I think the rule of thumb use to be that if you were going to set your initial and maximum page file the same, that they should be set to about 1 1/2 of whatever your total RAM is, so if you have 512 of RAM, you'd set both boxes in your page file to 768 or thereabouts.

tell me then, why when i have 1GB of RAM and a 360MB commit charge that i'm still getting hard page faults?

the answer is very simple. even with a commit charge slightly over 1/3rd of my available physical RAM, windows still feels it necessary to use the page file. the only way to stop actual page file usage is to turn it off, period.

so if all you care about is performance, buy more RAM and forget the page file. you can get 4GBs of admittedly low-end DDR2 RAM for $200. i don't want windows paging to disk while i'm trying to record, period, and on my next rig i'll get enough RAM so it won't.

if you don't care about performance then by all means buy however much RAM you want and use the page file. you probably won't notice the difference but i can, without even benchmarking it.

ORIGINAL: mwd
I don't really understand your numbers. Early post says you have 1GB RAM installed. Above post says that 360MB is 1/3 of your available RAM.

i said that 360MB is "slightly over 1/3rd" of my available RAM. If you divide 360MB by 1024MB, you will get 0.351, which in my estimation is "slightly over" 1/3rd, or 1.3334.

If you have 1GB physical RAM installed I can guarantee it's not all available. Probably more like 650MB which would mean right off the bat 55% of your memory is committed.

do you know what a commit charge is? it is the sum of all memory in use by programs in Windows, including Windows system programs itself. All my physical memory is available, it's not like Windows refuses to use a third of it. The remainder of physical memory is typically used by the "system cache", which is recently used data (such as document files) that Windows keeps on hand in case you use them again. Linux does the same thing, the difference being that Linux does not dip into the swap partition (on Linux it's a partition and not a file) until it runs out of physical RAM to use. Windows uses the page file all the time.

I agree with you in the regard that excessive hard page faults kill performance.

good. then you would also agree that any hard page fault diminishes performance, and that is my entire argument.

My first thought is you either have a memory leak or just flat don't have enough RAM installed for what you're doing. Which is kind of reinforced because you say you have excessive disk hits.

you don't think 1GB is enough RAM to run 360MB worth of Windows + applications? do you know what a memory leak is? i never said i had excessive disk hits, only that i had them. my point was that even with a commit charge not even close to available physical memory windows still goes to the page file needlessly.

In other words your a prime candidate for turning the page file on. Helping the situation you are in is one of the last valuable uses for the pagefile.

the only reason i still have my page file turned on is because i bought Drumkit from Hell Superior and didn't want to risk running out of memory on one of my projects. if all i ever did was surf the internet and use word, etc. i could easily turn off my page file and never have a problem.

i'm sorry, but i don't think you know what you are talking about. i say this not to be rude but because i have a double major in computing and have been working with computers professionally for several years. i know what i'm talking about because i've learned the theory and applied it in practice. i know pretty well how windows virtual memory works and how it is different from linux's approach to virtual memory (which i like a lot more).

An interesting article here claims cake isn't going to page to disk while your recording anyway but rather after initial commit goes directly from disk to interface.

Not my words, can't say if true or not... just interesting.

i only skimmed the article but in those 5 to 6 seconds i was able to determine the author has not actually experimented to back up his claims. here is a quote from the article:

OK, so how do we minimize page faults, and barring that, what can we do to decrease chances that a page fault will cause a disruption to our recording? The best answer is simple: install more RAM. If the computer never runs out of RAM, then paging WILL NOT HAPPEN.

that last sentence? b.s. again, i had a commit charge of 360mb, which is slightly over 1/3rd of physical memory. i'm getting hard page faults. just because it's on the internet in all caps doesn't mean the person knows what they are talking about.

Not trying to convert you Jack. I'm pretty much the first one to do what is necessary to make my system work the best. JS runs no pagefile says it's more snappy. Jose turned his off to fix a drum problem. If for some reason it makes your box run better I don't blame you.

Thing is I have the exact opposite experience of you. Much better performance, no errors, no blue screens now in 4 or 5 years.

if you don't have enough RAM and you turn off your page file then yes, you will have poor performance as windows and running programs fight each other for precious memory space. yes, it is much more efficient in terms of physical memory usage to use the page file, as most programs do ask for more RAM than they'll need and most programs have rarely used portions of program memory which can (almost) transparently be swapped back and forth from disk.

the fact remains though that windows 2000 and onward treat the page file as an extension of physical memory. thus, if you have 1GB of RAM and a 1GB fixed-size page file, you can replace the page file with 1GB of RAM and the only difference will be your programs run faster, context switching (going from one program to another) will be a little to a lot faster (depends on how much of the other program windows might have swapped to disk) and if you run out of memory the exact same thing will happen - the last program needing memory will crash or at least stop working as intended.

i know this because i've done it. i never had a single bluescreen while running without the page file, but then i've not had a bluescreen since removing my old soundblaster live.

So myself, anybody that actually might believe what Microsoft says or anybody that turned it back on after troubles... is full-o-crap and doesn't care about performance?

microsoft doesn't care about performance, they care about minimum hardware requirements. if you don't have to upgrade to run their latest software you'll be more likely to buy it. this is exactly why all microsoft articles on the subject will praise the page file for its efficiency (which it is, but less so than linux) and say that it's not safe to turn off your page file (which it isn't, unless you have enough RAM) and that you don't need as much RAM because of the page file (which you don't, unless you want to turn it off for extra performance).

you also referenced an article on NT4 which i hope you know is an old operating system. if memory serves there were some significant changes between how NT4 handles virtual memory and how 2000 did. however you should also keep in mind that memory was a lot more expensive, relatively, back then than it is now. as i said above, you can quite literally get 4GB of RAM for $200. getting a relatively huge amount of memory even a couple years ago would have been a lot more expensive.

That one kinda' bugged.

sorry, but microsoft wants to sell you software. they have a vested interest in making you use the page file rather than buying more RAM, because you'll have more money to give to them. it appears a lot of people follow their party line on the page file simply because they've never tried it themselves. i have, and i have a different opinion. it's not worth my time and money to back it up with benchmarks and independent research though, so take it for what it is.

But on the lighter side I really don't think you meant to insult or be malicious.

The one thing I'm in 100% agreement with is more RAM is the solution.

i do not mean to be insulting but for some reason i see a much bigger percentage of disinformation on this subject than others, from people who i doubt have ever done anything other than believe that disinformation from someone else. so it gets under my skin a bit. no harm intended on my part, and i'm more trying to get people thinking and trying stuff on their own than trying to "educate."

ORIGINAL: mwd
Ya' know Jack when you stated, via implication, that anyone that thought the pagefile might actually be a good thing was full of crap I pretty much took that as a global statement. Not so much directed at me as you had just included maybe dozens, hundreds, thousands, heck maybe hundreds of thousands of others in this revelation.

Wasn't so personal then, however in your last post you pretty well zeroed in.

let me clarify, because i think you got me wrong:

the page file is a good thing unless you are concerned only with performance. i did say as much in my last post.

i don't have anything personal against anyone who disagrees that turning off the page file improves performance given sufficient RAM. when i say "full of crap" i mean their argument. you see, there are a few simple facts which everyone can agree with, and the only logical conclusion is that given sufficient RAM it is better, from a performance standpoint, to turn off the page file. that is my only argument here, which i will fully explain later.

Not a problem. I got started in computers when DOS was in the 2's and there was no such thing as a major in computing. Now I work on them 8 hours a day and play with them another 4. Seven days a week... not five. But of course that means squat.

you know, every time i try to validate my credentials, it's a mistake. invariably, the other person has spent more time working with computers or whatever than i have...

i'm sure there are a lot of things you could run rings around me on regarding computers. windows virtual memory and page file is not one of them. sorry. there's nothing you have said about it that could lead me to believe otherwise, in fact a couple things you've said have convinced me that windows memory management isn't something you understand completely. i don't know how else to put it, i'm not good at being nice.

I look at the original post and realize how far the topic has strayed. I re-read your post and must say there's a bunch I agree with... and then a bunch I don't.

Normally I would say "ok, there is a debate" or "here is a discussion" but I don't see this thread (between you and I) going in that direction.

Seems pointless when, in your opinion, the people that conceived, designed, implemented and explained the operating system you use... are full of crap.

the people aren't but their opinions on the page file stem from a financial need to sell software. if you don't have to buy more RAM to upgrade windows, you will be more likely to upgrade windows. it's a simple economic fact, and so anything coming out of microsoft will naturally be favorably slanted in their own direction. do you think they'll say that the page file sucks? of course not! will they say their memory manager isn't as good as linux? of course not!

I mean, really, what could I offer?

to be honest? in this discussion, there's nothing you can say to me that will change my mind. i know that the page file hinders performance like i know that the sun rises in the east and sets in the west. i know that it can be turned off without causing problems because i've done it.

ORIGINAL: jacktheexcynic ~ i'm sorry, but i don't think you know what you are talking about.

No apology necessary... can't be the first time I've heard that... I've been married before and have a kid.

You seem to be blunt so please allow me the same.

It's not hard for me to admit error, especially from the comfort of my own home, to a bunch of people that don't know me. I just look at it as learning something new. Nothing lost and everything to gain. I read your views, that you don't have the time to validate, and it's sure hard to take them as anything but your own opinion or theories. Your validation is you have diploma's on your wall and some knowledge of Linux. Thing is, where we are identical, I hate misinformation too and I don't believe everything I read. Including forum post.

they are my opinions and theories. i've tested them at home and know they are fact, at least in my own environment. maybe my copy of windows is special but i doubt it, so i feel that extending those opinions and theories to other windows XP pro operating systems is valid. since windows 2000 and vista aren't fundamentally different in terms of memory management i feel pretty confident in extending belief in my theory to those OSes as well.

If I tell someone their full of crap I usually feel compelled to offer some tangible evidence that lends credence to, or concurs with, my observation.

what tangible thing would you like me to do? i could fill up this thread with screenshots but that wouldn't prove anything. i could set up a web cam on my computer at home, turn off the page file, buy some benchmarking software, but then i think my wife would probably divorce me.

In lack thereof (from you) I even sought this information out, from one end of the internet to the other, that might support this pagefile theory of yours.

I just can't find it. Can you send me a link?

i'd send you a link to this thread but i don't think you'd find it funny. =) seriously though, what makes someone else's opinion better than mine? do you think microsoft is going to agree with me? where do you think everyone else gets their opinions from? some guy on the internet, or microsoft? it's very unlikely you'll find anyone to agree with me who also has the tangible evidence you seek. in lieu of that, i offer you these simple facts:

1. the windows virtual memory manager sees the page file and physical memory combined as virtual memory. the two are interchangeable from a virtual memory standpoint.

2. the page file is stored on disk.

3. disk latency and access time are much slower than RAM.

4. windows does know the difference between RAM and the page file, and swaps out seldom used pages of memory to the page file.

based on these four facts (and if you disagree with any of them, feel free to research them), it is clear to me at least that maximum performance can be gained by turning off the page file.

the only question left is, will windows still use the page file given sufficient physical RAM? the answer is yes, and i've proven it to myself by trying it out on my own computer. again, some facts (which for the most part you have to believe that i'm telling the truth or not):

1. i have 1GB of RAM on my computer.

2. the "commit charge" number in task manager refers to all OS and programs in memory. (it does not include the system cache, which is swapped out to disk as needed and not necessarily part of this discussion.)

3. during the test i conducted, the commit charge was approximately 360MB.

4. pfmon.exe (available with the windows 2003 resource kit) determined that firefox had several hard page faults during operation (i would say less than two minutes of usage, but i didn't have a stopwatch).

so with an overhead of more than 60%, windows still hits the page file on the currently active process. bearing this in mind, how can someone argue that using the page file doesn't affect performance?

finally, i add these last facts in light of the link you provided regarding DAWs and virtual memory, which you can verify by research if you don't believe me:

1. each program in windows is told it has 4GB of memory to work with, but the windows virtual memory manager actually decides how to allocate real memory (physical RAM + page file) to those programs. this method was devised to protect important sections of memory (like physical devices and the OS kernel) from idiot programs.

2. the windows virtual memory manager alone decides what pages of RAM get sent to the page file for all programs.

3. the programs themselves simply ask for memory and have no idea where it will actually come from.

this means that sonar or any other DAW has no way to say "i want to use physical RAM only." it simply cannot happen, and what will happen is that the windows virtual memory manager will make sonar swap to disk if it feels like it.

again, my purpose here isn't to educate, i simply don't have time to find references for all these things as quite a lot of them stem from the senior level course i had on operating systems, over half of which was spent discussing virtual memory management. i suppose if you dig through the internet you can find support for each of these facts or try it out for yourself. i hope that's what you decide to do, and if you actually find out that i'm wrong, and can explain it logically to me, i'll gladly change my mind. if i don't have to buy 4GB of RAM for my next DAW i can spend that money on something else.

i honestly do not mean to insult anyone here, and "full of crap" was definitely a bad choice of words so i apologize for that. misinformed is a better way to put it.

you know i'll have to try moving DFHS to that drive and see if it's any slower. i only have a pair of western digital raptors (74GB a piece) and DFHS takes up like 35GB... would be nice to free that up.

i did a speed test on it (posted the results in another thread) and while it wasn't blazing fast it was pretty good for USB2. don't remember the results off-hand but it can definitely handle streaming media. i'm going to move all my music to that disk for sure.

"Keep in mind that when you are doing any type of render or compression on DV video the disk access it not the limiting factor. For example to compress to mpeg 2 it takes longer to compress then the drive takes to store it. So no matter how fast your storage is it's still "waiting" on the CPU to compress it. When there is enough CPU power to render faster then real time so the drive is at 100% throughput then I'll worry about how fast the storage is."

That's all very true. Although in this case I was using uncompressed AVI as apposed to DV-AVI, which in turn made it render much faster since it didn't have to go through a compress/decompress scheme, and that in turn just makes your point even more valid. I haven't tried working with large DV-AVI files from that drive yet.

For the record though, I did use a sharpening filter on that entire file, and everything rendered fine and speedy, so I have high hopes with DV.

no problem, and i'm glad i was able to make myself clear. =)