Paging file

Glyphic

[H]ard|Gawd
Joined
Apr 19, 2004
Messages
1,626
I have two drives, a 7200 pata and a 10k sata, windows is on the sata, what is the best way to set up my paging file? both drives? or the non windows drive?
 
Put one pagefile on each drive. Make them approximately the same size, considering that the windows drive is faster but has OS overhead.
 
you have a gig of ram....do you ever use more than a gig? if not, just turn the thing off...unless you do video editing or somethign like that
 
M11 said:
Put one pagefile on each drive. Make them approximately the same size, considering that the windows drive is faster but has OS overhead.

You definitely don't want to split your pagefile or span it across two non-RAIDed drives. This would be a lesson in system instability.

My advice, unless your using a specific application that requires one (Photoshop), then disable it. Hardware RAM is ALWAYS going to be faster than a pagefile, at least until we have solid state hard drives.
 
What a plethora of misinformation!

However, I won't mess with this can of worms. Suffice to say that you will gain no performance boost from removing the page file. There are other arguments for using it, most that I disagree with (though they are more solid arguments), but the performance argument is ridiculously incorrect and misinformation based on ignorance about how the virtual memory subsystem in OSes—not just Windows, but all operating systems—actually works.

There are different opinions about optimal settings and placement for the paging file. None of them have shown results using empirical data in repeatable environments sufficient to prove them absolute. It will always depend on loads of factors, from machine configuration to software running on it to how many users are accessing data at any given time.
 
djnes said:
You definitely don't want to split your pagefile or span it across two non-RAIDed drives. This would be a lesson in system instability.
Instability? I have never seen this instability that you speak of. It has worked fine evertime everywhere I've done it/seen it done.
 
GreNME said:
What a plethora of misinformation!
Your right, if your referring to your own post. It's been benchmarked, proven, explained, and demonstrated in countless guides online. Not to mention when you actually DO disable it, you can notice the difference right after the reboot. Apps open faster, switching between apps is faster, and level loads in games are faster. It's one of the few tweaks you can do to a PC that you can actually notice WITHOUT running synthetic benchmarks. The only misinformation posted on the [H] in the countless threads about this are those people who claim it's not going to change anything. First hand accounts by myself and many others on here beg to differ with your "claims".
 
M11 said:
Instability? I have never seen this instability that you speak of. It has worked fine evertime everywhere I've done it/seen it done.


Maybe on your one system, but it's been strictly warned against even back in the days of NT 4.0.
 
djnes said:
Your right, if your referring to your own post. It's been benchmarked, proven, explained, and demonstrated in countless guides online. Not to mention when you actually DO disable it, you can notice the difference right after the reboot. Apps open faster, switching between apps is faster, and level loads in games are faster. It's one of the few tweaks you can do to a PC that you can actually notice WITHOUT running synthetic benchmarks. The only misinformation posted on the [H] in the countless threads about this are those people who claim it's not going to change anything.
Bullshit. Find me the "countless guides" online, and let's take it a step further: put your money where your mouth is. I'll do the same thing I've already done on this forum before, which is have multiple people take their systems, and run multiple benchmarks for different things both with and without a page file. I will guarantee you that the results will come back the same as they did then, and the same as they have on the four different machines I've personally tested: no performance increase.

First hand accounts by myself and many others on here beg to differ with your "claims".
No, those "first hand accounts" are nothing but the wonders of placebo effect.

I'll await your acceptance of my challenge, and we'll work on finding benchmarking tools we can agree on, from 3D testing to image manipulation to whatever else your heart desires.

Afterwards, I will explain to you in a little more detail how even when you have used the Windows setting to disable the page file, you are still working with virtual memory, which means that the hard disk is still being used to handle memory calls. I get a little sick of the regular influxes of people spreading misinformation.
 
djnes said:
Maybe on your one system, but it's been strictly warned against even back in the days of NT 4.0.
Back in the days of NT4 everything was shit and everything was blamed at one point or another. And I have done this not just on my SIX personal systems with multiple drives, but on countless custom gaming PCs that I've assembled. In a few days when I wipe my system, I think it might be benchmark time.
 
GreNME said:
Afterwards, I will explain to you in a little more detail how even when you have used the Windows setting to disable the page file, you are still working with virtual memory, which means that the hard disk is still being used to handle memory calls. I get a little sick of the regular influxes of people spreading misinformation.
PF != VM, it's a subset of VMM. Yes, you still page, and use VM when you disable to PF.

The performance boosts would be difficult to measure, how do you measure alt-tabbing load times? I wouldn't go so far to say it helps with load times, however.

What removing the PF does do is increase paging operations, not general use of the PC.

Show me a benchmark that does measures that aspect and you have your tool. I'm aware of nothing that does this. Which is why it's hard to get a handle of what it does help. However, even if it can't be measured, that doesn't mean it doesn't help.

djnes, you know I agree no PF is better, I do it myself, but I have never seen this 'proof' on other tweaking guids. Do you have a link to tests run by others?

So if there is no proof that this works, and I still do it... Why?

Because, when you system IS paging, you can definatly notice it. The HDD starts going nuts when memory is demanded by a new application (IE loading a new app, this may be where djnes is getting 'faster load times').

Scenario1a:
1GB of RAM, 500PF. 700MB memory used in open applications. Opening a new application that requires 100MB RAM will cause the system to page 100+MBs of RAM (always pages more than requested to have extra available to the next app) to the HDD (writes 100MB of data to the drive). As the system pages, it executes the application, putting everything into RAM at once (there is enough to load it immediatly). So the application opens, and the system pages. If the application is also reading from that disk or IDE chain your application will not open as fast, since the HDD is active on more than one request.

Scenario1b:
1GB of RAM, 0PF. 700MB memory used in open applications. Opening a new application that requires 100MB of RAM will cause the system to immediatly open the app, no paging necessary (some VMM stuff still happens, and can be noticed if PF useage, which measures all paging, IIRC). No extra disk activity is happening.

Scenario2a:
1GB of RAM, 500PF. 700MB memory used in open applications (mostly a game). When you load a new level, or even an area in a new level your memory requirements change. Because of this you may start paging, because windows thinks it needs more RAM, even though it will never hit it's limit (because you have already tested the max RAM required). When this happens, your performance will go to poo in the game. 5fps anyone? I have seen this in UT2k3/4. Chugging along at 40-50fps, and you get 2-3 seconds of high disk activity, BLAM, 5-10fps. This is a problem in most games...

Secnario2b:
1GB of RAM, 0PF. 700MB memory used in open applications. When you load a new level, or even an area in a new level your memory requirements change. Because of this, nothing special happens, it just uses the RAM you already have. 40-50fps consistant.

Now, measure that... Sure you can run fraps, and measure FPS through the whole thing. However, you can't 'make' windows page on command, so you can't compare it vs. another setup. Run the same benchmark at the same time, and the system may not page at the same point, heck it may not page at all... That's the problem, I see, windows pages at inconsistant times.

The only way around this delay in paging would be to have a seperate disk and controller, so the disk activity from paging didn't hold the data you do need hostage until it was done. Most people can't afford/don't have a high speed disk and controller for swap files only.

Heck most people are lucky to have 2 HDDs, and one is generally older/slower, making it a poor choice for a page file, even then it's generally one the same controller...

Sans PF isn't for everyone, just people with enough RAM and a normal storage situation. This happens to describe a lot of people who have 1GB of RAM, but only 1 HDD IDE/SATA HDD.
 
Phoenix86 said:
PF != VM, it's a subset of VMM. Yes, you still page, and use VM when you disable to PF.
I never said that PF == VM. In fact, I am saying that those who use the "memory bus is faster than HDD bus" argument are missing the point that the virtual memory management is still happening (at the lowest level, through the HAL), so you're not skipping any steps in the path of memory management. In other words, Phoenix, that (PM != VM) is precisely what I'm saying. :)

The performance boosts would be difficult to measure, how do you measure alt-tabbing load times? I wouldn't go so far to say it helps with load times, however.
Well, I have some for-pay benchmarking tools that can do that, pretty much by running a series of scripted processes (using, for instance, MS Office), and outputting results. However, I can't expect everyone who wants to challenge this to go out and spend money on them.

What removing the PF does do is increase paging operations, not general use of the PC.

Show me a benchmark that does measures that aspect and you have your tool. I'm aware of nothing that does this. Which is why it's hard to get a handle of what it does help. However, even if it can't be measured, that doesn't mean it doesn't help.
Do you mean increase or decrease there when talking about disabling the PF? I thought the point was to decrease the number of page faults on running programs.

As for the benchmarks that may be able to measure this, let me double-check some of the licensing for the tools and make sure I'm allowed to publish results (yeah, weird license agreements). Otherwise, I would suggest running two low-level benchmarks together and have a script switch focus between them at intervals.

As for the scenarios, I would agree with some, and disagree with others (usually on minor things). Still, Phoenix makes a far more compelling argument than "it just works faster!" That I can understand. Phoenix is in favor of disabling the PF (in some situations, with caveats), but approaches it from an informed position rather than propagating misinformation. I can deal with that. :)
 
GreNME said:
I never said that PF == VM. In fact, I am saying that those who use the "memory bus is faster than HDD bus" argument are missing the point that the virtual memory management is still happening (at the lowest level, through the HAL), so you're not skipping any steps in the path of memory management. In other words, Phoenix, that (PM != VM) is precisely what I'm saying. :)
I just want to point that out before the discussion gets of track, so many people read one thing and apply it to another. And the "RAM is faster than HDD" still applies, just only to the part of VMM you move from disk to RAM, namely the PF. However, people thinking removing the PF will stop all paging are dead wrong. This is why people get confused when they look in the task manager and still paging activity (with no PF).

Phoenix86 said:
Quote:
What removing the PF does do is increase paging operations, not general use of the PC.

Show me a benchmark that does measures that aspect and you have your tool. I'm aware of nothing that does this. Which is why it's hard to get a handle of what it does help. However, even if it can't be measured, that doesn't mean it doesn't help.

GreNME said:
Do you mean increase or decrease there when talking about disabling the PF? I thought the point was to decrease the number of page faults on running programs.
I meant increase the speed of paging operations because it's move to RAM instead of the HDD. To decrease the actual number of page faults... *I'm* not sure that's possible without a major hack/change to the VMM system. The point isn't to reduce the number of page faults themselves, but when paging does happen, it's not noticed as much.

If you could reduce the number of page faults required, that would be an improvement on VMM, and I think MS would want to talk to you. ;)
 
To the people who insist on arguing about how we should still have page files, I offer this informal test. I had BF1942: Desert Combat loaded on my system. I eliminated the need for a cd by using the no-CD crack (relax, I own the retail game). I did this so it would be purely from the HDD. I ran the game WITHOUT a page file. My level loads were in the average realm of 10 - 12 seconds. WITH a pagefile, set static at 300 MB, my level load times were 25-30 seconds, and I saw extensive drive HDD activity during the game itself, as opposed to very very little during the same game and level without the page file.

My computer in question is a gaming system...pure and simple. Given the nature of the "tasks" it is asked to perform, having no page file has ALWAYS been better.
 
Prove it, djines. If you can't prove it, you are just explaining the typical "it feels faster" syndrome. Currently, there exists only one source I know of who claims faster performance without the PF, and that is Quack Viper, who doesn't know his ass from a hole in the ground to begin with. I offered my challenge. You can either take it or quit claiming that there is any real proof. Anything else would just be silly.

Phoenix:
If you could reduce the number of page faults required, that would be an improvement on VMM, and I think MS would want to talk to you. ;)
To be honest, this would more be a matter of having the programs themselves not paging as much, IMO. Obviously, the program can't dictate to the OS how to manage memory, and the OS can't arbitrarily assign times for page faults, but I think the answer lies somewhere in-between writing better management for the OS alone and providing better development tools for programs. Frankly, I think too many software companies are applying old-school thinking when developing, and OS vendors are trying to maintain insane backwards-compatibility with it, creating a vicious cycle. Both Linux and Windows, as far as I can tell, have very effective VMM systems, but most programs don't make active use of them. I would posit that has something to do with the cost of re-upping training and cleaning house of sloppy code.
 
GreNME said:
Prove it, djines. If you can't prove it, you are just explaining the typical "it feels faster" syndrome. Currently, there exists only one source I know of who claims faster performance without the PF, and that is Quack Viper, who doesn't know his ass from a hole in the ground to begin with. I offered my challenge. You can either take it or quit claiming that there is any real proof. Anything else would just be silly.


Well, ummm, I can count seconds...as I explained in my previous post. First off, spell my nickname right.....second, what possible to you hope to gain by trying to get me to believe your opinion? Are you that insecure? My opinion, based on testing and performance guides I have read in the past, when XP was first released absolutely recommended for gaming performance, disable the pagefile. I attended an event in the First Union Center in November of 2001 for the release of XP, and I listened to the small lecture given on Windows Gaming and XP. Microsoft's gaming reps themselves suggested it. You know...the ones who develop games for the PC? Furthurmore, it definitely does FEEL faster, which is more than anyone can say for most tweaks. I don't really have time for little pissing contests as your suggesting. I gave you a count of the seconds for load times in games. As I stated before, my system is a gaming machine ONLY....so if it speeds up my load times, you can be damn sure I will do it.

As with most things, maybe your results may vary. Sorry if you don't like my opinion, but for my system(s) it does make a difference. I proved it on my own system, and that's all I care about. If there was a true test of this, I would gladly do it when I return from my vacation. I'm finishing up my RAID vs non-RAID comparison as we speak.
 
djnes said:
Well, ummm, I can count seconds...as I explained in my previous post.
Which you cannot compare to other people's experience and hard evidence, as it seems to contradict repeatable results.

First off, spell my nickname right.....
My mistake, consider it done.

second, what possible to you hope to gain by trying to get me to believe your opinion? Are you that insecure?
Ad-hominem attacks only make your argument that much weaker. When in doubt, toss insults? Not very smart.

My opinion, based on testing and performance guides I have read in the past, when XP was first released absolutely recommended for gaming performance, disable the pagefile.
Your opinion offers no repeatable results, and is based on anecdotal evidence. Unless you wish to take my challenge and see if you can prove it with hard evidence, you are simply talking nonsense.

I attended an event in the First Union Center in November of 2001 for the release of XP, and I listened to the small lecture given on Windows Gaming and XP.
Who gives a crap? I've been to more Microsoft seminars than I can count, there are Microsoft programmers who frequent this board, and all of them will tell you the same thing: disabling the PF gives you no performance boost.

Microsoft's gaming reps themselves suggested it. You know...the ones who develop games for the PC?
Get a solid quote, not something you are dredging up from memory of three years ago. Get hard evidence of someone who actually develops making this claim. I will guarantee you that you cannot.

Furthurmore, it definitely does FEEL faster, which is more than anyone can say for most tweaks.
Very nice to hear you have convinced yourself that it is faster, even though empirical data shows otherwise. People used to say the Earth FEELS flat, too...

I don't really have time for little pissing contests as your suggesting. I gave you a count of the seconds for load times in games. As I stated before, my system is a gaming machine ONLY....so if it speeds up my load times, you can be damn sure I will do it.
Translation: you don't want to deal with the embarrassment of being wrong after so adamantly refusing to aquiesce, even when someone else who disable their PF (Phoenix86) has questioned your claim.

As with most things, maybe your results may vary.
1 + 1 = 2. 1 + 1 will always = 2.

Sorry if you don't like my opinion, but for my system(s) it does make a difference. I proved it on my own system, and that's all I care about. If there was a true test of this, I would gladly do it when I return from my vacation. I'm finishing up my RAID vs non-RAID comparison as we speak.
Sorry you don't like facts, but your opinion means jack and squat when compared to actual facts.

Like I said, at least Phoenix makes a cogent argument for why he does it. You might want to take some pointers from him.
 
GreNME....look at you giving me a lecture on testing results. Wow...I had no idea of those things...despite all my lab time in college on scientific testing. Imagine that.

Since you had trouble reading it, while you were dreaming up your ridiculous responses, I will say it again. Come up with a scientific test, and I will do it.

Let me play your game for a second, big man.....put YOUR money where your mouth is. Show me proof that it HURTS performance. Don't give me links saying it's not recommended. I am fully aware of the drawbacks. Don't lecture me on how XP really does page, even without an actual file. I am fully aware of all of this. Show me how MY system is performing less than it would if I do have a pagefile.

There's nothing more immature that accusing someone of something you yourself are doing. Your previous post is riddled with hypocrisies. I gave you the results of my own testing. Counting seconds is hardly anecdotal. Seriously man, you'll get no where in this industry putting down others to make yourself feel important. We have a name for you....we call them "Marketing Folks".

I have no trouble admitting when I'm wrong. I always praised RAID setups on a desktop....even after reading Anand's article. So, I did my own testing and have found out it really doesn't show much of a difference, except in the synthetic benchies, like PCMark04, Sisoft, etc.
 
I really do wish I had a camcorder...I'd record myself loading up games/maps as well as navigating menus with and without a pagefile. Memory will always load faster than a hard drive...so why keep your system on the hard drive?


Are there tests out there that measures the ability to switch applications/load game maps...etc?
 
Calm down, you guys may actually figure something out from each other as your both smart.

djnes, I can't tell you the number of times I read/typed djines only to correct it, odd...

A couple of things here... Anecdotal evidence is one thing, however a measured time is another. When a user tells me, "I pushed the power button in for 3-5 seconds and the computer won't turn off" I don't believe them... I say "OK, push the button in until I say to let go" and I count 1-1000, 2-1000, 3-1000 *click*

User: oh, it worked this time...

When I technician says it takes 13-20 seconds difference, I believe he measured it. A watch, stop watch, counting, I don't really care, with that much difference it isn't just 'feeling'; it's obvious. Very different than the earth is flat anecdote.

When I said I'm not sure about load times, it's because I have not measured it, or even noticed. Then again, I'm not running BF1942, which is a major RAM hog. Maybe there is poor coding causing excessive page faults when loading. If you speed up paging (move the PF to RAM), it might affect load times greatly.

The largest RAM hog I run is NWN, it likes to make my system page when I load an area with lots of spell effects on players, or when people with lots of spell efffects enter my line of sight. This isn't a good thing if your trying to kill/run from said player. Sometimes I would lag so bad (hdd churning away) I would see the person enter, and by the time I got a click in to move, I was dead.

That is anecdotal evidence. :D

Get a solid quote, not something you are dredging up from memory of three years ago. Get hard evidence of someone who actually develops making this claim. I will guarantee you that you cannot.
That's a bad statement. You cannot guarantee what others will say, regardless of who is correct on the facts. A better statement is you have not read this yourself, and/or read the opposite.

All of this debate, and no one has asked the most simple of questions...

WHY DID MS PUT THE OPTION IN THE OS? :p

I think the answer to the argument lies in that question.
 
Phoenix86 said:
WHY DID MS PUT THE OPTION IN THE OS?


You know, that is something that has been missed. Obviously, there is a reason for a user to have that option. Last time I checked...all joking aside, there's no option for "Enable BSOD on reboot", so one would take that to mean it's at the very least, not a negative thing. But as usual, thanks for the comments. Very good points brought up there.
 
djnes said:
GreNME....look at you giving me a lecture on testing results. Wow...I had no idea of those things...despite all my lab time in college on scientific testing. Imagine that.
Gee, shall I give my testing lab specs and credentials as well? Looks like you're the one who is trying to have a pissing match.

Since you had trouble reading it, while you were dreaming up your ridiculous responses, I will say it again. Come up with a scientific test, and I will do it.
There are already tests out there to handle programs in a manner that would test the page file. However, they cost money. Not a whole lot of money, but they cost. And as I already said, we can choose from an agreed-upon list instead of one person dictating the parameters. You seemed to have missed that little point.

Let me play your game for a second, big man.....
More ad-homs. Your position is wavering.

put YOUR money where your mouth is. Show me proof that it HURTS performance.
Reading is fundemental. I never said it hurts performance. I said it doesn't help. It's a bogus "tweak" that yields no discernable advantages.

Don't give me links saying it's not recommended. I am fully aware of the drawbacks. Don't lecture me on how XP really does page, even without an actual file. I am fully aware of all of this. Show me how MY system is performing less than it would if I do have a pagefile.
Nice straw man tactic. However, I'm not arguing about lessened performance, and I'm not on your machine. I am stating facts about how the virtual memory works across the board. You are taking my statements and turning them into an argument they are not. That's your problem, not mine.

There's nothing more immature that accusing someone of something you yourself are doing.
As I already stated, I've done it before. Here on this forum, with not a single result yeilding increased performance, not even from those claiming it would.

Your previous post is riddled with hypocrisies. I gave you the results of my own testing. Counting seconds is hardly anecdotal.
Yes, it is. Anecdotal evidence is where you are testing without repeatable and verifiable results, taken from a personal (I "feel" it) point of view. Counting seconds does not a scientific experiment make. Have the results recreated on other people's machines, under the same circumstances, and you will have data. To date, you have supplied none.

Seriously man, you'll get no where in this industry putting down others to make yourself feel important. We have a name for you....we call them "Marketing Folks".
Once again with the ad-hominem. I can assure you that you will get nowhere in the industry if you can't prove what you say. To date, you still cannot.

I have no trouble admitting when I'm wrong. I always praised RAID setups on a desktop....even after reading Anand's article. So, I did my own testing and have found out it really doesn't show much of a difference, except in the synthetic benchies, like PCMark04, Sisoft, etc.
Dropping names (like Anand) and naming benchmarks isn't giving you any more credence. Changing the subject isn't making you look any more believable. Show me where Anand has yielded recordable and repeatable test results for the page file, not another subject.

Fark_Maniac said:
I really do wish I had a camcorder...I'd record myself loading up games/maps as well as navigating menus with and without a pagefile. Memory will always load faster than a hard drive...so why keep your system on the hard drive?


Are there tests out there that measures the ability to switch applications/load game maps...etc?
No, but there are system stressing programs out there (from eTestingLabs, for instance) that load more than one app using a script and perform various tasks to put numerous loads on the system from different standpoints. This is done to see how program suites (like, say, MS Office) perform under different hardware configs and such.

Phoenix86 said:
That's a bad statement. You cannot guarantee what others will say, regardless of who is correct on the facts. A better statement is you have not read this yourself, and/or read the opposite.
Indeed, but I am seriously trying to give the benefit of the doubt here.

Phoenix86 said:
All of this debate, and no one has asked the most simple of questions...

WHY DID MS PUT THE OPTION IN THE OS? :p

I think the answer to the argument lies in that question.
For the same reason Linux, BSD, OSX, and Solaris (among other OSes) have a swap file. ;)

The question to follow that up with would be: What is the optimum configuration for it?

The answer there is the same as many OS questions: it depends on the user, the hardware, the tasks it will be performing, and the duration/load level of regular operation.

Therein lies the rub: it's not always a simple answer, and using anecdote as fact is not a recommended practice. Not in the small-business consulting realm, nor in the enterprise-level IT realm. Not in the development realm, either. This is why I strictly recommend against making any changes that are simply anecdotally praised. On the other hand, Phoenix, you have shown considerable arguments for your own suggested configuration, under the right circumstances. Also, the fact that you do not always recommend your configuration, as well as not claiming unmeasurable results, gives your argument enough credence that I have no choice but to accept it as viable. I don't have to agree with it, but I can equally not say it is without merit.

As for optimizing page faults, I have to admit I have quite a few other issues to handle before I'm up to the task of making a noticable change to something far more people are working on as a job (since I don't get paid to program... I have to be an everyman). Still, it's something to think about for me, and I've already got plenty of documentation downloaded to help me out when I look into it. It never occurred to me to actively look at such optimization though, so thanks for the comment. :)
 
djnes said:
Last time I checked...all joking aside, there's no option for "Enable BSOD on reboot", so one would take that to mean it's at the very least, not a negative thing.
There are, however, settings to not have it automatically reboot on BSOD, and to choose between dump file sizes. So the option is there, as well.
 
Now he's calling me a name-dropper because I referenced Anand's article about RAID setups? How F'N funny is that? PUUULEASE stop this madness now. Did you hear the sound of a toilet flushing? That was your chance at any credibility from me. OMG I can't stop laughing enough to even type this! Is that your mission to disagree and piss off everyone else in this thread? Your succeeding. He he he...I'm a name dropper because I reference an uber popular website on an unrelated topic. BAH HA HA HA HA!
 
GreNME said:
And the ad-homs just don't stop. :rolleyes:

For the third time, you come up with a test, and I will do it. Keep in mind when doing your "brainstorming" that my system is for gaming only, and that was my only point all along. My game level load times decreased greatly by disabling my page file. On workstations where ultimate stability is concerned, I always leave a page file. On servers, it's an absolute given to leave one. Like I said, come up with a testm since your all-knowing. I do not own a camcorder, so that is out of the question. You have yet to add anything productive or even insightful here. I'm asking for a test that will disprove my earlier findings that my level load times have gone way down.
 
GrenME, you beat around the bush so much, just admit it there is a reason... :)

I never said it hurts performance. I said it doesn't help.
First, take a stand. The change will change performance, good or bad, it will. It may or may not be measurable, but I guarantee this: changing your PF options to, static size, dynamic sizes, locations, spanning, or no PF, will affect performance.

I see your hedging from your position now, again, or maybe I'm reading closer... If "it (doesn't) hurt performance" and "it doesn't help performance" that only leaves one options left.

"it does help performance" :)

What is the optimum configuration for it?
The answer there is the same as many OS questions: it depends on the user, the hardware, the tasks it will be performing, and the duration/load level of regular operation.
That means, in some cases, with some hardware, and some software, the option IS a valid optimal setting. This is why I bring up the option, the fact it's there means it is likely to help in some cases. It also means it's not likely to negatively impact your settings.

EDIT:
Children... please. In between all your bickering is a valid thought. Pointing out how many ad-hominem attacks or if someone misspells your name doesn't resolve the question at hand. It also doesn't mean there isn't a point. I can call someone a jackass in the middle of explaining how the world isn't flat and that doesn't make me wrong...
 
What's that sound? Sounds like...someone coming. I can hear keys.
You don't think it's a mod with a lock, do you?

While the pissing match IS quite entertaining, I can't help but notice that it's also a bit pointless...how does that quote go? "Arguing on the internet is like winning in the special olympics...in the end, you're still retarded." While I'm not calling anyone retarded (you both obviously know something about the subject at hand), you guys have kind of "egged each other on."

djnes: GreNME's just not going to concede unless you run benchmarks that HE specifies, and gets results that HE can repeat.

GreNME: I would suggest that timing things with stopwatches IS data...just not a form you're comfortable with. And djnes just isn't going to concede that "it feels faster" isn't reason enough.

So how about this:
The next time someone asks "Should I disable my Page File", you guys say this:
I believe so/not, but there is a lot of disagreement on it. But there are areas of agreement...
1. It won't hurt it.
2. Some programs won't run (Photoshop, for example)
3. Some people say it helps, but only if you have a certain amount of memory (and there is disagreement on that too)
4. Some people say it doesn't help
5. I have yet to see convincing benchmarks either way
6. Try it, and see what you think

I'm headed out right now, or I'd disable it and see...how about a poll? Everybody disable the page file (temporarily) and measure (with a stopwatch) the loading times for some programs, and games (try a few levels too.)
So...what should we load (for uniformity), and which games? I call on [H]'ers to settle this! There are enough people on these boards, and enough combined knowledge, that we could actually come up with a reasonable answer...wanna try?
 
O[H]-Zone, your advice on how to answer these questions is very good, and is what we've followed in the past. When someone offline asks me that question, I immediately ask what the purpose of the machine is, and how much memory it has. Those two factors are how I determine if it's safe to disable the page file. I have my one gaming system with no page file. I have 2 test machines, one HTPC, one server, and one laptop that all have 1 GB of memory in them, and they all have the pagefile enabled. Maybe, for the sake of GreNME, I should have spelled out that I felt it was a gaming tweak.

And others have made valid points...a stop watch recording time is a very valid form of data collection, especially in computing comparisons. How long it takes a computer to run a task or set of tasks is a very common benchmark in most online and print review sites.
 
Stop watches work for the olympics...

O[H]-Zone, right on. :)

I have not seen people recommend no PF w/o discussing needs and hardware first, esp. not djnes.

OK, admittedly I wasn't as hyped about doing some benchies the first time GrenME wanted to, but if we can solve a long standing debate... I'm in.
 
Phoenix86 said:
Stop watches work for the olympics...

O[H]-Zone, right on. :)

I have not seen people recommend no PF w/o discussing needs and hardware first, esp. not djnes.

OK, admittedly I wasn't as hyped about doing some benchies the first time GrenME wanted to, but if we can solve a long standing debate... I'm in.


I'd be more than happy to count in "mississippi's" if that is more acceptable to GreNME.
 
Phoenix86 said:
First, take a stand. The change will change performance, good or bad, it will. It may or may not be measurable, but I guarantee this: changing your PF options to, static size, dynamic sizes, locations, spanning, or no PF, will affect performance.

I see your hedging from your position now, again, or maybe I'm reading closer... If "it (doesn't) hurt performance" and "it doesn't help performance" that only leaves one options left.

"it does help performance" :)
No, you're missing an option. This isn't binary. There are three outcomes in a true scientific test with a control and a variable:
  • Positive (or proactive)
  • Negative (or reactive)
  • None (no action)

If you leave out one of these possibilities, you are not conducting a proper experiment or test for a variable.

Phoenix86 said:
That means, in some cases, with some hardware, and some software, the option IS a valid optimal setting. This is why I bring up the option, the fact it's there means it is likely to help in some cases. It also means it's not likely to negatively impact your settings.
You may hold that opinion, but I have never seen a configuration where this is so. I have seen workstations, servers, renderers, home-internet-browsers, and game rigs. None of them would have performed better sans paging file. Some of them would perform worse, and others would gain nothing. Still well within the three options that changing a variable can produce.

O[H said:
-Zone]djnes: GreNME's just not going to concede unless you run benchmarks that HE specifies,
Wrong, as I've already stated.

O[H said:
-Zone]and gets results that HE can repeat.
Wrong again. Results that anyone, repeating the exact same steps, can repeat consistently. There is currently no data showing consistency of the claims, much like the disabling services issue.

O[H said:
-Zone]GreNME: I would suggest that timing things with stopwatches IS data...just not a form you're comfortable with. And djnes just isn't going to concede that "it feels faster" isn't reason enough.
A stopwatch isn't going to count as data unless everyone present or a third, objective party is present at each timing to give the results. I seriously doubt this is going to happen, considering the distances involved.

This is the most annoying part of the debate: everyone but Phoenix (it seems) is unwilling to take what is actually said and debate against it, so they end up debating against something that isn't there. It's ridiculous.

Phoenix86 said:
Stop watches work for the olympics...
Where everyone is present to both observe and request verification of results. If you all want to plan a get-together to hash this out, that's cool. We'll share a few beers over it.

Phoenix86 said:
OK, admittedly I wasn't as hyped about doing some benchies the first time GrenME wanted to, but if we can solve a long standing debate... I'm in.
Now all we have to do is find methods that all parties agree on from which to test. Without that, everything is still well in the realm of conjecture.

djnes said:
I'd be more than happy to count in "mississippi's" if that is more acceptable to GreNME.
You find a way that will count mississippis in a verifiable and consistent manner (meaning human counting is out), and you will finally be agreeing to my original challenge.

Gee, it only took a whole page.

Now, let's find benchmarks that can either be run in unison or separately, then let's get a thread started with screenshots to show results. In the lack of any image hosting, I can host it myself. I am leaving the benefit of the doubt to those who participate that they will not jack the results in favor of their position, and comport themselves professionally in the matter.
 
This is my one and only warning. If you guys cannot keep it civil this one will be locked.

If you want to both do benchmarks, then go right ahead and bench your systems. Post the results, system specs, and apps used to "bench" with and we'll go from there. Bickering back and forth doesn't solve anything. Results will.
 
On to the benchies!!! (and my problems with finding one...)

Things we need to agree on. First, we need a method. How do we 'induce' paging so we can test it's performance? Loading word, excel, and powerpoint isn't going to cause a 1gb system to page, but it might cause a system with 256MB RAM to do so...

Next, we need an application. We can compare specific things like djnes' example of load times of BF1942, but if we don't all have bf1942, I don't think we can get enough data to compare. Maybe not, maybe there are enough people out there willing to test the exact same thing. We would need, at least, 3 machines with the same software to get a basic comparison.

Then, we need a measurement tool. Here I'm at a loss, suggestions? There may be a tool/app combo that will work, but I doubt it because of what we are measuring, PF performance, not just load times, overall FPS.
 
aside from the flaming :rolleyes: , I got what I needed fromthis thread, thx for the replies though
 
Phoenix86 said:
Things we need to agree on. First, we need a method. How do we 'induce' paging so we can test it's performance? Loading word, excel, and powerpoint isn't going to cause a 1gb system to page, but it might cause a system with 256MB RAM to do so...
Agreed. However, having different numbers from different configurations (and RAM sizes) is acceptable to me.

Phoenix86 said:
Next, we need an application. We can compare specific things like djnes' example of load times of BF1942, but if we don't all have bf1942, I don't think we can get enough data to compare. Maybe not, maybe there are enough people out there willing to test the exact same thing. We would need, at least, 3 machines with the same software to get a basic comparison.
Once again, agreed. That's why I hesitated to suggest the benches that cost money. :(

Perhaps running things in tandem, with a small script ensuring they start at the same time?

Phoenix86 said:
Then, we need a measurement tool. Here I'm at a loss, suggestions? There may be a tool/app combo that will work, but I doubt it because of what we are measuring, PF performance, not just load times, overall FPS.
If we're using tests running in tandem, we could try to find something to test the total machine performance as a general measure of performance. It's not paging we have to measure—we already know paging goes on—it's overall performance.
 
You could use a photoshop script with different transforms and filter operations, with an image of sufficient size such that the system is forced to swap. Be sure though to split the PS page file across both drives as well.

this is the image I use for PS benchmarking.
 
GreNME said:
Photoshop requires a page file to work. It'll be biased.
I meant with regard to one drive or two drives
 
Back
Top