Windows 7

VulcanB2 wrote: and why removing it is a daft idea. WHY? Why is it daft? Please cite technical reasons, not just "they said so".

Jeez are there no mods about with an itchy delete finger

Sure not using a page file/swap file/virtual memory/swap partition may be fine if you know that you have enough memory to run the applications you're going to run, but to be honest I see no point in disabling the page file, I trust the operating system to make the best judgement on things that I can't control (can you accurately tell exactly how much memory all of your application will use at any given time?).

Nh - 'ts ne.

Bst rgards,
Vlan.

and forcing Windows and its memory manager to make decisions on how to manage the memory in a restricted way. There is nothing difficult to understand here, in fact, it's common sense.

Did you ignore my description of the memory manager? Exactly how does it restrict its ability to manage anything? Please stop being vague and start getting technical.

The memory manager tries its best, using algorithms, to keep resources available to you.

Please refer to my earlier post. It tries to free up memory when required by looking for data it can move out that is the same size or slightly larger, but no more than necessary.

It is as simple as this:

[1024][4096][512][23][56]

I want 2048 of memory

[1024][2048][FREE SPACE][23][56] [SWAP=4096]

My app does its thing.

Without the swap?

[1024][4096][512][23][56]

I want 2048 of memory

OUT OF MEMORY. Something does something. Who knows what as it is random.

Additionally there are certain Windows code pages and certain pages from whatever applications you are running that are best left cached in that RAM.

Yes, like Kernel memory. If you understood my previous post, you would know this is called NON-PAGED MEMORY. It an't moving no matter what.

Backed stuff, i.e. code, application stuff etc. that exists as files on your drive can simply be dropped when the mm decides it is advantageous to do so because, being backed, the mm can load it back from those files when required.

See, you made the same mistake again. IT DOES NOT DO THIS. Refer to my earlier posts.

the best thing would be for everything you are using to be in RAM with no need for anything to be paged out but unless you have gobs of RAM

Incredible. We're getting somewhere at last!!!! You now realize it is for cases where THE SYSTEM IS OUT OF PHYSICAL MEMORY.

The page file itself is used by the mm to page out those memory pages that have changed but are not existent anywhere other than in RAM

Not true. It will page out all memory necessary to complete a task, up to the limit of the maximum size of the swap file (ignoring race conditions talked about earlier with respect to dynamic swap files).

If it's changed and doesn't exists on disk then the mm can use the page file for temporary storage.

Not true. If it is pageable memory, it will move it if required. Again, you fail to recognize it does NOT care what that data is. It is just a bunch of 0s and 1s occupying space it wants to use.

As a consequence program code that really should be cached for better performance has to be sacrificed instead since it is backed and so it is paged out (note, paged out does not mean copied to the page file).

Huh?? This makes no sense. Specifically:

note, paged out does not mean copied to the page file

If it isn't copied to the page file, by definition the data is not paged out, but dumped. Freed. Deleted. Overwritten. Trashed. Ignored. Removed. Corrupted. Whatever. Anything but kept in a usable state for later.

Therefore I really can't see why you can in any way not see that there is a potential for a performance penalty.

See, I think you are mixing terminology, and getting confused by it.

Cache is data not in use, but *might* be needed, maybe (possibly never). The whole point of cache is to move that data to faster storage. Hard disk to RAM, then RAM to L2 cache, then L2 cache to instruction/data cache.

The system can cache all it wants - if I use something else, it makes NO DIFFERENCE.

When you initially start using the computer, all programs start out equal. They are all on the hard disk, and none are anywhere near memory.

The only time caching makes a difference is if I launch Word, write a letter, quit, and come back later (assuming I didn't restart). The next time I launch Word, it will already be in memory and it will start very quickly.

FYI: you don't need a swap file for this. Yesterday I had Acrobat Reader open, FS9, GIMP, FireFox, Media Player, etc.. open. I could close of them. If I statred them up again, the system did NOT load everything fresh off the HD - they all ran straight out of memory (cache). It didn't make one access to the HD to do this.

Ultimately, unless you have large amounts of RAM and always restrain yourself from going anywhere near memory limits then there is always the potential for performance hits caused by unneccesary paging out of commonly used program and OS code because the memory manager has no choice because it no longer has the use of the page file.

Everything you said was correct right up the the words "because it no longer has". You have this backwards. A system with little memory and a paging file will ALWAYS be slower than a system with lots of memory and NO paging file. Tell me how a HD with a read/write speed of 80 Mb/sec will be quicker than system memory with a read/write speed of 1.6 Gb/sec???

Your understanding is flawed.

What stuff will it now not have to be waiting for to move between the 'swap' file and memory without a page file that it would with?

Anything that is in the swap file that it requires. If it involves disk I/O, it WILL be slower. Basic physics. 80 Mb/sec is slower than 1.6 Gb/sec. Nothing on this planet will change that.

The page file gives it more choice on how to manage the memory, not less.

Manage what???? What is there to manage? Space? That is what 8 Gb of memory is for.

Any physical memory location takes exactly the same time to access as any other memory location.

You have a limited amount of memory, how can you possibly think that cutting off a source of potential storage (granted, a lot slower storage, but storage at least than allows the mm to juggle things around better) can possibly be better than leaving it use it as it sees fit?

You're making progress. How is 2.6 Gb of memory "limited" if you never use more than 1.5 Gb at any time? The same goes for most people, except graphic designers and architects for the reasons I mentioned earlier.

If I have a 2 L bucket and only ever fill it with 1 L of water, there is little sense pouring 500 ml of it into a small container just in case, is there? Yet this is exactly what Windows does. You can prove it by monitoirng HD activity with the swap file enabled, vs. with it disabled.

By disabling the swap file, you prevent Windows doing this. Instant increase in performance. Not only in terms of access to data in memory, but also the HD. Instead of the HD flicking between swap and loading data or a program, should the system be trying to do botyh at once, the HD can do its job of just being a store of user data and programs, thus speeding up the job further.

Every time that HD head has to move from one part of the disk to the other, it takes 800 ns. Compare the same operation to memory, that takes 1 ns. No competition.

Solved by adding more memory? Not a feasible solution in most cases and ultimately a waste of time and money since you have a technology available to you, built into the operating system, a core feature of the code that is designed to prevent that very problem, so why not use it.

Memory is so cheap now there is no excuse NOT to do it. The swap file was designed back in the days when 16 Mb of RAM cost £100 (~£300 in todays currency). It is slower, and will NOT prevent you running out of memory!!! Just read all the FS forums on the dreaded "OOME". This is exacerbated by the use of dynamic swap files, due to the race condition I was talking about.

If the swap file exists, Windows will try to use it whether it needs to or not. Just use yur favorite file monitoring program (or write your own to intercept read/writes to the paging file) and watch.

I do some graphical 3d rendering in my spare time - add a few models, scenery, large textures etc. etc. and click render.

I've seen those. They're excellent.

The image I produced for my son for Christmas was rendered 1:1 at 3622 x 2834,  16" x 12" (larger actually since it was for a canvas print and I wrapped it around the borders). It took several hours to calculate the lighting and ray tracing etc.  Without a page file I'd have had no chance...

See - now we're getting specific. You know you use an app with huge memory requirements, yet you are happy to use the swap file, a storage device some 800 times slower than main memory? In your case, you are right you can't get away without using it.

3622 x 2834 x 32 = 328,471,936 Mb of RAM (at least). Still well inside 2.6 Gb. Let's allow 500 Mb of the development environment you're using, and another 500 Mb for temporary storage (remembering that ray tracing is primarily heavily computational rather than heavily storage orientated). Allowing 256 Mb for the rest of the OS, that is 1.6 Gb or so of memory usage total. Wow - we still have 1 Gb to go.

who do you think contribute to the tech net articles...?

As I wrote, he is hiding behind his MVP badge. It is his site, he wrote it. MS aren't behind anything. I didn't see any links to any articles from his site - only his comments. Note also that storage manangement is not nearly the same as understanding the ins and outs of operating system design and memory management. Storage management likely means he studied how to manange an iSCSI array from Windows Server 2008 using a GUI, not how a microprocessor manages memory - that is far beyond the scope of an MVP.

Since the Microsoft default is to set a system managed page file I think that means that MS think Windows works best with the page file enabled.

Oh dear... and I suppose the default MS security settings are the best, too? Don't ever work for HM Government!!

Best regards,
Vulcan.

built into the operating system, a core feature of the code that is designed to prevent that very problem, so why not use it.

 

 

Oh dear... and I suppose the default MS security settings are the best, too?

 

If it's the settings I think you mean, they work for me, never had an issue. I’ve also never disabled the swap file and no issues thanks. I think you have issues though don't you Pointy? 

VulcanB2 wrote: It has the option of using the page file, you take that option away. You have restricted its ability to manage the memory. How? You keep repeating this but never say how?

It doesn't take any knowledge of this subject to use common sense

Uhhmmm.. I'm done with this conversation. Common sense? Hmm - common sense is knowing that if you walk in front of a truck doing 60 MPH, you will get killed. Common sense doesn't enter into it here.

You are just plain avoiding answering my question now.

It can always drop the stuff that already exists on the drive (backed store)

Where do you get this rubbish from???????? WHAT IS YOUR TECHNICAL SOURCE FOR THIS BAD INFORMATION????

that's your mistake

No mistake. Re-read the question.

Please don't get involved in this discussion, unless it is to add technical reason to blow my description and knowledge out the water.

I can only summize that the reason MM is avoiding answering my question and has failed to produce any technical articles countering my position, either in whole or in part, means I am indeed correct in my assertions.

Seeing as this thread is now playing like a broken record, I'm done.

Best regards,
Vulcan.

VulcanB2 wrote: OK - I've given you a chance to kill my arguments. Now I'm going to kill yours.

Yep, keep believing that...

The link's you've posted are mostly irrelevant to the page file

No they're not. To understand the page file you need to understand memory management in general.

Even the links you've posted disagree with what you're saying

OK then - quote my error, then quote the correction from those links. You obviously spotted something - let's see it.

assuming you have unlimited memory is a big mistake

No....... you obviously don't understand the point of me saying this, which is why you are saying it is a "mistake".

When you conduct experiments, you assume a set of conditions, no? Are you not familiar with scientific method? Assuming an unlimited memory space is one condition for the experiment.

Are you lot really this ignorant or are you just acting that way?

As I say, if you spotted an error - quote it. I even provided links for you!

Vulcan.

VulcanB2 wrote: Are you lot really this ignorant or are you just acting that way?

I always act that way, because I'm a twunk and argue about anything