Putting Paging file on it's own partition

[CjStaal]

If you put the Paging file on it's own partition, EVEN ON THE SAME HDD it will dramatically improve performance. Just make the Paging File partition 4.01Gb and FAT32 and make the min and max space 4Gb and that way it won't fragment from changing it's size all the time, and it will never get filled.

 

[Alec§taar]

Agreed, 110% - been doing it that way for decade++ now

See my subject-line/title above...

I've been doing that for years, since NT 3.51 & 1993 or so!

(& that same technique, with a LITTLE variation? It performs even BETTER if done on a disk separate from your main program & OS bearing HDD, & even moreso on a solid-state ramdrive disk).

NOTE, that in my signature below, I use this technique?

E.G.-> How & where I do it, w/ what types of files or data, & on a solid-state ramdisk drive...

I.E.-> Exactly what you're stating & THEN some... all for performance-related purposes!

APK

P.S.=> It's going to be even BETTER here this year @ some point, once the DDRDrive x1 PCI-e DDR-2 Solid-State drive releases:

That Solid-State RamDrive uses faster memory than my current SSD, a CENATEK "RocketDrive" does first of all!

(DDRDrive uses DDR memory, & the CENATEK "RocketDrive" uses PC-133 SDRAM memory)

Plus, the DDRDrive also uses a faster bus type!

(DDRDRive uses PCI-Express bus (x4 slot = 1,000MBps), vs. PCI 2.2 which my CENATEK RocketDrive uses (133mb/sec))

Faster by far as well...

Thus, not only will I have the fast seektimes ramdisks afford, but faster seeking due to faster memory type utilized over the kind my current SSD has on it, but also faster BURSTING for read/write I-O as well, because the bus type in PCI-e is MANY ORDERS FASTER than PCI 2.2... by far! apk

 

[CjStaal]

sweet ramdisk man...

(& that same technique, with a LITTLE variation? It performs even BETTER if done on a disk separate from your main program & OS bearing HDD, & even moreso on a solid-state ramdrive disk).

Yeah I know it performs better especially on a different HDD that's why I emphasised "even on the same HDD" because some people think that it won't help if it's on the same HDD no matter what you do

 

[Alec§taar]

sweet ramdisk man...

Better than the one I use now even... I, for one, eagerly await the DDRDrive x1 PCI-e model release date!

Yeah I know it performs better especially on a different HDD that's why I emphasised "even on the same HDD" because some people think that it won't help if it's on the same HDD no matter what you do

Yes, it can help some, by avoiding fragmentation of ITSELF, AND other files too (as it grows/shrinks as you mentioned...) by placing it into its OWN separate partition on the SAME disk, but faster IF done on a 2nd disk on its FIRST partition (fastest outermost tracks is where partition #1 forms) & FASTER on a SSD by far.

APK

 

[Boneface]

how does one do that....?

 

[CjStaal]

well you make a different partition, goto control panel, click system, go to advanced, performance, settings, advanced, virtual memory, change, *DISABLE PAGING FILE ON C* and then set a 4096mb Min and Max page file on the partition you are going to use.

 

[Boneface]

ok question mine is set at just over 2000mb....does that matter? and by disable u mean to do what here...this is a pic of what im seeing

 

[CjStaal]

Yopu should put it on a seperate partition

 

[trog100]

with 2 gig of real physical memory your system wont be useing the pagefile.. most of the stuff u read about improving its performance is years out of date and goes back to the days when a PC didnt have enough real memory..

alec has his on a ramdisk not a very slow hard drive.. when windows says pagefile in use it dosnt meant its useing the hardrive part of its virtual memory..

the bottom line here is simple.. if u dont have enough real memory and windows has to resort to its 300 times slower hardrive memory your system will grind to a stuttery halt..

so if u aint useing it.. it dosnt matter where u put it or how big it is..

i set mine to a minium size of 200 megs.. it never grows any bigger which quite simply tells me it never gets used with the amount of real memory i have and the software i use..

u can switch it off but somethings look for one even thow they dont use it..

i can force it to grow by loading say fear.. quake 4 and oblivion all at once.. he he he

but my 2 gigs of real memory will accomodate everything i sensibly do without having to resort to its hardrive memory..

if u hit contol/halt/delete that will tell u how much memory u have in use at any one time.. as long as the amount in use is less than the amount of real stuff u have u are okay.. if it aint buy some more of the real stuff..

trog

 

[Alec§taar]

with 2 gig of real physical memory your system wont be useing the pagefile.. most of the stuff u read about improving its performance is years out of date and goes back to the days when a PC didnt have enough real memory..

Think so? Ok... try to run Adobe Photoshop, w/ out a pagefile.sys present.

(Try to startup your OS itself, & not have it complain about it... because, it will... both apps WILL!)

alec has his on a ramdisk not a very slow hard drive.. when windows says pagefile in use it dosnt meant its useing the hardrive part of its virtual memory..

I do, & it makes a difference in performance... not just in benchmarks, but one I can literally see/feel... it only makes sense!

E.G. -> Std. electro-mechanical HDD's are 1000's of times/many orders of magnitude slower than RAM is, even on a SSD like I use.

the bottom line here is simple.. if u dont have enough real memory and windows has to resort to its 300 times slower hardrive memory your system will grind to a stuttery halt..

Well, it WILL form a "temporary" pagefile... if you don't have one present.

AND, that will fragment itself as it "grows/shrinks", & other files on the disk platter it is on as well, as they too form &/or grow-shrink.

(After all, you have to remember: This IS what "Virtual Memory" based Os' are about & the nature of filesystems too - there is "no perfect world" & fragmentation results... & it IS, a "performance killer" & a non-static pagefile.sys is subject to fragmentation of itself, & other files as well - which is what this person is stating saying you are 'best served' by putting it on a separate partition - to avoid fragmentation)

Creating the "illusion" of more RAM than is physically present in RAM chips on your mobo... by 'swapping in & out' of memory, but not entire programs, only sections of them that are NOT presently in use/locked & in 4kb/4096 sized "pages"... it has a "downside" with a dynamically sizing/resizing pagefile.sys:

Pagefile.sys fragmentation & fragmentation of other files as well, as it grows &/or shrinks in response to paging.

so if u aint useing it.. it dosnt matter where u put it or how big it is..

That's WHERE you go wrong: It is always in use!

Again, keep in mind: That this IS what modern OS' are about, paging... to create the illusion of more RAM present than there actually is.

i set mine to a minium size of 200 megs.. it never grows any bigger which quite simply tells me it never gets used with the amount of real memory i have and the software i use..

That's the catch - you have to "analyze" it, & THAT? Takes time... & effort.

You have to first see how often you page & with what apps & then set your pagefile.sys size accordingly.

You can "get away" with pagefile.sys SIZE minimization thru it... that is, until you hit an app or dataset using an app, that is TOO MUCH for what you set.

Personally? I go for the "old-school rule", albeit WITH SOME VARIATION:

OLD-SCHOOL RULE = 1.5x physical RAM present should be the size of the pagefile.sys, up to 4gb limitations of 32-bit code memory addressability.

AND YES, I go static in size to avoid fragging the pagefile.sys (and other apps, but mine is on its OWN partition, & that is all, a 1gb sized one on a Ramdisk/SSD).

I did a combining of the total of my RAM physically present on the mobo, & pagefile.sys size, to total up to what your system CAN address as a limit in 32-bit environs:

4gb, as that is ALL you can really use anyhow, in any 1 single app.

See, I did a bit of what you're stating, & a bit of the old-school rule-of-thumb:

I have no apps that ask for more than 4gb first of all, they can't after all, as the Win32 API will fail on requests larger than 4gb typically iirc, from my own experimentations there.

So, I took my RAM amt. (512mb) x 1.5 = 765mb pagefile.sys & then went up to 1/2 again that (1gb, the size of my SSD partition for a pagefile.sys)!

I have YET to get an "out of memory" based err. here... that is nearly 12 years running using this "rule" of mine, no less.

Again though - DO see Adobe Photoshop though, as it is a HUGE "hog" for memory given larger data to work on & of course, an exception to any rule exists most times.

u can switch it off but somethings look for one even thow they dont use it..

Right, photoshop being one.

i can force it to grow by loading say fear.. quake 4 and oblivion all at once.. he he he

Right...

but my 2 gigs of real memory will accomodate everything i sensibly do without having to resort to its hardrive memory..

Perhaps for YOUR use patterns, but you also note exceptions. Rightfully so, there is always exceptions to rules... and outliers on any analyzation curve.

This is also one of the MOST debated topics there is online, that I have ever seen... these are good topics to debate, they make you THINK!

if u hit contol/halt/delete that will tell u how much memory u have in use at any one time.. as long as the amount in use is less than the amount of real stuff u have u are okay.. if it aint buy some more of the real stuff..

trog

Yes, taskmgr.exe & its "performance" tab is one of the tools you can use to analyze your typical personal use-patterns in applications & data you typically use, to try to "trim" the size of your pagefile.sys... but, again:

What happens when you hit an app or data in an app, that blows away your typical use-patterns & expectations?

You're unprepared... and, get "out of memory" type errors.

APK

 

[trog100]

we agree on some things alec but will have to dissagree on others..

i have the little 200 meg swopfile now called pagefile.sys just in case i ever run things like photoshop that look for the presence of a swopfile.. but i dont use photoshop much i use psp-pro cos i am used to it.. i repeat my pagefile.sys never gets used with the amount of physical memory i have and the software i run..

u dont have to analyze your swopfile to see if its in use and windows will happily run fine without one and no it wont create a temporary one if u tell it not to..

and if u do actually run out of memory xp unlike earlier versions of windows dosnt even blue screen.. it just politely tells u that u dont have enough memory to run the application u have attempted to load..

when windows 95 first came out new machines designed to run it came with a standard 16 megs of physical system memory.. not so new machines had either 6 megs or less..

windows virtual system memory system was designed around such a scenario.. running on machines that did not have enough physical memory.. back in those days a 4 meg stick cost 25 english pounds.. 16 megs came in at 100 quid..

back then games did not run in the windows enviroment one booted out of windows and fired up in dos mode..

at present my machine has 250 megs of memory in use.. i have 1750 megs of spare physical memory doing bugger all..

if i load a memory intensive thing like a game.. lets say oblivion.. this takes up about 800 megs of my physical memory.. add this to my 250 already in use and i still have nearly a thousands megs spare..

i have put this to the test many times.. windows will only start to use its fake hardrive memory when it runs out of the real stuff..

i dont know what u use photoshop for either dude but most of the jpeg images i edit come in at about 4 meg max.. 1 gig of system memory is enough for.. errrr.. 250 seperate images or 250 undoes..

i also know that if my system did run out of real memory whilst working on large images.. instead of an image appearing in a couple of seconds it would take a couple of minutes to get hauled of the normal hardrive..

i also know that if my system runs out of real memory whilst playing a game my game will instantly become totally unplayable..

the answer is pretty simple.. i make sure i have enough real physical memory to handle the software i use.. if my system ever did grind to a swopfiling stuttery halt i would simply buy some more of the real stuff.. its cheap enough..

windows does not use its swopfile unless it has to.. a hardive is that slow compared to real physical memory that if windows is forced to use its hardrive based fake virtual memory part of things.. things become slooooo slow as to be unusable..

whatever u do with your swopfile.. if its hardrive based (unlike yours alec) your system will become a waste of space if windows has to resort to useing it..

i have 2 gigs of physical memory for a reason dude.. the reason is a simple one.. so that windows dosnt ever have to start swopfiling..

2 gigs is enough for what i do.. but i could and would easily install some more to stop my machine from having to use its hardrive to fill in for lack of the real stuff..

trog

 

[Alec§taar]

I'll say this, as I have here a few times now:

It ALL boils down to how you use your rig, point-blank!

(Application mix, tasks @ hand, etc. et al)



* I'll "get more into it" manan... I have to "attend to a female personnage/guest" who is being insistent on my attention!

APK

 

[trog100]

"It ALL boils down to how you use your rig, point-blank!"

it does.. u are perfectly correct..

but the real bottom line is the whole idea of unlimited system memory (virtual memory) by way of useing the hardrive to augment real physical memory is outdated rubbish and not really valid on a modern machine..

most younger folks.. those who belong to the cheap memory era and have never tried to run XP on 64 meg of ram or 95 on 16 megs have never seen swopfiling in action.. they read all this outdated junk about how to speed up their "virtual memory" and think it means something..

we have real physical memory (the hare) and we have the windows swopfile (the tortoise).. the two dont work that well together..

what makes matter worse is over the years the hare has been made to run a lot faster than it used to.. the speed dfifference tween the two is now so great the idea of useing the toroise when u run out of hares dont make a deal of sense any more..

1 or 2 gigs of (hare) memory is more than enough for most folks.. the silly tortoise part can be ignored and pensioned off.. whatever u do to the poor old thing aint gonna help if u have to start useing it..

trog

ps.. those still useing 250 megs of physical memory to run xp might benefit a little by speeding up the toroise a smidge.. but buying a couple more hares will benifit a million times more.. he he he

 

[Alec§taar]

Trog100, read ALL OF THIS, because the "devil's in the details", & an experiment you can perform & observe, yourself, will illustrate to you JUST HOW OFTEN PAGING OCCURS & how often "Virtual Memory" on disk in pagefile.sys is really used... here we go:

i have put this to the test many times.. windows will only start to use its fake hardrive memory when it runs out of the real stuff..

Not true, & I'll let you see for yourself how it is not. Explorer.exe is a PERFECT EXAMPLE thereof in fact, & its page faults will show you this (it is paging, all the time nearly, generating "page faults" & this is the KEY to this statement of mine - even though unused RAM is present? It is PAGING!)

You'll see below, perform the test I am noting... you'll see I am correct on this note.

windows does not use its swopfile unless it has to..

Untrue, and the experiment I am about to have you do? Will show you otherwise...

You're always using it whether you know it, or not.

Again: This IS the nature of these machines using the OS that they do... modern, virtual memory utilizing OS like Windows NT-based ones, & Linux (& others too).

In fact, let me FINALLY prove this to you, by letting YOU prove it to yourself, ok?

We'll do a little experiment below:

we have real physical memory (the hare) and we have the windows swopfile (the tortoise).. the two dont work that well together..

AGAIN: First of all did you know, that the pagefile.sys is "RAW WRITTEN" & bypasses normal filesystem access patterns & API calls? It read/written to FASTER THAN ORDINARY READ/WRITE ACCESS for your files & programs because of it & also because it is OS kernel subsystem driven (memory manager & iirc, partially cache mgr. they operate superclose with one another).

Au contraire, mon frere... They work well together & stop you from getting "out of memory" errors & the real "raison d' etre" for Virtual Memory Modern OS designs!

VERY well in fact, especially considering HOW the pagefile.sys is accessed (faster than normal read/writes to ordinary files on disk in fact)!

Let me give you an example you can SEE for yourself:

A simple way to see this, is to start a NEW program! Leave it onscreen as a normal sized window.

Then, open up Taskmgr.exe, & get into its PROCESSES tab...

Use its VIEW menu, SELECT COLUMNS submenu, & check off "Virtual Memory Size", "Memory Usage", & "Peak Memory Usage", & "Page Faults Delta" columns. They will show in the processes tab as VMSize, Mem Usage, Peak Mem Usage, & Page Faults, respectively.

Go back to PROCESSES tab in taskmgr.exe & pay close attention to the new program you started by first hiliting its row, & then, note the VMSize & pagefaults columns!

(That is what is allocated to that process in Virtual Memory (& again, keep in mind: THAT ALL MEMORY? Is "Virtual" in today's OS', both RAM in chips AND pagefile.sys space combined)).

Now, minimize the process to the startbar!

Tell us what happened to the numbers in the Mem Usage column...

(They went down for that program you just minimized, didn't they?)

Now, when you restore that window (to normal size) they go up again.

Where do you think the RAM went for that program when you minimized it?

Did the program just "get rid of it"??

NO!

(EDIT PART - possibly SOME, this is known as "discardables" to program, but usually iirc? That happens when an app starts up, or dumps data it was using, but no longer is anymore, currently).

NOTE - this doesn't happen on Win9x, I never saw it happen (while doing the min/max of a window program running test)...

THIS, is a BIG diff. in Windows-NT based OS exists here from my experiments in the past w/ Win9x vs. NT!

It paged down to disk, in pagefile.sys, what is not immediately used resources it has!

(Things like its screen images, buttons, tabs, controls of all kinds etc. from its interface & more when NOT visible/normal window view OR maximized, vs. minimized)

Right down to your Pagefile.sys & it went into the block of size allocated to it (shown in taskmgr.exe as "VM Size" in fact!)

What can illustrate that in taskmgr.exe?

The PAGE FAULT column!

As you minimize, maximize/restore a Window? You'll see it move LIKE MAD!

Here is what a "page fault" is by its definition:

--------------------------------------------------------------------------------

An event that occurs when a thread refers to an invalid (out-of-date) VM page. The VM manager must refresh the page from the page file. See VM.
docs.rinet.ru/NTServak/glossary.htm

A page fault is an exception which is raised by the memory management unit when a needed page is not mapped in physical memory. This exception is passed on to the operating system which will bring the required page into physical memory.
en.wikipedia.org/wiki/Page_fault

--------------------------------------------------------------------------------

That entire memory block is still claimed by the process, but paged out because it is no longer the foreground process & its resources (window elements like buttons, tabs, etc.) are no longer in use if not visible on screen!

They got paged out to disk into that VMSize allotted area which you have visible! Explorer.exe especially (your GUI shell)?

You'll note it is nearly CONSTANTLY generating page faults & thus, paging its data in & out of the pagefile.sys - EVEN IF FREE, UNUSED RAM, IS PRESENT!

VM Size, iirc, is based on its PEAK MEM USAGE (usually @ startup of a program, it takes a LITTLE BIT MORE to load itself than it needs to keep running from then onwards).

i also know that if my system runs out of real memory whilst playing a game my game will instantly become totally unplayable..

Again, did you know, that the pagefile.sys is "RAW WRITTEN" & bypasses normal filesystem access patterns & API calls & is driven by preferred CPU status OS kernel subsystems in the OS kernel memory mgr (ring 0/rpl 0/kernel mode operations, get pref. over usermode/Ring3/RPL3 stuff)?

I.E.-> It read/written to FASTER THAN ORDINARY READ/WRITE ACCESS for your files & programs because of it.

Secondly:

Your program that is generating PAGE FAULTING (more details exist on that above, important & ones you CAN SEE)?

It is thus, always "useable" (albeit possibly not enjoyable if paging data in & out of the pagefile)... but, it will NOT just "die" either as long as there is diskspace open for pagefile.sys growths, that is.

It will contine to run & not throw "out of memory" style errs because of it...

It will run, albeit more slowly than if it was TOTALLY in RAM & had all of its memory requirements satisfied by PURE "RAM IN CHIPS"... not paging it in & out of RAM in 4kb/4096 bytes increments.

When you have a foreground process demanding more memory than is physically present in RAM chips on your mobo that is unused then, the memory manager will page to disk what is NOT immediately locked by other running processes in 4096k increments until sufficient memory is available for this NEW process coming into existence, &/or one that is demanding more RAM to do its work.

but the real bottom line is the whole idea of unlimited system memory (virtual memory) by way of useing the hardrive to augment real physical memory is outdated rubbish and not really valid on a modern machine..

It's not "unlimited", it is limited to the 4gb max memory addressability of 32-bit OS, & goes up in 64-bit ones.

That's where you're "off" man! Today's OS ARE VIRTUAL MEMORY BASED & ALL MEMORY TO THEM? IS VIRTUAL! Even RAM chip memory + pagefile.sys in total of them both.

Even IF you don't tell Windows to make a pagefile.sys? It will, a temporary one, that is subject to fragmentation of itself & yes, other files as it "grows/shrinks".

And, the very idea behind today's modern operating systems? IS Virtual memory use because to Windows? ALL MEMORY IS "VIRTUAL" AS FAR AS THE OS IS CONCERNED!

Which is exactly what you say is obsolete!

(the use of diskspace as memory (temporary swapped out/paged out in 4kb/4096kb increments of portions of programs &/or data that are NOT in use or locked)).

most younger folks.. those who belong to the cheap memory era and have never tried to run XP on 64 meg of ram or 95 on 16 megs have never seen swopfiling in action.. they read all this outdated junk about how to speed up their "virtual memory" and think it means something..

It does! Espcially what the post starter outlined... & the fact you are ALWAYS paging in today's VM using OS'... The technique he outlines avoids fragmentation of the pagefile.sys itself, generally, & other programs on disk with it, & it works.

In fact, this whole experiment & scenario I let you do above?

That is WHY I can see my system is faster (I rarely hear the HDD in fact, mechanical RAID 0 array here with a 128mb cache onboard it & managed by an onboard CPU on the RAID controller ontop of Windows diskcache OR what I use (SuperCache-II, better cache for NT-OS)) than normal ones using std. HDD bound diskdrives as a pagefile.sys location!

Off the SSD with my pagefile.sys on it? FAR FASTER ACCESS/SEEK to pagefile.sys interior data is why too.

Others not doing this "hit disk" which are 1000's of times (literally) slower than the SSD I use for my pagefile.sys location by far...

I.E.-> I page program data & resources in & out of it 1000's of times faster than most folks can & thus, see a performance diff. in the operation of Windows, especially in terms of Virtual Memory use & your Explorer.exe GUI shell? Probably the BEST EXAMPLE of a program constantly doing it... generating page faults (read/writes to pagefile.sys).

"It ALL boils down to how you use your rig, point-blank!"
it does.. u are perfectly correct..

Yes I am.

DO read this closely, & in its entirety.

The experiment above I noted for you to try?

That should illustrate to you the concept @ hand:

It WILL show you HOW paging is used & when (quite a lot, process by process in fact & that's what I will let you illustrate to yourself below using taskmgr.exe - explorer.exe your GUI shell? CONSTANTLY @ IT, paging data in & out of the pagefile.sys in fact as one example of how often you use your HDD pagefile.sys virtual memory area).

APK

P.S.=> It's pretty easy to see, if you do what I noted above as an experiment... but, moreso with an application like Photoshop, & even Windows itself if you bootup w/ out a pagefile.sys present (permanent one), & it WILL complain... & just end up forming a temporary pagefile.sys anyhow!

I think that will illustrate to you, for yourself to see, just how often you really ARE paging, & using VM on disk in pagefile.sys (page faulting WILL illustrate that to you, without question)... apk

 

[Alec§taar]

UP for trog100 & others to read the above, & try the experiment showing how often paging occurs, & why/how, on modern Virtual Memory utilizing OS...

APK

 

[trog100]

face to face (verbally) i could probably converse with u for hours alec.. but sadly my keyboard skills aint up to doing it on here.. he he

trog

 

[Alec§taar]

The main points of yours for me to disprove were these, vs. the statement you made above quoted next (they are in error & so is the one I quote in the next post below it):

i have put this to the test many times.. windows will only start to use its fake hardrive memory when it runs out of the real stuff..

Not true!

I'll let you see for yourself how it is not in fact & gave you the experiment to do so, above.

Explorer.exe, your GUI shell, is a PERFECT EXAMPLE thereof in fact!

Its constant generation of page faults will show you this (it is paging, all the time nearly, generating "page faults" & this is the KEY to this statement of mine - even though unused RAM is present? It is PAGING!)

You'll see above for yourself:

Do perform the test I noted using taskmgr.exe, it's PROCESSES tab, & having "VM Size", "Memory Usage" (all of them present, might as well), & "Page Faults" visible on screen...

Do that, you'll see I am correct on this note.

windows does not use its swopfile unless it has to..

Untrue!

The experiment I had you do above?

Will show you otherwise... guaranteed.

You're always using it whether you know it, or not.

Again: This IS the nature of these machines using the OS that they do... modern, virtual memory utilizing OS like Windows NT-based ones, & Linux (& others too).



APK

P.S.=> Drink it in, & digest it - because it IS how things work on these machines, & modern OS...

Whether you like it or not, pagefile gets used, ALL THE TIME, & pagefaults shown happening illustrate it perfectly in fact (as well as the experiment I ask you to do above), especially in your GUI explorer.exe shell (constantly generating pagefaults - which are access read/write IO to the pagefile.sys itself, proving paging goes on ALL the time, even if FREE RAM IS PRESENT)...

It's tough to deny the facts, ones you can see for yourself no less... & so can others, if they do what I noted above... easily analyzed! apk

 

[Alec§taar]

u dont have to analyze your swopfile to see if its in use and windows will happily run fine without one and no it wont create a temporary one if u tell it not to..

That's another one you are incorrect on (in a few cases, see below)...

The OS forms in VARIOUS circumstances when a permanent pagefile.sys is NOT present, & for the very reasons I state above & which the experiment showing explorer.exe GUI shell paging even IF free ram is present!

See here:

http://support.microsoft.com/?kbid=257758

SYMPTOM:

"When you set the paging file (Pagefile.sys) on your computer to a size that is lower than the recommended size of 12 megabytes (MB) plus the amount of random access memory (RAM), a temporary paging file (Temppf.sys) may be created, and you may receive the following error message after you log on:
Limited Virtual Memory

Your system is running without a properly sized paging file. Please use the virtual memory option of the System applet in the Control Panel to create a paging file, or to increase the initial size of your paging file. Back to the top

CAUSE

This issue can occur when the temporary paging file uses a substantial amount of free space on the hard disk, and the remaining available hard disk space is less than the initial size of the paging file setting that you configured in Control Panel."

&

"The error message listed earlier in this article can occur when the initial paging file size is reduced to an amount considerably smaller than the recommended amount. Windows recognizes that the paging file size is insufficient and creates a temporary paging file of up to 20 megabytes in the Winnt\System32 folder. The System dialog box in Control Panel is then displayed, forcing you to address the paging file problem."

APPLIES TO

• Microsoft Windows 2000 Server
• Microsoft Windows 2000 Advanced Server
• Microsoft Windows 2000 Professional Edition
• Microsoft Windows 2000 Datacenter Server
• Microsoft Windows NT Server 3.5
• Microsoft Windows NT Server 3.51
• Microsoft Windows NT Server 4.0 Standard Edition
• Microsoft Windows NT Advanced Server 3.1
• Microsoft Windows NT Workstation 3.5
• Microsoft Windows NT Workstation 3.51
• Microsoft Windows NT Workstation 4.0 Developer Edition
• Microsoft Windows NT Workstation 3.1
• Microsoft Windows NT Advanced Server 3.1
• Microsoft Windows NT Server 3.5
• Microsoft Windows NT Server 3.51



(And, yes, XP + Windows Server 2003 & even VISTA too... VM & how it works? Still the same for the MOST part, & definitely in the regards we have been discussing them in, here in this thread!)

APK

P.S.=> This is a result of NT-based OS' VMS based heritage, because VMS does the SAME THING for the most part... and, I have already proven you are ALWAYS paging, via a simple experiment you were free to try & witness...

Especially how in taskmgr.exe (with Page Fault column, Memory Size, VM Size, etc present in the PROCESSES tab) will show you that, for example, Explorer.exe your GUI shell is constantly generating pagefaults (access to pagefile)...

EVEN IF FREE UNUSED RAM IS PRESENT! apk

 

[trog100]

perhaps it does always use one alec.. but in normal uses it dosnt seem to.. perhaps if i tell it not to use one it ignores me and creates an invisible one that dosnt show up on my hardrive.. i dont know..

i do know what happends when a machine genuinely dosnt have enough physical memory and windows really has to use its swopfile in earnest.. the whole machine grinds to an unusable halt..

i also know that if i set my swopfile at a minimim of 200 megs as i have.. then i force load say one game of fear (highest quality textures) plus a game of say quake 4 and create a situation where task manager tells me i have a larger pagefile in use than the actual amount of physical ram i have the 200 meg swop file will grow in size accordingly..

if i have 2 gig of physical ram.. and task manager tells me i have 3 gig in use my little 200 meg pagfile.sys file will increase in size by the neccessary 800 megs or so to accomodate the needed 800 megs and grow to about 1 gig..

once enlarged it dosnt shrink back to 200 meg even if both games are shut down until the next reboot..

i know it does this by simply looking at it in explorer..

its this kinda thing that makes me say windows doesnt use its swopfile unless it has to..

i can force it to grow by loading multiple copies of games but during normal (sensible) usage it remains at the 200 meg minimum i set it at.. this lead me to believe windows dosnt use it unless it has to.. a reasonable conclusion from where i sit..

also my windows does seem to work perfectly okay without one.. xp seems to give me the option of not having one.. if as u say it then ignores what i tell it and creates one... where does it put it and what does it call it cos i cannot find a trace of it..

i also know that if i play the multiple game loading trick without a swopfile at all.. windows lets me carry on until i exceed my 2 gigs worth of physical ram then it polites tells me i dont have enough memory to load the next applcation..

this dosnt kinda go along with windows ignoring my no swopfile command and creating one.. it suggests exactly the opposite..

are we talking about the same thing here alec cos to me it seems we aint..

i see positive evidence of one thing.. u see positive evidence of another.. admittedly we are useing different methods to arrive at our conclusions.. but something is odd..

i have run your little test by the way.. three apps.. firefox.. thunderbird and winamp.. only winamp showed different figures maximized or minimized.. the other two stayed the same..

but the test dosnt cause me to disbelieve the evidence i see with the different kinda tests i run..

trog

ps.. i will also say what i see does suprise me.. it dosnt fit with what i commonly read about windows and its pagefile usage.. ..

 

[Alec§taar]

perhaps it does always use one alec.. but in normal uses it dosnt seem to.. perhaps if i tell it not to use one it ignores me and creates an invisible one that dosnt show up on my hardrive.. i dont know..

It does, temppf.sys... see above & why plus in what conditions.

BUT paging? It's ALWAYS happening in your GUI shell, this you can see for yourself... & it has to be paging to SOMEWHERE man... someplace on a disk!

You may not note it, but you have to keep this in mind:

Pagefile.sys access FASTER than normal files do (see above) for many reasons... raw writes/reads, & iirc, using NtAPI native calls & driven by OS kernel subsystems (which DO get preference over usermode stuff) in the memory mgt. & cache mgt. phase (these 2? work REALLY close together).

i do know what happends when a machine genuinely dosnt have enough physical memory and windows really has to use its swopfile in earnest.. the whole machine grinds to an unusable halt..

Yes, of course... which is WHY I went w/ the "technique" I use (Solid-State Ramdrive placement, & that? Will only get FASTER here, once I get the DDRDrive vs. the CENATEK "RocketDrive" I use now for that).

i also know that if i set my swopfile at a minimim of 200 megs as i have.. then i force load say one game of fear (highest quality textures) plus a game of say quake 4 and create a situation where task manager tells me i have a larger pagefile in use than the actual amount of physical ram i have the 200 meg swop file will grow in size accordingly..

Right... this is what the post starter was talking about - & how to AVOID fragmenting the pagefile.sys during growth (or, shrinkage), & thus, also avoid fragging other files too!

if i have 2 gig of physical ram.. and task manager tells me i have 3 gig in use my little 200 meg pagfile.sys file will increase in size by the neccessary 800 megs or so to accomodate the needed 800 megs and grow to about 1 gig..

Again, note what the poster stated who started this thread... he's correct on that account, & using a separate partition on your first drive CAN help... but, putting it on a 2nd disk, on the FIRST partition (forms on outermost fastest tracks) speeds it up more, & also doesn't interfere with your OS & Programs housing main C: drive... speeding it up in effect, & NOT fragmenting other files there, due to pagefile.sys size changes.

once enlarged it dosnt shrink back to 200 meg even if both games are shut down until the next reboot..

Right... that's IF you use the reghack to "wipe" your pagefile.sys @ next bootup for security purposes!

(This is also doable in Secpol.msc, graphically iirc as well, & if not THAT MMC.exe snapin? Then it is in gpedit.msc (not in Windows XP Home, but PRO upwards)), especially.

It will probably reform to "normal size" @ bootup too, I will take your word on THAT account as you see it happening.

its this kinda thing that makes me say windows doesnt use its swopfile unless it has to..

Trust me, it does... again:

Do the SIMPLE experiment I show last page, showing paging occuring, in taskmgr.exe using the columns I noted? PageFaults are the key (that IS paging).

i can force it to grow by loading multiple copies of games but during normal (sensible) usage it remains at the 200 meg minimum i set it at.. this lead me to believe windows dosnt use it unless it has to.. a reasonable conclusion from where i sit..

You'd think it... but, paging being shown to be occuring by explorer.exe constantly is contrary & concrete proof, otherwise.

Also, doing the other RESTORE-MINIMIZE experiment with any app you use will show it as well, & for the reasons I noted last page.

DO TRY THEM - see for yourself, this is the BEST proof I can offer you!

also my windows does seem to work perfectly okay without one.. xp seems to give me the option of not having one.. if as u say it then ignores what i tell it and creates one... where does it put it and what does it call it cos i cannot find a trace of it..

Depends on what you do, but next time you do that? Seek temppf.sys under %SystemRoot% (%Windir%\system32)... make sure Explorer.exe can SEE "System & Hidden Files" first though.

(Some "conditions" have to be met for this to occur though, see above in my quote from MS pages in regards to it forming)...

i also know that if i play the multiple game loading trick without a swopfile at all.. windows lets me carry on until i exceed my 2 gigs worth of physical ram then it polites tells me i dont have enough memory to load the next applcation..

It should have formed temppf.sys when you booted up WITHOUT a pagefile.sys, when it warns you are "running without a properly sized pagefile" etc. (again, see the conditions MS outlines above, for THIS to happen though).

this dosnt kinda go along with windows ignoring my no swopfile command and creating one.. it suggests exactly the opposite..

All I can tell you is, look for what I told you, & do those tests... they are proof of paging happening... it's what VM using modern OS, are about... again:

PageFaulting - it's key!

are we talking about the same thing here alec cos to me it seems we aint..

Sure we are, try the tests I note last page - they're EASY to do!

(Plus, the 2nd post of mine in regard to temppf.sys? Well, that was the best proof I could give of that... Windows making a temporary pagefile when needed @ bootup, when you don't have a permanent one(again, see the conditions MS outlines above, for THIS to happen though))

i see positive evidence of one thing.. u see positive evidence of another.. admittedly we are useing different methods to arrive at our conclusions.. but something is odd..

Sure seems that way - I used Microsoft's documentation in regard to temppf.sys formation... perhaps you should test this, after reading the conditions in which it is generated on the last page I quoted?

It would be GOOD to see how "accurate" that info. from MS, is... it is fairly current too!

i have run your little test by the way.. three apps.. firefox.. thunderbird and winamp.. only winamp showed different figures maximized or minimized.. the other two stayed the same..

It's possible... but, again: Explorer.exe (your GUI shell/desktop)?

It's constantly paging, generating PAGEFAULTS.

but the test dosnt cause me to disbelieve the evidence i see with the different kinda tests i run..

Well, then, I don't know what to tell you man, I really don't!

I can only let you see evidence of paging (which IS "pagefaults" by its very definition & visible/monitorable in taskmgr.exe when the pagefaults column is selected & visible).

trog

ps.. i will also say what i see does suprise me.. it dosnt fit with what i commonly read about windows and its pagefile usage.. ..

It's from Microsoft... the horses' mouth, as to temppf.sys formations (temp pagefile.sys when you don't have one period) & the test I noted?

Well, pagefaults? They ARE paging... literally!

APK

P.S.=> Good discussion (one of the most debated online in fact, almost as much as sizing a pagefile.sys)...

Perhaps others will read this, & offer THEIR take on it...

Cj_Staal states it is a "dramatic" performance increase... I'll disagree on that, but it does boost things potentially (fragging-wise & preventing it) on the same partition ON THE SAME DISKDRIVE as pagefile.sys size changes can occur...

It WILL offer better performance if done on a 2nd HDD on its first partition (fastest, outermost tracks is where partition #1 will/should form)!

Better still, & even MORE on an SSD as I do it currently! apk

 

[Alec§taar]

OTHERS ARE DISCUSSIN G THIS HERE TOO, & others are attempting what I & CJ_STAAL state

See here, very similar thread here today:

http://forums.techpowerup.com/showthread.php?p=136233#post136233



APK

 

[trog100]

alec i have read that much conflicting rubbish about windows and its pagefile i now dont believe any of it..

take this bit from your last post..

"Originally Posted by trog100 View Post
i also know that if i play the multiple game loading trick without a swopfile at all.. windows lets me carry on until i exceed my 2 gigs worth of physical ram then it polites tells me i dont have enough memory to load the next applcation..

It should have formed temppf.sys when you booted up WITHOUT a pagefile.sys, when it warns you are "running without a properly sized pagefile" etc. (again, see the conditions MS outlines above, for THIS to happen though)."

okay it dosnt do as u say it should.. it simply functions perfectly normall providing i dont exceed the amount of physical ram i actually have.. then when i try and bung more data into memory i dont have (before i ran this test i assumed it would blue screen) it simply tells me i dont have enough memeory and aborts the application loading..

i am now gonna remove my little 200 meg minimum size swop file and look very hard for this "temppf.sys" file u tell me windows should create..

i have also tried the fat32 partition and all sorts of tricks following what i read on the net in the past.. not a jot of difference can i tell..

just for interest.. whats your take on the "pre fetch" thing.. another xp thing that tons of completely wrong rubbish is written about..

anyways i am about to turn off my swopfile and run my machine without one.. and if i cant find any trace of the "temppf.sys" file u tell me windows should create.. "I'll be back" to quote a well known movie charactor.. he he

trog

 

[CjStaal]

Trog, search google on the benifits of having the PageFile on a FAT32 partition.

 

[Alec§taar]

alec i have read that much conflicting rubbish about windows and its pagefile i now dont believe any of it..

Because you're not paying attention to the constraints the MS page lays out that I referred to (dated in year 2004, fairly current) dude!

You are not seeing this errmsg:

"Limited Virtual Memory

Your system is running without a properly sized paging file. Please use the virtual memory option of the System applet in the Control Panel to create a paging file, or to increase the initial size of your paging file."

When you bootup... & thus, it MAY not form. With 2gb of RAM?? Most likely not...

I mean, lol, w/ 2gb of RAM present, + a 200mb pagefile.sys? You may NEVER see it form!

You would have to have a SUB 12mb pagefile.sys, according to the URL, not 200mb one like you have... read the definition of the constraint (one of them).

The URL from MS: I post it again below... for your reference!

okay it dosnt do as u say it should.. it simply functions perfectly normall providing i dont exceed the amount of physical ram i actually have..

Gosh man, of COURSE not, you haven't exceeded your RAM in chips... that's part of the constraints for temppf.sys to form!

See below!

then when i try and bung more data into memory i dont have (before i ran this test i assumed it would blue screen) it simply tells me i dont have enough memeory and aborts the application loading..

Well, that's up to YOU, if you don't mind you apps crashing/abending/erring out/aborting on you...

i am now gonna remove my little 200 meg minimum size swop file and look very hard for this "temppf.sys" file u tell me windows should create..

DO! That's the test... verbatim, from that page @ MS:

http://support.microsoft.com/?kbid=257758

SYMPTOM:

"When you set the paging file (Pagefile.sys) on your computer to a size that is lower than the recommended size of 12 megabytes (MB) plus the amount of random access memory (RAM), a temporary paging file (Temppf.sys) may be created, and you may receive the following error message after you log on:

Limited Virtual Memory

Your system is running without a properly sized paging file. Please use the virtual memory option of the System applet in the Control Panel to create a paging file, or to increase the initial size of your paging file. Back to the top

CAUSE

This issue can occur when the temporary paging file uses a substantial amount of free space on the hard disk, and the remaining available hard disk space is less than the initial size of the paging file setting that you configured in Control Panel."

&

"The error message listed earlier in this article can occur when the initial paging file size is reduced to an amount considerably smaller than the recommended amount. Windows recognizes that the paging file size is insufficient and creates a temporary paging file of up to 20 megabytes in the Winnt\System32 folder. The System dialog box in Control Panel is then displayed, forcing you to address the paging file problem."

APPLIES TO

• Microsoft Windows 2000 Server
• Microsoft Windows 2000 Advanced Server
• Microsoft Windows 2000 Professional Edition
• Microsoft Windows 2000 Datacenter Server
• Microsoft Windows NT Server 3.5
• Microsoft Windows NT Server 3.51
• Microsoft Windows NT Server 4.0 Standard Edition
• Microsoft Windows NT Advanced Server 3.1
• Microsoft Windows NT Workstation 3.5
• Microsoft Windows NT Workstation 3.51
• Microsoft Windows NT Workstation 4.0 Developer Edition
• Microsoft Windows NT Workstation 3.1
• Microsoft Windows NT Advanced Server 3.1
• Microsoft Windows NT Server 3.5
• Microsoft Windows NT Server 3.51

(And, yes, XP + Windows Server 2003 & even VISTA too... VM & how it works? Still the same for the MOST part, & definitely in the regards we have been discussing them in, here in this thread!)

I mean, lol, w/ 2gb of RAM present, + a 200mb pagefile.sys? You may NEVER see it form!

You would have to have a SUB 12mb pagefile.sys, according to the URL, not 200mb one like you have... read the definition of the constraint (one of them).

i have also tried the fat32 partition and all sorts of tricks following what i read on the net in the past.. not a jot of difference can i tell..

That's pretty easy to explain - computers move in ms/ns ranges, human beings, don't... benchmark programs & such may be able to see/show us diff.'s in scores & such, but our sense usually won't - especially when say, 2 systems are VERY close in speed, period.

just for interest.. whats your take on the "pre fetch" thing.. another xp thing that tons of completely wrong rubbish is written about..

Placements of files during bootup based on usage... placed nearer to the outermost tracks of the HDD, so they are accessed faster (physics really, like how a larger tire will move you faster than smaller tires do, due to greater radius etc. & surface area covered per revolution (sort of like that)).

anyways i am about to turn off my swopfile and run my machine without one.. and if i cant find any trace of the "temppf.sys" file u tell me windows should create.. "I'll be back" to quote a well known movie charactor.. he he

trog

There you go! It cannot hurt, provided you meet the conditions noted above... If you do, you SHOULD see the temppf.sys form under %Windir%\system32 (same as %systemroot%)!

APK

 

[trog100]

i have searched google.. many times.. i know the theoretic benifits.. i have even had my pagefile.sys on a small fat 32 partition.. i have had my pagefile every size and every place.. nary a jot of difference have i detected..

as i said in my last post i now have no pagfile and am about to search for the one alec tells me windows will make.. i came back here to find its name..

trog