Paging file

whatever happened to this? I do not see any results!

I just read the Myths thing...and am still calling shenanigans on it. Let's see some hard facts before we do or do not call it bogus.
 
Well, I'm still waiting back on some others' free time and suggestions. Also, djnes has said he is willing to contribute as well (which is great), so we're just trying to work our schedules together. :)

I'm thinking maybe those of us who are into trying to come up with something should trade e-mails (you can get mine either in my profile or by PM-ing me).
 
If I had a back up PC, I'd help. And which version (SP1/2) would you be using?

-J.
 
Fark_Maniac said:
whatever happened to this? I do not see any results!

I just read the Myths thing...and am still calling shenanigans on it. Let's see some hard facts before we do or do not call it bogus.
Click to expand...
We still need a method to test paging. We all know paging happens, heck if a large page is done, you see it when your HDD lights up a lot more than normal. The problem is, it doens't happen on command. IE I can fire up a game and test average/min/max FPS over a given time and level, but you can't say 'OS, start paging' and test performance there. I have enabled a swap file on my work laptop (beginning to think that was a mistake) and it paged for a good 15 seconds last week alt-tabbing between apps. That was a good 15 seconds where I was doing exactly jack shit on this machine (low end PC, 550mhz, 512 RAM).

Now if I could make that happen on command, then test the same scenario with and without a PF, we would have our test.

I have been thinking about this a bit, and I may have come up something, it's kinda complex. It's also something that has to be sized per machine its being tested on. So it may not be something 'everyone' can run unless you setup the test machine exactly the same. The reason being the OS's method for when it pages. In general, the OS will try to page 'enough' RAM to disk for whatever app your now running.

Lets assume a 1GB system and ~120MB is taken by the OS. That leaves ~900MB free RAM.

OK, so what we need to do is make it page, preferable when we launch a game or while it's running. So we need to starve the machine a little for memory. However, as suggested before a single large file open may not get us the results we want. In fact it may skew the results to the no PF side of things. Since it is one large piece of data, the OS will likely page it, all or nothing. This may be a good test for RAW PF performance, but it won't be very realistic. How many of us run one app taking a lot of memory, then fire up a game. None, it's much more likely we have several smaller applications taking chunks of data.

I'm thinking a spread of data in various sizes would be better. If we load a 25MB, 50MB, 75MB, 150MB file (does the app matter? I'm thinking a pic loaded in simple viewer, not photoshop), that leaves ~600MB-(overhead of the pic programX4) free. If we launch a game which requires ~500MB RAM it should cause several of the picture files to page to disk to free up extra space for the game. However on a no PF system there is enough RAM to load everything.

Thoughts?
 
I've got myself a copy of Winstone 2004 made by VeriTest. I'm running it on my game machine. It has a fresh copy of SP2 on it, but windows was installed months ago. For me, this test will be usefull as it is testing the environment it will be currently using.
 
For the original poster.....

Just put windows/programs/etc on the raptor, and make a 2G partition at the front of the IDE drive, format Fat32 4K cluster size. And set this to be your pagefile partition.

Commonsense dictates that the one place you will get a small performance increase is by having the pagefile on a seperate drive/controller so normal disk accesses and pagefile accesses can be done in parallel....up to the 133MB/s PCI bus limitation, assuming you are using a PCI based (even motherboard builtin controllers) sata controller, like NForce2/Sil3112, etc.

Then get a 2nd Raptor, setup raid0 and we can have another 50 page tiraid thread on that :D
 
I have completed 3 tests using Winstone 2004 ver 1.0.1. They are as follows:

withoutpagefile.jpg


with512pagefile.jpg


with1gbpagefile.jpg


pagefilesummary.JPG


So from what I can gather, according to Winstone 2004, a pagefile makes little to no difference to my machine. Granted, I've got a pretty responsive machine, these figures may differ with a much slower machine.
Since these set of tests indicate negligible diff with a pagefile or with out one...I'll keep the "placebo" feeling that no pagefile feels quicker in some (important) aspects.

**Disclaimer**
Your dog has cooties / this pertains to my computer specs only / I do not know how much pagefile swapping is involved as I did not sit and watch it all the time...it takes too frickin' long.
 
GreNME did similar testing, and found similar results. This doesn't surprise me because you likely weren't paging during testing, it's also why he say's it 'doesn't help or hurt performance.'

You can only test the speed of FTP, when your FTPing. You can only test a tweak to paging operations, when paging...
 
while the test was running, it had excel speadsheets getting created/building a website on building your own machine/archiving emails...wouldn't that make a machine pagefile?
 
Depends on your memory usage. Even if that caused paging, did you do the *exact* same thing in each program during each test?

You have 1GB RAM, how much of that was used when testing was started? How much when it ended?

If you started with ~200, and ended with ~400, your system may have never paged. If you started with ~200 and ended with ~900, it should have. This is one of the things I don't understand about paging, when does the system page, and how does it decide what to page? That's why I suggested to fill memory with the pattern above, and give the OS a selection of things to page to disk.
 
Fark.... dude, you need to hit the gas pedal on those KHX3000 sticks... Im at 240Mhz with same 11,2,2,2,cas2.0

Here's a test.... with 512M of ram, run FARCRY or DOOM3 both of which hit peak commitcharge's of 800-900MB. So it will be paging its a$$ off. Like on TimeDemo1 you actually CARE about the first run, as this slowdown is caused by the disk accesses. So you could use this to get a real FPS measure of the affect of different pagefile setups.

Just a thought.
 
uwackme said:
Fark.... dude, you need to hit the gas pedal on those KHX3000 sticks... Im at 240Mhz with same 11,2,2,2,cas2.0

Here's a test.... with 512M of ram, run FARCRY or DOOM3 both of which hit peak commitcharge's of 800-900MB. So it will be paging its a$$ off. Like on TimeDemo1 you actually CARE about the first run, as this slowdown is caused by the disk accesses. So you could use this to get a real FPS measure of the affect of different pagefile setups.

Just a thought.
Click to expand...

Man..I did. Had them up to 230 running 3v stable....for about 2 weeks. Then one stick started getting memtest errors. RMA'd...I'm happy for now...until I get the itch again to tweak it out. I'm running an ancient stick of HyperX ram...and now a very new stick of HyperX. I have no idea if either is BH5. I've seen pics on the inet that some are not.
 
Yeah I got 6 sticks of the KHX3000 from ZipZoomFly.... all 6 were BH5. All started out 185Mhz max, but are now 245Mhz 11,2,2,2,2.0 @3.3V after burnin. I yanked the heatspreaders, they cost ~10Mhz in OC... they are heat BLANKETS ;-) I have a nice 120mm fan(side door) blowing on them, and my 9800pro's ramsinks.... all are nice and barely warm to the touch at full speed. Best $228 x 3 I ever spent :D At this point I wont ever RMA (for fear of no BH5 chips for repair). Instead I have a spare 256M stick of BH5's that I will use for repairs myself if the need arises.... helps haing access to top-notch SMD rework equipment!

I took my own advice, and moved the Pagefile to a 2G partition at the outer edge of an 80G/8 Seagate IDE (primary master), while Windows/etc is on the sata raid0 92x maxtor 80G/8 on Sil3114. Seems nice and smooth, hard to quantify the value, but its only 2G, and the drive is there for other reasons anyway .....ex: VirtualPC dont do raid, etc.

So not sure if its a big deal, but its easy enough to do, so why not.
 
Just upgraded 512mb to 1gb ram and am gonna try shrinking my PF from 768MB to something much smaller like maybe 40MB. Then I'll try a free ramdisk program and move the PF into a ramdisk.
 
Moving to a ramdisk is POINTLESS.

You taking away RAM from the pool to create this ram drive, only to RE-ADD it back to the pool in a different form. This makes no sense, just disable the page file and you WILL move this operation to RAM without the overhead of a RAM drive.

edit: OK, let me break this down a little.

Ex. You have 1GB RAM.+ 200MB PF=1.2GB of VM, but you never use over 1GB of RAM and want to remove the PF and leave 1GB of VM.

Scenario1: Delete the PF. You now have 1GB of VM, all RAM.

Scenario2: Delete the PF. Create a 200MB PF on a RAMDrive, which reduces your RAM by 200MB. You have 800MB RAM+200MB PF (also located on RAM). You now have 1GB of VM, except to access the 200MB in the pagefile, the system must process the overhead of a RAMDrive (which isn't much, but it's there).

Scenario1 is clearly better, it gives the system the same amount of resources as Scenario2, except there is no RAMDrive.
 
OS = 2000 Pro

No option to disable PF completely, doesn't seem to want to let me keep the PF on the ramdisk anyway but that may just be specific to the app I'm using to create the ramdrive.. Anyway I wanted to put my IE temp files in the ramdisk as well.
 
The advantage of no PF is reduced time to swap. You system isn't going to choke when it's accessing a 2MB file. However, when you have a 700MB file, and it needs to swap, it takes time.

I'd bet there is little difference between 0, and 2MB PFs. I'd still avoid the RAMDrive, it just seems useless. It would be worth testing, if we had a method to test...
 
It seems like every test indicates that running without a pagefile offers no performance benefits, and at the same time can cause serious performance deterioration if the physical memory is maxxed out.

So what's the point of running without a PF? I see none and I see no reason to recommend to others to run with none.

-MrD
 
If I turn pagefile off, reboot, I see no pagefile.sys, but whatching Taskmanager[Performance], there's even more paging going on.....Where is this going ? :confused:
 
wallijonn said:
Gentlemen,

May I suggest that you use the Admin tools' "Performance Monitor"?

http://windows.ittoolbox.com/browse...g&r=http://www.wown.com/j_helmig/wxpperfl.htm

http://www.wown.info/j_helmig/wxpperfm.htm

You can set whatever parameters you agree to (page faults, page file access, disk access, etc.), set an interval for however long you want to, run a suite of your choice and output to a file to be read after the test completion.
Click to expand...
Monitoring it has not been the problem. The problem is how to enduce it acuratly so it can be measured.
 
man... what a can of worms... after reading the first 20 posts... i felt the need to reply. Time to hose the testosterone off the deck. Sheese! :rolleyes:

Windows (of anytype) must have a pagefile. it can be 20 MB or 2000MB but it has to have one. Eliminating the pagefile does make the system unstable, because some of the kernel can be corrupted in a memory low condition leading to a crash or instability. Having 4GB of memory doesnt eliminate the need for a pagefile.

Also, the person that stated that having multiple page files causes instability is wrong. Windows keeps stats on how busy a volume is and if there is a pagefile on a volume that is less busy, it has a bias to use that one before the others,

Doctor X
 
Doctor X said:
man... what a can of worms... after reading the first 20 posts... i felt the need to reply. Time to hose the testosterone off the deck. Sheese! :rolleyes:

Windows (of anytype) must have a pagefile. it can be 20 MB or 2000MB but it has to have one. Eliminating the pagefile does make the system unstable, because some of the kernel can be corrupted in a memory low condition leading to a crash or instability. Having 4GB of memory doesnt eliminate the need for a pagefile.

Also, the person that stated that having multiple page files causes instability is wrong. Windows keeps stats on how busy a volume is and if there is a pagefile on a volume that is less busy, it has a bias to use that one before the others,

Doctor X
Click to expand...
There is PLENTY of empirical evidence that counters your statement. Try reading the whole thread where several people state their systems are stabe as can be with no PF (me included).

OK, you made the claim, back it up. Prove that disabling the PF causes instability.

Next, I open WXP and see a radio button in the PF settings that clearly reads "No Paging File." If windows must have a page file, what does this option do?
 
http://www.xoxide.com/cooler-master-cool-drive-6-silver.html

That hard drive cooler shows you your maxed MB/s transfer rate, and your sustained/current MB/s transfers using the specified HD thats in it.

I have done tests and it showns that without a pagefile, the io access is very low meaning its not swapping to the HD, meaning load times are very quick.

Turning on the pagefile shows the a very high sustaind transfer rate meaning its swapping to hd, or "hd thrashing", in turn makes loading slower.

We are talking about load times. Using ram is about 100 times faster than using a hd to store the ram contents onto.

I'm also guessing you're saying solid state hds are about the same speed as eide/sata drives too right?
This thread is childish..In all of GrenME's posts he's asking people to post their NON HUMAN results, yet he has not even posted any of his own. Show us your non-human error proof pesamistic results, please?

PS. I must be blind when that digital reader is showing my sustained transfer rate is higher with a page file on correct?
 
Phoenix86 said:
There is PLENTY of empirical evidence that counters your statement. Try reading the whole thread where several people state their systems are stabe as can be with no PF (me included).

OK, you made the claim, back it up. Prove that disabling the PF causes instability.

Next, I open WXP and see a radio button in the PF settings that clearly reads "No Paging File." If windows must have a page file, what does this option do?
Click to expand...

It's getting to the point of worthless, man. You know, as well as I, that some people on here can't bear to me wrong, and they will continue to believe their misinformation, no matter how much proof you give someone. Doctor X is leaning towards this infamous group. I have 7 systems in my possesion at the moment, all used for various purposes, running XP. Not a single one of them has a pagefile, and not a single one of them is unstable. As the argument goes, it all depends on the amount of physical memory, all of which have 1 GB of memory. Its becoming useless to try and prove people wrong. Back in the good old days, when you were wrong, you laughed about it, and learned something new. Now, these people insist on not backing down, despite a lack of proof or evidence. Amusing at times....frightening at others.
 
You know, djnes, I wouldn't have so much of a problem with it if the ones who refuse to back down would at least understand the opposing argument. Hell, I don't agree totally with Phoenix on the subject, but we both understand what each other is saying enough to see the relevance of our positions (I forget where you stand on it).

As the specs of the general end-users' systems go up, this increasingly becomes more of a non-issue. I like the one Phoenix got stickied because it gives explanations and details to the actual questions of the PF instead of blanket statements. Much more constructive than this thread, IM(ns)HO.
 
djnes said:
It's getting to the point of worthless, man. You know, as well as I, that some people on here can't bear to me wrong, and they will continue to believe their misinformation, no matter how much proof you give someone. Doctor X is leaning towards this infamous group. I have 7 systems in my possesion at the moment, all used for various purposes, running XP. Not a single one of them has a pagefile, and not a single one of them is unstable. As the argument goes, it all depends on the amount of physical memory, all of which have 1 GB of memory. Its becoming useless to try and prove people wrong. Back in the good old days, when you were wrong, you laughed about it, and learned something new. Now, these people insist on not backing down, despite a lack of proof or evidence. Amusing at times....frightening at others.
Click to expand...
There will always be ignorance, myself included. I'll gladly try to expose BS when I can. It's trying at times, but it's the good fight. :)
 
Phoenix86 said:
There is PLENTY of empirical evidence that counters your statement. Try reading the whole thread where several people state their systems are stabe as can be with no PF (me included).

OK, you made the claim, back it up. Prove that disabling the PF causes instability.

Next, I open WXP and see a radio button in the PF settings that clearly reads "No Paging File." If windows must have a page file, what does this option do?
Click to expand...

There was a knowledge base article that stated the very fact. I will try to hunt it up. I never had that much memory in my pc's but i have had plenty in my servers.. (and i ran stuff that was critical so I didnt experiment.)
 
djnes said:
It's getting to the point of worthless, man. You know, as well as I, that some people on here can't bear to me wrong, and they will continue to believe their misinformation, no matter how much proof you give someone. Doctor X is leaning towards this infamous group. I have 7 systems in my possesion at the moment, all used for various purposes, running XP. Not a single one of them has a pagefile, and not a single one of them is unstable. As the argument goes, it all depends on the amount of physical memory, all of which have 1 GB of memory. Its becoming useless to try and prove people wrong. Back in the good old days, when you were wrong, you laughed about it, and learned something new. Now, these people insist on not backing down, despite a lack of proof or evidence. Amusing at times....frightening at others.
Click to expand...

Well, are you special. A whole 7 systems. I have no idea why you insist on flaming me and being rude. Running XP eh? Well, in case you didnt know ... you ARE running a pagefile even though you said no. Because XP creates one when booting if it cannot find one. It will create a temporary one about 1.5 times the amount of memory installed. And will remove it at shutdown I think (I had clear pagefile at shutdown turned on). I just tested this last night. To test, create a separate parition. Create a page file on that new partition. reboot. then shutdown and use partition magic to delete that partition. When you come up , windows will create a new temp pagefile on c even though you said no. I tested this on xp sp2. In prior versions, windows would just die in a memory low condition. I got that information from a knowledge base article. I will see if I can find it.
 
I havent found the knowledge base article, but I did find some other references. You make your own judgment.

http://www.gotapex.com/forums/archive/index.php/t-76697

I did discovered, even though I had a temp pagefile running (because I deleted the part with the pagefile on it) I could not start some of my games. Some check to see if a pagefile exists,.

Under win2000 there was an issue that windows would die on startup.

Can the Virtual Memory be turned off on a really large machine?

Strictly speaking Virtual Memory is always in operation and cannot be “turned off.” What is meant by such wording is “set the system to use no page file space at all.”

Doing this would waste a lot of the RAM. The reason is that when programs ask for an allocation of Virtual memory space, they may ask for a great deal more than they ever actually bring into use — the total may easily run to hundreds of megabytes. These addresses have to be assigned to somewhere by the system. If there is a page file available, the system can assign them to it — if there is not, they have to be assigned to RAM, locking it out from any actual use.

Another article:
http://www.winnetmag.com/Article/ArticleID/42035/42035.html

I just googled these... Your mileage may vary. Make up your own minds.
 
Doctor X, your severly misinformed (no offense, there is LOTS of confusing, misleading information on this subject). First, take a gander at this thread.

Well, in case you didnt know ... you ARE running a pagefile even though you said no. Because XP creates one when booting if it cannot find one. It will create a temporary one about 1.5 times the amount of memory installed. And will remove it at shutdown I think (I had clear pagefile at shutdown turned on). I just tested this last night. To test, create a separate parition. Create a page file on that new partition. reboot. then shutdown and use partition magic to delete that partition. When you come up , windows will create a new temp pagefile on c even though you said no.
Click to expand...
OK, you just learned (if you read that link) that VM and the PF are two seperate things. In your example you specify that a PF should exist (first bolded comments), then delete it (outside the OS) and reboot. It's normal for XP to create a PF on boot, if one doesn't exits *AND one is speficied. The KB article you are probably thinking about is the one that says the PF doesn't need to be backed up, and the OS will recreate it if needed.

However, if you don't have one speficied, it won't regenerate it. To test, set the PF option to no PF., reboot, and search for pagefile.sys.

BTW, you were not flammed.
 
Doctor X said:
Can the Virtual Memory be turned off on a really large machine?

Strictly speaking Virtual Memory is always in operation and cannot be “turned off.” What is meant by such wording is “set the system to use no page file space at all.”

Doing this would waste a lot of the RAM. The reason is that when programs ask for an allocation of Virtual memory space, they may ask for a great deal more than they ever actually bring into use — the total may easily run to hundreds of megabytes. These addresses have to be assigned to somewhere by the system. If there is a page file available, the system can assign them to it — if there is not, they have to be assigned to RAM, locking it out from any actual use.
Click to expand...
PF != VM.

The discussion going on here is about disabling the PF, not VMM.
 
Phoenix86 said:
Doctor X, your severly misinformed (no offense, there is LOTS of confusing, misleading information on this subject). First, take a gander at this thread.

OK, you just learned (if you read that link) that VM and the PF are two seperate things. In your example you specify that a PF should exist (first bolded comments), then delete it (outside the OS) and reboot. It's normal for XP to create a PF on boot, if one doesn't exits *AND one is speficied. The KB article you are probably thinking about is the one that says the PF doesn't need to be backed up, and the OS will recreate it if needed.

However, if you don't have one speficied, it won't regenerate it. To test, set the PF option to no PF., reboot, and search for pagefile.sys.

BTW, you were not flammed.
Click to expand...

ok then, can you explain what i experienced then? because I had the system partition set to no pagefile? Are you saying since I had one specified, even tho not on the system partition it would recreate it? Just trying to figure out what I saw
 
Doctor X said:
ok then, can you explain what i experienced then? because I had the system partition set to no pagefile? Are you saying since I had one specified, even tho not on the system partition it would recreate it? Just trying to figure out what I saw
Click to expand...
You saw the OS regenerate the PF becaues one was specified when you shut it down. You removed the PF w/o the OS knowing. You said:
Create a page file on that new partition. reboot. then shutdown and use partition magic to delete that partition. When you come up , windows will create a new temp pagefile on c even though you said no."
Click to expand...
In that scenario, you never disabled the PF, you deleted it.

Never said it was about disabling VMM.
Click to expand...
Then stay on topic. Many people confuse VM with PF, and reply to PF issues with VM quotes, like you did.
 
oh brother.... I give up. Everyone is such a great mood! Man... I hope you guys are not this rude in real life. This is just a conversation, it is not like your life is depending on it. I didnt confuse the two... I cut and pasted that quote from a site while I was at work. So I didnt look at it closely.
 
Don't ask questions if you don't want answers, and don't spread the FUD, otherwise you'll get countered.

If you just want to chat, I'd suggest AIM or IIRC.
 
Phoenix86 said:
Don't ask questions if you don't want answers, and don't spread the FUD, otherwise you'll get countered.

If you just want to chat, I'd suggest AIM or IIRC.
Click to expand...

Heh...you were right...very trying indeed. No matter how good of an explanation or how many links you provide, some people will insist on harping the same old points. GreNME and I had a small battle back and forth, but in the end we talked, understood each others point...and have built a respect for each other. This is rare on here, unfortunately.

Phoenix86...have a cold one on me tonight!
 
Doctor X, you're being intentionally obtuse and rather rude yourself. Give it a rest, unless you want someone who has less patience than Phoenix does to chime in.

For what it's worth, I am firmly in the "should have a page file" camp as far as that goes, and yet I still find your comments on the issue horribly misinformed and twisted to argue against straw men.

The page file is not the whole of virtual memory management, and even though paging still takes place when a PF is removed, it does not go to a page file in the file system. Then you post two ridiculous links. One is simply a forum discussion in the same vein as this thread, adding no new information. The winnetmag page offers flawed information (not all programs in i386 or later can address 4GB), then a reg 'hack' that does not actually do what he claims (it reduces core kernel files being paged, but does not totally do so). That doesn't totally negate the answer he wrote, but it gives an indication that while he may have some valid points, he does not fully understand that which he is referring to.

Your arguments are very much the same here: I agree that a PF should exist, but the way you are arguing it is flawed.
 
You must log in or register to reply here.
Back
Top