page file on SSD

[flamenko]

I dont get a chance to jump in to the forums much any more but somehow couldn't resist this. I have been intricately involved in the page file issue since the early Vista days and, myself, have been running without pagefile since I installed 4Gb RAM and was playing with it even on 2GB. Vista was an excellent framework however Win7 is that much better because of the way it prioritizes software that you have running.

An example of what I mean is simply that I can run 50 programs in Win 7 with 4GB ram without a pagefile without any problem whatsoever and never a crash.

Using a pagefile on a hard disk is self defeating because, thinking through logic alone, you are slowing the process as the hard drive is somewhat of a bottleneck.

The simple truth however is that, IMHO, pagefile serves absolutely no purpose in systems that utilize 4GB ram or more. I can agree to the fact that, if your system crashes a dump file is available for inspection if the pagefile is on, but would then state that systems just dont crash anymore and the lack of pagefile surely wouldnt be the reason.

I then come to the question of why would I want a process running that is serving no purpose which is the case with many things when running an SSD.

With an SSD, services such as pagefile, hibernation, Prefetch, Superfetch and even indexing seem to provide no benefit further. If it doesnt do anything than why would you want to run it?

Next, I would think that an argument could be presented that turning off moot services can do nothing but benefit your system as there is less opportunity for error in the machine overall.

I can suggest that shutting down above, as I suggest, will help your start times and also provide you with a great deal more SSD storage capacity which we want in our ssds.

As always, I invite return and know that there will be some who think that I am way out in left field with some of the services I shut down. My response would be that right this second I am running 11 ssds on my systed to include two RAID 0 configs and even an SSD powered through satadimm and its as solid as a rock.

In a few days I am going to try and step it up to include an LSI 9260 with 8 SATA 3 ssds installed.

On an amuzing note, I can remember a time when I was totally alone in my views on pagefile and look back to some of the NBR reviews now in amuzement.

Just MHO and hope this provides some benefit.

 

[hanspeter]

#26

With 4GB some can disable it, others cannot. So it is not a final conclusion.

With 4GB plus a page file, you might have more ram available for cache, which can speed up some things. No page file, less cache.

 

[flamenko]

#26

With 4GB some can disable it, others cannot. So it is not a final conclusion.

With 4GB plus a page file, you might have more ram available for cache, which can speed up some things. No page file, less cache.

If I can add, I have asked people through dozens of exchanges to provide any proof that they can find that shows Page providing any sort of performance increase and have yet to have anyone return such.

4GB or more...you are golden.

 

[dualsmp]

If I can add, I have asked people through dozens of exchanges to provide any proof that they can find that shows Page providing any sort of performance increase and have yet to have anyone return such.

4GB or more...you are golden.

Only thing I've encountered is when running chkdsk without a pagefile you can run out of memory, so the pagefile needs to be temporarily enabled. Also some games will run out of memory even with 4 GB, so the pagefile might have to be enabled depending on the game.

 

[postmortemIA]

Only thing I've encountered is when running chkdsk without a pagefile you can run out of memory, so the pagefile needs to be temporarily enabled. Also some games will run out of memory even with 4 GB, so the pagefile might have to be enabled depending on the game.

Consider workstation scenario, where you use dozen or two dozen of tools at same time. And few of them have high I/O and memory footprint, you'll run out of RAM pretty quickly. Games - not much so. There are rare ones that consume more than 2GB, and you don't run more than one.

Anyway, user with SSD has to have mechanic drive for media storage, unless it has no such data. It makes sense to put backup pagefile on such drive.

 

[Dadofamunky]

I don't use a page file. I don't like how windows writes stuff to disk that in background and takes it out of memory. YMMV.


Anyway here is a good SSD optimization guidewhich will cover your concerns
http://thessdreview.com/optimization-guides/the-ssd-optimization-guide-2/

Extremely useful link...

 

[Zebo]

I dont get a chance to jump in to the forums much any more but somehow couldn't resist this. I have been intricately involved in the page file issue since the early Vista days and, myself, have been running without pagefile since I installed 4Gb RAM and was playing with it even on 2GB. Vista was an excellent framework however Win7 is that much better because of the way it prioritizes software that you have running.

An example of what I mean is simply that I can run 50 programs in Win 7 with 4GB ram without a pagefile without any problem whatsoever and never a crash.

Using a pagefile on a hard disk is self defeating because, thinking through logic alone, you are slowing the process as the hard drive is somewhat of a bottleneck.

The simple truth however is that, IMHO, pagefile serves absolutely no purpose in systems that utilize 4GB ram or more. I can agree to the fact that, if your system crashes a dump file is available for inspection if the pagefile is on, but would then state that systems just dont crash anymore and the lack of pagefile surely wouldnt be the reason.

I then come to the question of why would I want a process running that is serving no purpose which is the case with many things when running an SSD.

With an SSD, services such as pagefile, hibernation, Prefetch, Superfetch and even indexing seem to provide no benefit further. If it doesnt do anything than why would you want to run it?

Next, I would think that an argument could be presented that turning off moot services can do nothing but benefit your system as there is less opportunity for error in the machine overall.

I can suggest that shutting down above, as I suggest, will help your start times and also provide you with a great deal more SSD storage capacity which we want in our ssds.

As always, I invite return and know that there will be some who think that I am way out in left field with some of the services I shut down. My response would be that right this second I am running 11 ssds on my systed to include two RAID 0 configs and even an SSD powered through satadimm and its as solid as a rock.

In a few days I am going to try and step it up to include an LSI 9260 with 8 SATA 3 ssds installed.

On an amuzing note, I can remember a time when I was totally alone in my views on pagefile and look back to some of the NBR reviews now in amuzement.

Just MHO and hope this provides some benefit.

Another viewpoint.

http://lifehacker.com/#!5426041/understanding-the-windows-pagefile-and-why-you-shouldnt-disable-it

this issue is as old as page file itself. There is no right or wrong just what works for you.

I personally never use one because of minor caching issues I talked about earlier but there is some who greatly need it such as photo shop users with massive files. Performance as in benchmarking and FPS matter none what you do. In sum there is no right. default is probably best for most people who don't micromanage their system and keep abreast of usesages and capacity.

 

[Zebo]

Extremely useful link...

I would not do all of them...use discretion. Like I like indexing.

 

[Zebo]

If I can add, I have asked people through dozens of exchanges to provide any proof that they can find that shows Page providing any sort of performance increase and have yet to have anyone return such.

4GB or more...you are golden.

Open CCC with it on then reboot open it with it off and let me know how it goes. I cold show you a million other tests of caching fail built into windows but this one is simple.

 

[Rubycon]

Programs that use huge amounts of memory (PS) should be using a scratch disk which can be on its own SSD, for example.

 

[Nothinman]

The pagefile doesn't improve performance; putting it on an SSD wastes the most expensive storage media in your system. Buy enough RAM that you don't need a pagefile at all, and/or put one on a spindle disk if you feel they're really necessary.

Actually it can improve performance because it allows the system to page out data that hasn't been used in a long time making more room in main memory for whatever you're currently doing.

And since with VM even on a 32-bit system you essentially have a need for infinite memory since each process can use up to 2G of VM. Running without a pagefile is like walking a tight rope without a net, sure it's possible but it's a terrible idea.

Programs that use huge amounts of memory (PS) should be using a scratch disk which can be on its own SSD, for example.

No, they should be using VM like any other app and letting the OS manage it for them. Each app implementing it's own memory management would be a total nightmare. PS only does that as a bandaid for the terrible memory management on Mac OS before OS X and Adobe never saw fit to fix it.

 

[hanspeter]

No, they should be using VM like any other app and letting the OS manage it for them. Each app implementing it's own memory management would be a total nightmare. PS only does that as a bandaid for the terrible memory management on Mac OS before OS X and Adobe never saw fit to fix it.

Without such a scratch disk, the application would not be able to allocate more than 2GB in a normal 32bit setup. Sometimes you need more than that. PS can allocate more than 2GB by using the harddrive as storage.

 

[Voo]

Without such a scratch disk, the application would not be able to allocate more than 2GB in a normal 32bit setup. Sometimes you need more than that. PS can allocate more than 2GB by using the harddrive as storage.

And Nothinman's whole point was that not every application should roll out it's own memory managment but let the OS do its work. The photoshop scratch disk is exactly the same as the page file (at least conceptionally, for the actual implementation all bets are off, but it hardly is better..).
Well in theory they could use the Kernel address space for themselves and distinguish somehow between the two, but I extremely doubt they did that - pretty much work, who knows what kind of complications and they had enough problems getting the relatively straightforward 64bit transition done..

 

[hanspeter]

And not every application does it. PS does it, because 2GB VAS is what a 32bit app normally gets. If PS needs to handle image data way beyond that, it would fail without a scratch disk - that is the point.

 

[Voo]

And not every application does it. PS does it, because 2GB VAS is what a 32bit app normally gets. If PS needs to handle image data way beyond that, it would fail without a scratch disk - that is the point.

You would first have to show somehow that PS internally uses a own pointer implementation >32bit and then maps those into the given 2gb address space.
Considering that Adobe had pretty large problems to get PS 64bit ready, I extremely doubt that (if they already have pointers > 32bit, where was the problem to use not 3x but 64 bits for it and change the mapping? Should make it easier considering that mapping would become more or less a nop).

But if they do it, then yes that would make their implementation valid since it fulfills a different purpose than the page file.

 

[hanspeter]

They are not using memory pointers above 2G. They are resuing the address space (swapping data in/out). It is the reusing that Windows cannot provide.

 

[flamenko]

Running without a pagefile is like walking a tight rope without a net, sure it's possible but it's a terrible idea.

Same jiberish we have been listening to for years. Pagefile was great for XP and it served a purpose but there is abs no purpose for it, even in a 32 bit system when ample RAM is available.

If one wants to claim increased performance, please provide supporting references.

 

[Voo]

If one wants to claim increased performance, please provide supporting references.

Use a program that uses lots of RAM and leaves lots of dirty pages. Start other program that needs RAM. See how scheduler pages out dirty pages and increases the working set of the new process.

 

[flamenko]

I can repeat myself if you like but, "Show me the money!"

Use Windows 7 and bench with and without pagefile. I am not trying to be 'forthright' but I have heard such for years. The only thing I still have seen is that no performance benefit is apparent with pagefile on and especially in typical user scenarios.

I even had someone once tell me it was required in Photoshop 7...but didn't realize I was using that very program.

 

[Rubycon]

PS USED to complain about it but current versions don't.

The default pagefile size is quite outdated though. Install Win7 on a system with 128GB of RAM with a 160GB SSD and you will cry when you have no disk space for programs! :awe:

 

[flamenko]

Wait 128GB of RAM???? I always suggest a simple test to win people over. Shut down pagefile and then click on any programs until 50 are running simultaneously and be amazed. Many of my programs include Ofice, Acrobat, Photoshop and so on... I will run multiple instances of many including songs and video.

Yes...video will get choppy but it will with Pagefile on as well. IMHO it all comes down to Windows prioritization of active programs.

 

[deimos3428]

Actually it can improve performance because it allows the system to page out data that hasn't been used in a long time making more room in main memory for whatever you're currently doing.

Not so. The pagefile can buy time, but any system that would benefit from a pagefile has a problem to solve, not one to ignore.

By stating the above you're assuming a system that is already RAM deficient. Ideally you'd want to fix the root cause if possible: provision more RAM (or move applications to another machine) and/or figure out why programs are consuming more memory than they rightly should if that's the case.

A pagefile should be considered a last resort when you can't add more memory, not a safety net.

 

[Rubycon]

Can a pagefile go on a removable disk? Get a cheap 8GB flash drive and you're all set.

 

[Nothinman]

Without such a scratch disk, the application would not be able to allocate more than 2GB in a normal 32bit setup. Sometimes you need more than that. PS can allocate more than 2GB by using the harddrive as storage.

On Windows with AWE it can. It can't all be addressed at once, but it can definitely use more than 2G of VM even on a 32-bit system. SQL Server has been doing this for a very long time.

Same jiberish we have been listening to for years. Pagefile was great for XP and it served a purpose but there is abs no purpose for it, even in a 32 bit system when ample RAM is available.

The problem is that it's impossible to have enough memory to cover all cases. On a 32-bit system each process gets 2G of VM, so you would essentially need an infinite amount of memory to cover every possibility.

Not having a pagefile forces the system to leave every modified page without a backing store in memory which wastes memory. If you haven't used your browser in X amount of hours because you're playing a game, wouldn't you rather the memory claimed by that browser could be paged out to make room for the game? If you already have enough memory to handle both then you don't have to worry about the pagefile being used, so why remove it?

I can repeat myself if you like but, "Show me the money!"

You're the one arguing in favor of people configuring their system in a potentially unstable manner so the burden of proof is on you. Please prove that not one detrimental effect will be had by people disabling their pagefile.

By stating the above you're assuming a system that is already RAM deficient. Ideally you'd want to fix the root cause if possible: provision more RAM (or move applications to another machine) and/or figure out why programs are consuming more memory than they rightly should if that's the case.

And because of how VM works every system is potentially memory deficient. You can try to curb it's use and sometimes control individual process' memory usage, but you can't account for everything that might happen. Leaving a pagefile there just in case is just smart.

A pagefile should be considered a last resort when you can't add more memory, not a safety net.

Obviously building an overcommited system from the start is dumb, but so is disabling the pagefile when it provides a safety net and costs you nothing.

Can a pagefile go on a removable disk? Get a cheap 8GB flash drive and you're all set.

Not on Windows, MS specifically disallows that AFAIK. And those cheap flash drives are usually slower than a decent spindle drive so it would be worse anyway.

 

[hanspeter]

On Windows with AWE it can. It can't all be addressed at once, but it can definitely use more than 2G of VM even on a 32-bit system. SQL Server has been doing this for a very long time.

Of course there is AWE, but that is something else and something a bit more special.

Also, AWE is physical memory only. Most people have access to more harddrive space than they have physical memory.