Optimize XP - A Windows XP Optimization Guide

[KoolDrew]

Yeah set correctly as in set to a static size

When I said set up correctly I did not mean like that. Setting the initial and max the same is stupid. You should however set the min high enough that Windows does not have to resize the pagefile when more of is needed.

That is covered here Xp Myths

You are missing many myths such as Virtual memory (which happens to be in yoru tweak guide), DisablePagingExecutive, IoPageLockLimit, IRQ8Priority and AlwaysUnloadDlls. And I think all of these tweaks (except Virtual memory adjustements) are made with that program called Cacheman and that program is not reccomended.

I wouldn't need a website for that I built about 1000+ PCs with it installed before XP came out.

And I am a 15 yr old who has only built 1 PC. Building a PC has nothing to do with your understanding of the Os at all.

I dont want to take credit away from you GeneralAres. You have a decent site with some pretty decent information. You provide a lot of information that might be common knowledge for most of us geeks but most average people should be able to walk away with a better, more secure operating system.

While the site does have some decent information it is tweak guides like this that causes everyone to have a misunderstanding of how virtual memory actually works or what it is.

 

[Crusty]

Maybe I can help persuade....

I just got back from my Computer Systems Architecture class. My professor is an adjunct professor who is a senior architect for Intel's Mobile division. Today we covered compiling, assembling, linking, and LOADING. We were specifically dealing with MIPS which is a 32-bit cpu, so just like the IA32 it has 4gb of addressable memory. He must have said at least 10 times during the lecture that once a program gets loaded it lives in it's own little world of 4gb Virtual Memory. The program doesn't care where it really is stored, whether it be in physical RAM, or in a swap space. It doesn't care, nor does it matter to the program. What matters is how the operation system handles the Virtual Memory. A swap file is not Virtual Memory, it is simply an extra storage medium that the OS can utilize in order to present a program with it's 4gb of Virtual Memory as designated by the 32-bit Addressing of a 32-bit CPU.

 

[Sunner]

Originally posted by: MCrusty
Maybe I can help persuade....

I doubt it, but one can always hope.

 

[CTho9305]

If you're interested in Virtual Memory, I'd recommend reading the lecture slides for lectures 20, 21, and 22 here. Note that this is a computer engineering course, so it's covering mostly the hardware perspective. There's a computer science lecture here (lecture 18). The CS slides are probably easier since the CE class has that CS class as a prerequisit

 

[GeneralAres]

No where in that article does it say to set the min and max sizes the same. It does say that fragmentation can cause a performance drop and I don't deny that. But I do deny that the performance drop will be anything noticable, if you're already paging that much then a few extra seeks here and there won't be noticable. It's sort of like pounding out a small dent on the hood of your car to get better airflow and performance.

So why then did Mark Russinovich the author of Win Internals write Page Defrag and post on his site:

Paging and Registry file fragmentation can be one of the leading causes of performance degradation related to file fragmentation in a system.

See what your saying doesn't hold up PERIOD. It doesn't hold up to references you quote.

If you want I can setup an XP Myths page describing what VM really is on my comcast webspace and link to it too, would that be credible to you?

Nope you need a source considered credible by more then some Anandtech Forum members.

You just lost credibility. ME was a piece of crap. Sure, you cant consider any of the 9x OSes stable, but which OS COULD you consider 100% stable? All OSes have their issues.

Right and how many installations of it did you use? This has to be the biggest Myth propagated by computer forums in regards to an OS. Windows ME was meerly an upgrade to Windows 98SE that include WDM Driver Support, System Restore and made an attempt to hide DOS since it was long past dead anyway. I'm sure I would think it is crap too if I tried to use Windows 95 and 98 (not SE) drivers with it.

When I said set up correctly I did not mean like that. Setting the initial and max the same is stupid. You should however set the min high enough that Windows does not have to resize the pagefile when more of is needed.

Right, so stupid that by default Windows sets the maximum to roughly what it is recommended to in the guide. Your whole comment makes 0 sense, If you have a different min and max size then at some point the pagefile can become defragmented.

You are missing many myths such as Virtual memory (which happens to be in yoru tweak guide), DisablePagingExecutive, IoPageLockLimit, IRQ8Priority and AlwaysUnloadDlls. And I think all of these tweaks (except Virtual memory adjustements) are made with that program called Cacheman and that program is not reccomended.

You didn't read the guide. DisablePagingExecutive is not a Myth. Well Cacheman is recommended thus why it is in the guide.

And I am a 15 yr old who has only built 1 PC. Building a PC has nothing to do with your understanding of the Os at all.

It obviously does.

 

[KoolDrew]

See what your saying doesn't hold up PERIOD. It doesn't hold up to references you quote.

We already told you fragmentation of the pagefile has very little affect on performance. This is because the pagefile is NOT read as a whole file (from one end to another). Instead it is accessed in very small chunks. And when I say very small I mean a few tens of a KB at a time. The pagefile is also not the only file involved with paging. Many other files such as exe's and dll's are also involved with paging, because of this the head of the hard drive is jumping all over the place anyway so fragmentation has no affect on performance and is irrelevant. Also as Nothinman stated under the most odd circumstances it may increase performance. Think about it. If the head is jumping all over the place and that little chunk in the pagefile that needs to be accessed is closer to the other data being accesses it would be faster. Again this not probably not be noticeable though.

Nope you need a source considered credible by more then some Anandtech Forum members.

And those dictionary definitions you posted was a real "source considered credible"....??

If you have a different min and max size then at some point the pagefile can become defragmented.

If the initial size is high enough the pagefile will NOT increase unless it needs to be. And there may be a time when the pagefile needs to be increased in size and that is why setting initial and max the same is just stupid and you do NOT benefit from it at all. Fragmentation of the pagefile does NOT affect performance as I already stated.

You didn't read the guide. DisablePagingExecutive is not a Myth. Well Cacheman is recommended thus why it is in the guide.

DisablePagingExecutive is not in the guide at all and as it may not be a myth it is completel useless and does not do anything Windows does not do on its own. Also Cacheman applys many of those memory tweaks including DisablePagingExecutive and others that are not reccomended. The only things I recall that would be fine to use is Menushowdelay which I have actually found useful.

It obviously does.

Then go build yourself another 1,000 PC's and then you might understand exactly what Virtual memory is...

I find that many authors of tweak guides are the same way as you as Ib have talked to a few and one in particular was arguing with me about Iopagelocklimit. He was saying he has done his research on the tweak and said he has noticed performance increases after using the tweak. It's funny how a registry entry that is ignored by the OS since Windows 200 SP1 can provide a performance increase..

 

[Nothinman]

So why then did Mark Russinovich the author of Win Internals write Page Defrag and post on his site:

For fun? Lots of people code for fun.

Nope you need a source considered credible by more then some Anandtech Forum members.

My point was that your links are no more credible than anything I could create in 10 minutes.

Right, so stupid that by default Windows sets the maximum to roughly what it is recommended to in the guide. Your whole comment makes 0 sense, If you have a different min and max size then at some point the pagefile can become defragmented.

So what? If you use your filesystem at all things can become fragmented, might as well turn off the computer to save everything in an 'OMG optimized' state.

DisablePagingExecutive is not a Myth

No, but it is also completely useless to end users.

It obviously does.

Does not.

 

[jvarszegi]

Originally posted by: Nothinman
The difference has been explained here many times. Yes, MS uses the term VM incorrectly in their UI. The only place they seem to use it correctly is in MSDN articles.

VM, like the name suggests, is the virtualization of memory addresses. Each process sees it's own set of memory addresses, on a 32-bit system there's 4G worth. 2G is reserved by the NT kernel and the other 2G is available for the process to use.

The pagefile is just a backing store for some data, NT requires everything in memory to have a backing store on disk so that it can free up that memory if it ever needs to. Most things can be paged back in from their original respective files (i.e. executables, shared libraries, non-changed files, etc) but any data in memory that's been altered needs a place to go on disk and that place is the pagefile.

The pagefile is just a backing store for some data

It largely does NOT act as a backing store; for something to have a backing store means that it is duplicated in storage, not flushed to storage.

NT requires everything in memory to have a backing store on disk

You've said this before, and it's just as obviously false now as ever. Not only is the page file used to provide virtual memory storage (and not a "backing store"), but NT (and here I assume you mean all Windows 32-bit versions; forgive me if I'm wrong) doesn't require you to have a page file at all. I'm a little rusty on NT 4, but Windows 2000 and XP definitely do not require a page file. Without a page file, how could it run if it required this "backing store"?

any data in memory that's been altered needs a place to go on disk

The alteration actually has nothing to do with needing disk storage.

 

[Zebo]

:::YAWN::: another n00b w/o a clue/

 

[GeneralAres]

We already told you fragmentation of the pagefile has very little affect on performance. This is because the pagefile is NOT read as a whole file (in other words from one end to another). Instead it is accessed in very small chunks. And when I mean very small I mean a few tens of a KB. The pagefile is also not the only file involved with paging such as exe's and dll's. Because of this the head of the hard drive is jumping all over the place anyway so fragmentation has no affect on performance and is irrelevant. Also as Nothinman stated under the most odd circumstances it may increase performance. Think about it. If the head is jumping all over the place and that little chunk in the pagefile that needs to be accessed is closer to the other data being accesses it would be faster. Again this not probably not be noticeable though.

Your dodging the question, why would Mark Russovich the author of Win Internals (which is quoted as one of the sources) write a PageDefrag utility and write: "Paging and Registry file fragmentation can be one of the leading causes of performance degradation related to file fragmentation in a system."

You haven't told me anything other then you read a book on something related and are trying to contradict an Author who is reputable. Your not.

Those dictionary definitions you posted was a real "source considered credible".

And they still are. Virtual Memory Addressing is just that, how Virtual Memory is addressed. Wether it be Hardware or Software. Saying a Page File is not Virtual Memory is nonsense considering it serves no Purpose outside of the Virtual Memory Manager.

If the initial size is high enough the pagefile will NOT increase unless it needs to be. And there may be a time when the pagefile needs to be increased in size and that is why setting initial and max the same is just stupid and you do NOT benefit from it at all. Fragmentation of the pagefile does NOT affect performance as I already stated.

Read above about fragmentation.

DisablePagingExecutive is not in the guide at all and Cacheman applys many of those memory tweaks including DisablePagingExecutive and others that are not reccomended. The only things I recall that would be fine to use is Menushowdelay which I have actually found useful.

Who said it was? I see a comprehension problem here, first you mention the IRQ8 Tweak as NOT being in the guide that actually IS in the guide, then you quote something and don't even read it! I wonder how this translates into the "information" you have "read" to provide your flawed advice?

Then go build yourself another 1,000 PC's and then you might understand exactly what Virtual memory is...

I already have and understand it fine.

I find that many authors of tweak guides are the same way as you as Ib have talked to a few and one in particular was arguing with me about Iopagelocklimit. He was saying he has done his research on the tweak and said he has noticed performance increases after using the tweak. It's funny how a registry entry that is ignored by the OS since Windows 200 SP1 can provide a performance increase..

Great irrelevant comment! Here is one I've talked to many forum members who manage to read concepts wrong and misinterpret them. Then declare everyone else misinformed.

For fun? Lots of people code for fun.

Right :roll:

My point was that your links are no more credible than anything I could create in 10 minutes.

That's nice, now that you claim dictionaries are no longer relevant I will invent my own definition for a word's usage, oh wait you already have! :thumbsup:

So what? If you use your filesystem at all things can become fragmented, might as well turn off the computer to save everything in an 'OMG optimized' state.

Nice way to contradict yourself.

No, but it is also completely useless to end users.

Even if something only improves performance by 1% its useful.

Does not.

Sorry I meant building one properly which includes installing and configuring the OS.


So far you guys have proven that you can reiterate the concept of Virtual Memory Addressing and how it works with the Linux Kernel, fantastic. Give yourself an award. You have also proven that you can reference a source and Author, then condradict his claims when it does not fit your argument. That is about as uncredible as it gets.

 

[jvarszegi]

Originally posted by: Zebo
:::YAWN::: another n00b w/o a clue/

::YAWN:: It's catching. I know you weren't talking about me.

 

[Sunner]

Originally posted by: GeneralAres Saying a Page File is not Virtual Memory is nonsense considering it serves no Purpose outside of the Virtual Memory Manager.

Well, a CPU in itself serves little purpose outside of a computer.
I guess a CPU == a computer then, by your reasoning.

 

[GeneralAres]

:roll:

 

[CTho9305]

And they still are. Virtual Memory Addressing is just that, how Virtual Memory is addressed. Wether it be Hardware or Software. Saying a Page File is not Virtual Memory is nonsense considering it serves no Purpose outside of the Virtual Memory Manager.

No, actually, you can have page files without virtual memory - you just have to manage it in your application instead of having the OS automagically handle it. Before Virtual Memory, if you wanted more space than there was physical memory, you would have to do exactly that. It just so happens that most OSes that take advantage of VM also use pagefiles, because they're a good idea.

quote:
For fun? Lots of people code for fun.

Right <sarcastic smiley>

I do. here is the list of bugs in Mozilla that I've fixed and am working on fixing. I get no course credit or money for doing these - I happen to enjoy programming. Just because YOU don't enjoy computers enough to learn how the actually work, and program for fun doesn't mean other people don't.

I don't have time to respond to the rest of the incorrect replies here.

edit: fixed link

 

[Nothinman]

It largely does NOT act as a backing store; for something to have a backing store means that it is duplicated in storage, not flushed to storage.

It is a backing store for things that were changed in memory. The fact that the data in memory may be cleared to have the space reused by something else is irrelevant. And it is largely duplicated storage, what do you think pagefile reservations are?

I'm a little rusty on NT 4, but Windows 2000 and XP definitely do not require a page file. Without a page file, how could it run if it required this "backing store"?

Have you tried running them without one? Without a pagefile NT 4 would usually BSOD or give other random errors, I believe Win2K and XP are smart enough to create a small one automatically for you if there isn't one but it's been a while since I tried it on Win2K.

Without a page file, how could it run if it required this "backing store"?

Of course it's technically possible (XP's PE is an example, but it took MS until XP to get a well working PE) but it would mean that any data in memory that had been altered would be pinned there since it would have no place on disk to flush it in a low memory situation.

Saying a Page File is not Virtual Memory is nonsense considering it serves no Purpose outside of the Virtual Memory Manager.

Just because pagefile usage requires VM doesn't mean they're the same thing. Again I ask, if I disable the swap on my Linux box here do you really think I've disable VM?

I already have and understand it fine.

That would seem to be false.

Right

You don't think so?

That's nice, now that you claim dictionaries are no longer relevant I will invent my own definition for a word's usage, oh wait you already have!

I never said they were irrelevant. But you have to admit that they weren't written by people who've actually designed a modern OS so there's a huge space for error, and VM is one of the most confusing things in OS design, especially since MS misuses the term all over their UI.

Nice way to contradict yourself.

It was a sarcastic attempt to prove how idiotic your advice was.

Sorry I meant building one properly which includes installing and configuring the OS.

Neither of which require any real understanding how VM works. This isn't 1960, most things the OS does are "magic" that "just works" for most people.

 

[KoolDrew]

Read above about fragmentation.

Read above where myself and nothinman told you a fragmented pagefile will have no noticable performance impact. Do you even know how the pagefile is acessed? If the pagefile was read as one whole file a fragmented pagefile could result in degraded performance, but this is not the case.

Who said it was? I see a comprehension problem here, first you mention the IRQ8 Tweak as NOT being in the guide that actually IS in the guide, then you quote something and don't even read it! I wonder how this translates into the "information" you have "read" to provide your flawed advice?

I read your reply wrong and thought you were saying DisablePaginExecutive was in the guide. I also missed the IRQ8Priority tweak because all I did was search the page for "IRQ8Priority" and siunce it came up with nothing I thought it was not in there.

I already have and understand it fine.

NO you don't...

Right

Alot of people code for fun. And when your dealing with something like defragmenting the pagefile I would imagine it would be fun. Especially for someone who is actually interested in how the OS actually works.

Sorry I meant building one properly which includes installing and configuring the OS.

Installing and configuring the Os still would not give you an understanding of how the OS works. My mom can install and configure Windows XP. Do you think she has any idea what Virtual Memory or the pagefile actually does?

So far you guys have proven that you can reiterate the concept of Virtual Memory Addressing and how it works with the Linux Kernel, fantastic

If you would read you would know it works the same as in Windows.

Even if something only improves performance by 1% its useful.

More like 0% which would not be useful

Your dodging the question, why would Mark Russovich the author of Win Internals (which is quoted as one of the sources) write a PageDefrag utility and write: "Paging and Registry file fragmentation can be one of the leading causes of performance degradation related to file fragmentation in a system."

I don't know. You would have to ask Mark Russovich, but my guess is he made up the description to make people download the program and use it. If he put "Pagfile fragmentation will not degrade performance. I just made this for fun" would you think people would download it and use ? No they would not.

 

[GeneralAres]

I don't know. You would have to ask Mark Russovich, but my guess is he made up the description to make people download the program and use it. If he put "Pagfile fragmentation will not degrade performance. I just made this for fun" would you think people would download it and use ? No they would not.

Now you accusing him of lying? :laugh: I've done enough here. I've proven you contradict the sources you use to try to prove your point. Anyone reading this can see Mark Russovich the Author of Win Internals wrote PageDefrag and can see what he clearly wrote on his page. Anyone with half a brain knows he did not write it for fun. Propaganda is just that and people can see it when you guys write it, sad.

 

[Nothinman]

Someone write a defrag tool for ext2 also, but does anyone use it? No, because it's not necessary and it was only done as proof-of-concept.

Anyone with half a brain knows he did not write it for fun

Anyone with a quarter of a brain would know that you can't make that assertion without asking him.

Propaganda is just that and people can see it when you guys write it, sad.

You're the one who came in here promoting a web page, we just decided to correct you but it's obvious you're not willing to learn something new or admit when you're wrong.

 

[KoolDrew]

I emailed Mark about it. Let's hope he replies.

 

[GeneralAres]

Time to get over yourselves, I checked my library of books. it doesn't surprise me Linux users don't understand definitions properly.

From the:

Microsoft The Windows 2000 Server Operation Guide:
Part No. 097-0002722

Page 698:
Virtual Memory:
The Space on the hard disk that Windows 2000 uses as memory, the amount of memory taken from the perspective of a process can be much greater than the actual physical memory in the computer. The operating system does this in a way that is transparent to the application, by paging data that does not fit in physical memory to and from the disk at any given instant.


Page 298:
Set the Same Initial and Maximum Size:
Setting the pagings file's initial size and maximum size to the same value increases efficiency because the operating system does not need to expand the file during processing. Setting different values for initial and maximum size can contribute to disk fragmentation.


And Mark Russovich's Site:

Pagedefrag

One of the limitations of the Windows NT/2000 defragmentation interface is that it is not possible to defragment files that are open for exclusive access. Thus, standard defragmentation programs can neither show you how fragmented your paging files or Registry hives are, nor defragment them. Paging and Registry file fragmentation can be one of the leading causes of performance degradation related to file fragmentation in a system.

Class Over, Have Fun.

 

[KoolDrew]

The Space on the hard disk that Windows 2000 uses as memory, tha....

First of all that definition is wrong and and actually defines the paging file. Second of all what kind of definition does not spell "the" correctly...

Set the Same Initial and Maximum Size:
Setting the pagings file's initial size and maximum size to the same value increases efficiency because the operating system does not need to expand the file during processing. Setting different values for initial and maximum size can contribute to disk fragmentation.

If the min is high enough then the pagefile will not keep resizing. Also by doing this you keep one of the major benefits of having a pagefile. If Windows needs a larger pagefile it will increase it's size. So setting the min high enough will give you the same benefit you speak of about a static pagefile, but will be better as the pagefile can expand when needed.

it doesn't surprise me Linux users don't understand definitions properly.

It does not suprise me some "tweaker" doesn't understand definitions properly...

 

[GeneralAres]

So now your claiming Microsoft Published books are wrong too? This is laughable. BTW I made the Typo not the book. I've given more then adequate proof, if someone does not believe me they can go get the book themselves, look at the links I provided ect... I know it is hard to admit your wrong but hey thats life. Like I said it does not surprise me Linux users would want to act Elitist.

 

[Nothinman]

it doesn't surprise me Linux users don't understand definitions properly.

Right, because Linux sure doesn't use Virtual Memory :/

So now your claiming Microsoft Published books are wrong too?

Sure, most of their documentation and website use the terms incorrectly. The only places I've seen it used properly is MSDN and I doubt MS has their engineers review every book they publish.

I've given more then adequate proof

All you've proven is that there are a lot of people as misinformed as you on the subject.

 

[GeneralAres]

Right Microsoft's Knowledge Base is wrong, Microsoft Published books for its Server Operating systems are wrong, Mark Russovich who wrote the Windows Internals Book is wrong, Dictionaries are wrong. :laugh: You see a trend here? The only people who are right are some Anandtech Forum members! Please, get over yourselves, it's embarassing.

 

[Nothinman]

Right Microsoft's Knowledge Base is wrong, Microsoft Published books for its Server Operating systems are wrong, Mark Russovich who wrote the Windows Internals Book is wrong, Dictionaries are wrong

Yes and no. I'm sure they know better, but they're using the terminology in a way that people expect. Like when someone calls their computer case the CPU, do you correct them? Maybe once, but if they continually do it you'll just give in and do the same thing eventually. And I'm sure the KB articles are only written with that wording because the UI uses it incorrectly and it's better to be consistently wrong than to contradict yourself in different places.

If you look at the Sysinternal page you'll see that Mark also wrote 'contig', an app to defragment a single file. Why do you think he wrote that one? Wouldn't it make more sense to just use the built-in defrag tool to defrag all of the non-special files? He also put up there NTFSFlp to format a floppy with NTFS, putting an NTFS filesystem on a floppy eats over 25% of the space making an already minute storage device even smaller, do you think he really believes it's a good idea to format all of your floppies NTFS?

And I like how you skipped most of my questions, like if I disable swap (or the pagefile, if you must use Windows) on my machine do you really think I've disabled Virtual Memory?

And you've peaked my interest, I know I'm correct in my usage of VM, but I've emailed Mark to see what he says about his PageDefrag tool and how he would define VM. If/When I get a response I'll let you know and I can forward it too you if you'd like.