Windows Virtual Memory - Pagefile.sys and can it be safely disabled

This is a thread where we will discuss Windows Virtual Memory and in particular, the infamous windows pagefile and whether or not it can be disabled and improve performance. Primarily this thread will deal with Pagefile.sys and the Windows OS. After having it suggested to start a thread about this , I had felt that this was a good idea, being that there is such controversy regarding this issue. My objective is to prove that on some systems that it can be safely disabled and have a positive effect on system performance.

 

I always disable mine (on systems that have enough ram to pick up the extra load of course) and it definitely improves system performance and reduces disk fragmentation.

 

Personally, I have it disabled on my desktop server which has only 2.5 gig of ram, so far it has not caused me any troubles, nor did I noticed any decrease in performance. I am still in the stage of testing the change (removal of page file completely), I will report back in a few days, or when I encounter performance issues in the next few days.

 

Does anyone recommend any good benchmarking programs that I can use to do a before/ after results regarding the pagefile?

 

Thanks Pai

 

Try PCmark05 for benchmarking.

 

I have a Satellite L25 w/ 2GB RAM.

I disabled the page file for a week or so and found absolutely no difference. No performance increase or decrease.

I keep in enabled just in case I ever need the extra kick. I don't need the 2GB of hard drive space anyway.

I don't really understand why it would serve any benefit to keep in disabled. I've never heard an explanation that seemed logical either.

However, if you are a fan and you have an explanation I'd be happy to hear it.

EDIT: While benchmarking, make sure you open up tooooons of stuff. Every program you can find. Testing it while running nothing would seem rather useless.

 

Back in the old days apps were designed assuming a pagefile and would get cranky or flaky without one. I'm not sure it's so critical anymore, but I still insist on using one. On anything up to 4GB RAM I let it do the automatic settings. Over 4GB I set it to 1GB fixed size.

Figure if something is using the pagefile on your system now enough that you believe it's impacting performance, while you're SURE you have enough RAM that you don't need the pagefile, why is the app using it? Maybe the app knows something you don't.

 

cool, finally a possibility of names/numbers of some about-to-be-named systems.

And the concrete effect description at that ... cool.

I hope to "see" a software environmnet as well, besides the "standard" of "more than so or so gigo ram".

btw, noticed that u r "folding" ... mine always present from boot up to shutdown.
. 1st question: r u running this with NO pagefile? If so, what are the current app that u run in parallel? I did a test on this myself, on my Sager NP4750, 2gigo, XPSp3, Catalyst 8.6 - crashed on me when trrying to open a 80mb powerpoint file.
. 2nd question: running "folding" alone, and with no pagefile, have u seen a non-issue between finishing the project and when contacting for new work? Mine would hang, depending on the size of the submitted file when finished.

cheers ...

 

The only reason why I keep mine around is so I can run a bunch of VMs.

 

I have 2.5gb of ram so I deem the pagefile rather unnecessary. I noticed little or no performance increase/decrease, however disabling the pagefile makes your computer access your hard drive less, meaning more battery life. I'd suggest passmark for benchmarking, and as what a previous poster said, make sure you open up perhaps wmp and about 10 internet explorers before you start the test. Hopefully this will make the difference more pronounced.
the trial passmark is fine:
http://www.passmark.com/download/pt_download.htm
hope this helped!

oh yeah and I forgot make sure you wait about 5 minutes after reboot (in order for pagefile changes to take effect), so that all background processes are stopped. also the tests should be repeated 3 times with the pagefile off and three times with it on

 

I've disabled my pagefile but still see 841 MB PF Usage in the Windows Task Manager .

how come?

 

have u restarted your computer?

 

. can u provide some concrete numbers based on ur personal experiences on ur comp? and the way u set them up as well. if not
. can u provide some links as where i can reference "more battery life" by "disabling pagefile"? it might be cool if once can really really confirm this

cheers ...

 

off-course.....please help me

 

hmmm... I am unwilling to charge my laptop then discharge it with pagefile on and charge again then discharge with pagefile off- it wud take too long.

this quote comes from:
http://74.55.96.66/vb/all/windows/t...e-usage-easily-and-quickly-please-anyone.html
it is common sense knowing that the pagefile is virtually a substitute for ram. if you have insufficient ram to hold your applications then your computer will have to draw even more power to use the hard drive as a substitute for ram. There are other ways to test battery life other than discharging a system. One option is to use nhc, which will show how much power is being drawn from the battery.

 

too bad, no concrete numbers .. was hoping to see some, or at least that u had somehow documented ur experience.

i was hoping for a link with concrete information, and not just another quote. But if u do come across a new one, do share

cheers ...

 

To What extent do the readers of this thread want documentation?
In my report will:

Battery time with and without be desired?
Software environment and system tweaks included?
I have already performed one benchmarking test, and need to do another.
I am using the passmark test and I have problems with the complex 3d test so far with and without the pagefile.
What else would be desired? Your Suggestions Please

 

I disabled it a week or two ago and have had no issues. The main reason being that Windows sticks stuff in the page file even if you have plenty of RAM, it's using your hard drive unnecessarily. Windows task manager performance tab still shows me has having a 617 MB page file though, not sure why.

 

The paging file, I have heard, is put into RAM

 

Correction, System Cache. Sorry.

 

concrete personal number & description on these would do good for a start.

regarding Folding@Home, do u have any feedbacks from my queries earlier?

cheers ...

 

I suppose you can try to use PCMark05 as a benchmarking tool (mentioned on 1st page), but I don't know how well it will reflect the disabling of the page file. It will be interesting to see the results. Thanks for taking the time to do this, Schmi.

 

There is a valid reason to have a minimal page file on the OS partition, say 200mb. Windows needs the pagefile to create minidumps when your computer crashes. No pagefile = no minidumps = no debugging info.

 

okay dokay, we are good to go for 2nd round:

extraction from OP

with a confirmation (?) from one of many inputs:

and being concrete about numbers/statements:

i went ahead with my limited, and charlatan contribution:
. emachines m6805, 1.5gigo ram, 3700+amd, 40% diskspace free
. battery fully charged, set at 10% low warning
. power scheme set to laptop with extended battery life, probably making no difference, AMD throttles down cpu when on battery anyway.
. avast, comodo
. off the net
. XPSP3 - got my restore point, just in case
. no extra process other than the avast and comodo; the rest are windows related.

with pagefile, system controlled:
a) vlc open, watching 4nonBlondes clip "what's up" (cool one, if u ask me)
b) word doc opened (40mb size)
c) wmp11 open, listening to local playlist (dig a mix of rock&roll, classic, metal and cowgirl/cowboy)
d) played some card games
. ding-dong: 1hr42min and some seconds (ok, the battery is old, just like the rest of the machine, so please no straying away with smart remarks)

battery recharged!

with NO pagefile:
a) to d) repeating- boring,i know. Well, being creative aint my trademark )
. there were some hesitations in some part of music playing.
. ding dong: 1hr44min and some seconds
. wow, we are on something here!

. charlatan approach ON!
. battery re charged

. repeat a) to d)
. ding dong: 1hr34min and some seconds ... yikes!

. recharge battery, tried again
. ding dong: 1hr41min and some changes

taking advice from:

me stopped!

in a nutshell, i failed in showing and in convincing myself that the NON-Pagefile dingy was stretching the battery life and, in the above test, improving music enjoyment.

I did not even notice the so-called disk trashing when the page file was on (could be due to hard hearing or getting used to the comp noises?)

anyway, wait to hear and see the next concrete positive effect in system performance of a non-pagefile environment.

cheers ...

 

problem is with the 10% low warning, is that this is rarely the truth. My system can run for 2hours before reaching 10% and then 30 minutes before completely turning off. Since you (qhn) have a relatively large amount of ram your extension in battery life / performance will be less pronounced.
For your test to be completely valid, you'd have to repeat with pagefile on and off at least three times, then average out the results. You'd also have to be running severely intensive programs, for example running 10 browsers, 3 videos, and 5 office documents.
Lol if you do this, we will really have a decisive conclusion on battery life.

also to test change in performance:
1) boot your computer once with pagefile on and once with it off.
2) run the full passmark test 3 times from each configuration-
you can download from:
http://www.passmark.com/download/pt_download.htm

average out all the results and then post!

 

wow, big words from someone that does not even have some numbers to show.

my charlatan tests showed me enough that not having a pagefile, gave me no gain in battery life. It might not be valid, according to ur not.yet.to.show.ur.own method, but it served a personal and factual purpose of having some numbers to back my saying. And not hot air talk!

and borrowing ur words, put up some numbers and then post! i never claimed the benefit - u did, prove it!

cheers ...

 

@qhn

Reguarding your queries from earlier in respect to folding @ home, I had one instance of that situation happening where I will note that i have only had one problem recieving a new core from F@H. I ignored it because I had inputted a new firewall rule in Windows Firewall and I was not sure if this was the cause for i could not recieve a download for the F@H service. So i deleted the rule and restarted F@H and it works fine.The Program has worked fine on mine with the pagefile disabled. And to answer your question, yes I normally run it without a paging executive and the service starts with Windows. Although they usually tend to send me larger files - takes forever. But the last one i recieved works ok. I will monitor this more closely (F@H). I must admit that I do keep a readyboost drive handy should i need alittle more system cache. But I have not needed it up to this point.

 

@Schmi
thx for the feedback. especially about the hanging when large core was being sent down.

cheers ...

 

Here we go again...

Let me just say this. Do NOT disable the pagefile. It is an integral part of the virtual memory system, and unless you design and code operating systems for a living, you really don't know what's going on. The guys that do design and code operating systems for a living put it there for a reason, and you are not smarter than them.

You cannot start a "scientific" study by starting out with a statement like that. You have already biased the results by stating that your objective is to prove something. The correct way to approach this is to ask the questions, "Does it make a difference?" and "What difference does it make?"

What you are seeing is the placebo effect, nothing more.

Common sense is usually wrong, and in this case even more-so. The pagefile is NOT simply an "overflow bucket" that gets used when you run out of RAM. It is an integral part of the system that makes sure everything runs smoothly.

Your hard drive is there to be used. "using it unnecessarily" makes no sense. The drive is already spinning and taking up energy. The small amount of reading/writing for the pagefile is insignificant.


PS. Anecdotal statements of, "well, I disabled it and don't see any problems" are irrelevant. It only means that you don't know what problems it might cause, as well as how to see them. Even so, that evidence does not lend you license to say that ALL systems would be safe doing so.

 

Amen, Orev. Amen. I have been pushing at every opportunity to get one of the folks who claim to know it is ok to turn off the pagefile, to show why they might want to. The claim is always a faster machine or less battery use or both. I have ignored the part about it being "a good idea" (knowing it is not) and instead have concentrated on debunking the so called advantage. As of now, not a single benchmark has been show to "prove" the assertion. I am still waiting.

Gary

 

@ALL
Here is the results of the Benchmarking that was used using the Passmark Performance Test v6.1

All these are performed on the system in my Signature.

http://www.mediafire.com/?4zdcedlnz2m

The results are downloadable at the moment because i wanted to get these results up as soon as possible. I will post the pictures from the benchmarking in another post. My apologies in advance for the hassle in having to download this. i will attempt to have the pictures up within an hour.

These are the following user initiated programs and system processes and services that were running during the tests:

Internet Download Manager v5.12 Build 11 Full
10 Instances of Firefox 3 RC2 V5.0
1 Instance of Windows Media Player v11
Windows Side Bar
Passmark Performance Test 6.1 (1017) WIN32

in the background as processes and program oriented Services:

OpenOffice.org
ATI Catalyst ControlCenter V2007.1109.2138.38670
IOBit Smart Defrag
Gateway Camera Assistant
Gateway Games Process - WildTangent
Synaptics Touch Pad

Processes:

Agrsmsvc.exe
Ati2evxx.exe
Ati2evxx.exe
Audiodg.exe
CCC.exe
CEC_MAIN.exe
Csrss.exe
Csrss.exe
dmw.exe
explorer.exe
firefox.exe
IDMan.exe
IEMonitor.exe
IObit SmartDefrag.exe
lsass.exe
lsm.exe
MOM.exe
MSASCui.exe
pt.exe
pt.exe
rpc.net
services.exe
sidebar.exe
SLsvc.exe
smss.exe
spoolsv.exe
stacsv.exe
sttray.exe
svchost.exe
svchost.exe
svchost.exe
svchost.exe
svchost.exe
svchost.exe
svchost.exe
svchost.exe
svchost.exe
svchost.exe
SynTPEnh.exe
System.exe
System Idle Process
taskeng.exe
taskeng.exe
traybar.exe
Wininit.exe
winlogon.exe
wmplayer.exe
wmonetwk.exe
wnpnscfg.exe

Services:

AgeremodemAudio
Ati External Event Utility
AudioEndPointBuilder
Audiosrv
BFE
BITS
CryptSvc
DcomLaunch
Dhcp
Dnscache
DPS
EapHost
EMDMgmt
Eventlog
EventSystem
gpsvc
hidserv
KeyIso
LanmamServer
Imhosts
MMCSS
MpsSvc
Netman
netprofm
NlaSvc
nsi
PcaSvc
PlugPlay
ProfSvc
RasMan
rpcnet
RpcSs
SamSs
Schedule
SENS
ShellHWDetection
slsvc
Spooler
SSDPSRV
SstpSvc
STacSV
stisvc
SysMain
TapiSrv
TermService
Themes
upnphost
UxSms
W32Time
WdiSystemHost
WinDefend
Winmgmt
WMPNetworkSvc
wscsvc
wuauserv
wudfsvc

It should be noted that roughly half of the tests had the Folding @ Home program/service running at the same time as the 10 instances of firefox and the one instance of windows media player.

Also note that for majority of the tests, the RAM was set for the better use of Programs, not System Cache. The ones that were changed so the better part of RAM was used for the System Cache are marked as the file indicates.

The processes and services that were provided were supplied so the software environment could be seen. I hope that this attempt helps everybody, no matter their view of memory management.

Enjoy.

 

good testing! worth reading and analysing over.

Small question: the services and processes (instances) that u used. Were these active or passive? F@H is a given being active. But FF open windows and WMP, were u listening/accessing media file(s) and/or active browsing during testing time (when Passmark was running)?

cheers ...

 

hence the word 'virtually' in my statement.
anyway nice tests Daniel.

 

@qhn

You posted:
"good testing! worth reading and analysing over.

Small question: the services and processes (instances) that u used. Were these active or passive? F@H is a given being active. But FF open windows and WMP, were u listening/accessing media file(s) and/or active browsing during testing time (when Passmark was running)?

cheers ..."


Thanks, i am glad you liked them.
Regarding the question of FF and WMP activity, I was actively browsing and listening to my music library.

As to the FAHCore_82 process, it was in truth active in all the tests that state FAH, and Passive in all the tests that state otherwise. I merely paused the work so it was inactive during the tests that do not state "FAH".

As to the passivity and activity, maybe you could clarify what you mean. If you mean what i think you mean, then they were passive.

@orev

The people that make a living making these systems make OS's for the masses, what they believe that the Majority of poeple would need.
They do not make OS's for individuals, Hence the popularity of tweaking and optimizing systems (and should I dare say in some cases going to Linux)Standalone laptops with wireless connections systems do not require all the services that a server, or a desktop might need. It under normal circumstances might not require some services, then these services are there for no good reason. Optimizing ones System settings and removeing the things you do not use is like cleaning out a house of things that are just cluttering up the house and in this case a perfectly good Operating system. And to change things so performance is better all around and customized to ones particular hardware system is just common sense.
It is as redundant to ask why a homeowner wishes to remodel their kitchen when the one that was installed is good and was made by professionals.
Then why ask why in regards to an OS?
The base operating system (Vista Premium) does not know what my computing habits may be or what i really need. So hence, out with whats not needed and in with what is. Stop what i don't use and inprove or let be the things I do. i would have figured this to be "common sense".

and Orev your comment on common sense is just as "Anecdotal " as you claim our statements to be. We are not here to attack each other, attack views if you must, not people.
What i will agree with is a wrong group mentality. If the group wishes to explore an avenue that seems "taboo" by programmers but is harmless to others, then let them. The very fact that we have some of the wonderful advances we do is because people were willing to break the mold and break a few rules. Is this not the American way?
And Orev i believe the questions you suggested are being answered as we speak. Yes it may have been biased to state such a view, but as the original poster, that was one of my objective. and an objective can be proved and disproved upon as Scuderia puts it "hard evidence". As people are willing to provide that evidence is also in preportion to the strict opposition. In the pursuit of this particular view, I am not seeking conflict, but challenge.
And placebos have been proven to heal as well in some cases.

@jissac

Thanks

@ Sculeria

You stated:

Amen, Orev. Amen. I have been pushing at every opportunity to get one of the folks who claim to know it is ok to turn off the pagefile, to show why they might want to. The claim is always a faster machine or less battery use or both. I have ignored the part about it being "a good idea" ( knowing it is not) and instead have concentrated on debunking the so called advantage. As of now, not a single benchmark has been show to "prove" the assertion. I am still waiting.

Gary"

I would have to respond to your reasonable "why" with the simple "because I want to and I can". I won't tell you it is a good idea, but i can say it may be a good idea for an individual. I will not state it is good for all.
I will agree with you and orev that some programs might need a pagefile.
I will disagree that i claim to "know". I am merely experimenting and testing what people with some views might claim is a myth. I have heard of systems running no pagefiles for years, but i could never test their claims myself on any prior system that i owned because i did not have enough system RAM. But I did something that is different than you appear to be doing (at least in regards with my posts) - Having an open mind and attempting to show some civility. Be open to the possibility as i am to yours. Everyone knows a pagefile is safe, but safe is not grounds for learning firsthand. This means to me that some of these software rules can be bent. Rules are for safety, yes i will agree. But nothing ventured nothing gained.
I could listen to what everyone else thinks , but when it comes down to it, it will depend on my outlook on computing and what my system can handle.

 

But I also will consider others views

 

Above are some screenshots of the benchmarks saved as baselines.

Enjoy

 

screenshots are kind of small. Can't read them.

I think another test would be regarding Windows Updates. While it is checking for updates and downloading them it creates a new instance of svchost.exe which uses up an enormous amount of RAM and processing power. Why it does that I have no idea. Without the paging file it renders my computer useless.

Also, could you summarize some of these reports? I have no clue what any of the stuff it talks about means.

 

@schmi:
once more, very good tests and analysis - enjoyed reading it myself.

so far, my personal observation of ur detailed tests is that it is "proving" to have a positive effect on some software/working environment more than depedency of hardware (if this was what you meant by "systems"). Which is pretty much what i have long and always suspected.

active and passive: we do understand the meaning i tried to convey.

cheers ...

 

I enlarged the photos, I hope this helps.

 

This is only a partial upload of pictures because my file hosting will only allow 1.5MB upload.

 

@qhn

Thank you. I am glad my efforts were appreciated!

 

Thanks for making them bigger. If I'm reading those correctly the tests seem to suggest that there were a lot of performance improvements when using a paging file, correct?

 

They are all up now. This should help.

For the CPU tests, I found this.

What you saw was a partial list - they are all up now.
I apologise for the scrolling - I wanted the tests to be easily readable.

Enjoy

 

i m sure that i m not the only one in this department. it is always cool to receive and see personal experiences with factual inputs (scientific or charlatan).

my personal remarks are limited to ur OP and ur efforts in showing (and confirming to) me that while it is possible to turn off the pagefile and gain some (or more) performances, it is very much user working specific environment that determines the gain.

Ur great personal efforts have shown/confirmed to me (again) that:
. more system ram allows one the flexibility to consider and play with these tweak/implementation;
. more ram does not translate to an automatic non-pagefile system, as many many have believed, quoted, with zero factual backing up.

i see that ur personal test results can be used and referenced upon, for anyone who wants to consider to turn off the virtual memory. This is a far cry from the current so-called generalisation of "having a non-pagefile (hardware) system= performance gain"

cheers ...

 

the result doesn't seem too clear.
essentially, it really depends on the programs you run?

isn't it better to have more operations per second?

 

No, the result is clear. Windows comes out of the box with the pagefile enabled, and almost all computer scientists you will talk to will tell you that this type of memory system needs a pagefile.

Try as they might, those who claim that disabling the pagefile creates some sort of performance benefit, not a single benchmark has been able to prove it.

As for operations per second, that discussion is not even on the same planet as this one, so I suggest you leave the pagefile enabled.

 

then is it better to let windows manage the size of the pagefile, or to set a fixed size yourself?

 

Let Windows manage it. Part of the improvements of operating systems is that they manage themselves and you should not need to mess with them. Let it use the defaults and forget about it. The same goes for defragmenting.