on a seperate volume if that volume is on a seperate physical disk, putting it on it's own partiton on the same drive will just increase seek times and make things worse.

 

[Phoenix86]

Originally posted by: Nothinman

True, but it is in it's nature to expand, so fragmentation is possible if not likely over time. If you do use your PF a lot (you don't have enough RAM) a fragmented PF can cause additional slowness.

Sure, but the fact that you're using the pagefile at all will kill performance, having the pagefile be fragmented will be the least of your worries.

Agreed, but just because every system your using has enough RAM doesn't mean everyone's does.

 

[Nothinman]

Agreed, but just because every system your using has enough RAM doesn't mean everyone's does.

I still think the fragmentation is irrelevant. If memory is tight and you're using the pagefile you'll also be paging in from the tons of executables and shared libraries that you're running so seek time is going to be killing you no matter what.

 

[Phoenix86]

Originally posted by: Nothinman

Agreed, but just because every system your using has enough RAM doesn't mean everyone's does.

I still think the fragmentation is irrelevant. If memory is tight and you're using the pagefile you'll also be paging in from the tons of executables and shared libraries that you're running so seek time is going to be killing you no matter what.

Well, when I was supporting NT back in the day (not too different with regards to paging), we would see horrid problems with fragmented PFs. The systems were generally short on RAM, everyone needed a PF. We noticed MAJOR improvements by defragging a *heavly* fragmented PF. This was on systems using the PF, so I'd tend to think PF fragmentation can make a slow system much slower. Again, that was on NT 4.0...

It wasn't uncommon to see machines crashing, defragging the PF fixed it.

 

[JEDI]

laptop hibernation mode:

pagefile must be >= amount of memory, right? else laptop wont suspend/hibernate?

 

[Phoenix86]

No, hibernation uses hiberfil.sys, not pagefile.sys and that file must = RAM.

edit: Live it, love it, learn it.

 

[JEDI]

Originally posted by: Phoenix86
No, hibernation uses hiberfil.sys, not pagefile.sys and that file must = RAM.

edit: Live it, love it, learn it.

cool!

so if i up my laptop to 1gig from 512meg, i dont need to change my 768meg pagefile?

i just need to change my hibernation file?

 

[NeophytexDMD]

Hmm, from personal experience, it is good to have windows manage the paging file. First time was when I was on a computer that was struggling to play movies. Setup System Managed and windows was playing the movie flawlessly after a couple of reboots to have windows adjust the paging file size. Second time was when I helped my best friend build his computer and when windows would start up, you clicked start->my computer and nothing would happen, for minutes, then hours eventually. 3ghz proc, 1gb ram, 120gb hdd, Hmm, can't be that more ram is needed for windows to open up a harddrive and read it's contents within the hour then display them to you. Upgrade to 2gb, same thing happens. Turn on system managed, no problems. Comes up right away. Lesson learned, system managed size is there for a reason. Stop being lazy and maintain your harddrives DEFRAGMENT THEM!

 

[NeophytexDMD]

bump!

 

[Nothinman]

Why did you bump this a minute after you posted?