Time to get an SSD.

Dublin_Gunner

Unregistered. wrote: »

I absolutely agree with this - random reads from an SSD will be significantly faster than from HDD. However, if OP does not want it on their SSD, then it should be moved to a hard disk and not disabled all together.

One of the biggest myths in computing - disable the pagefile.

marco_polo

Dublin_Gunner wrote: »

One of the biggest myths in computing - disable the pagefile.

I agree, it is based on the faulty premise that disabling the pagefile somehow keeps you RAM 'full' at all times, which is exactly what Windows memory manager WMM does anyway. What alot of people don't realise is that most available memory is not actual empty, but rather has been marked as being available to other process while still containing the original page data. If nothing else needs the memory the page can still be accessed by it original process if needed.

The only difference is that if WMM needs the memory for something else, with the pagefile disabled, the old cached page data will simply get discarded, rather than getting written out to the pagefile.

Either way if you get a hard page fault the page data will have to be read in again from disk, either the pagefile or the filesystem if disabled.

Dublin_Gunner

marco_polo wrote: »

I agree, it is based on the faulty premise that disabling the pagefile somehow keeps you RAM 'full' at all times, which is exactly what Windows memory manager WMM does anyway. What alot of people don't realise is that most available memory is not actual empty, but rather has been marked as being available to other process while still containing the original page data. If nothing else needs the memory the page can still be accessed by it original process if needed.

The only difference is that if WMM needs the memory for something else, with the pagefile disabled, the old cached page data will simply get discarded, rather than getting written out to the pagefile.

Either way if you get a hard page fault the page data will have to be read in again from disk, either the pagefile or the filesystem if disabled.

I went and explained this on a thread on Guru3D.com all about 'performance results disabling the pf' and I got berated.

Even though I quoted references to one of the Windows engineers basically stating that disabling it will not give you a performance benefit, but will definitely at some stage cause a crash.

8GB RAM with NO PAGEFILE CLUB

Starts off a bit messy 'cause I cracked a joke. But even though I quoted references from an MS dev, they still tried to say it worked, without quoting or posting a single valid benchmark or reference lol

One guy who's advocating the whole charade didn't actually realise the pagefile was controlled by the OS :rolleyes:

Monotype

I'm glad I don't hang out at guru3d. Waaaaay too much bitching and moaning.

Dublin_Gunner

Monotype wrote: »

I'm glad I don't hang out at guru3d. Waaaaay too much bitching and moaning.

The on guy that did post performance links, (from Tom's) actually showed software crashing when the PF was not enabled :rolleyes:

The hilarity!

marco_polo

Dublin_Gunner wrote: »

The on guy that did post performance links, (from Tom's) actually showed software crashing when the PF was enabled :rolleyes:

The hilarity!

I'd take the expert opinion over that of a guy who states (on page 2) "I don't really see the point, because windows uses RAM as virtual memory if the PF is turned off" and "Another thing, when PF is disable;ed, the OS does NOT use the HDD for swap space, it uses RAM."

Err right, so when windows need to swap something out of RAM it swaps it out to RAM?. Genius

Dublin_Gunner

marco_polo wrote: »

I'd take the expert opinion over that of a guy who states (on page 2) "I don't really see the point, because windows uses RAM as virtual memory if the PF is turned off" and "Another thing, when PF is disable;ed, the OS does NOT use the HDD for swap space, it uses RAM."

Err right, so when windows need to swap something out of RAM it swaps it out to RAM?. Genius

Some people know what they're talking about. Others believe everything they kinda read on the internetz.

Serephucus

If you do end up putting browsers and the like on your SSD, there's a very handy tweak that'll store FF's temp cache in RAM instead of on a drive.

FF address bar - about:config, and confirm that you know what you're doing.

find browser.cache.disk.enable - set to false.
right click, new, int, broswer.cache.memory.capacity, and set to whatever you want in KB. I suggest at least 128MB.

The disadvantage with this obviously is that your RAM usage for FF will go up more than usual, but since most comps no-a-days have 4GB+, who gives a crap?

Oh, I've also been running without a pagefile since I got my SSD (32GB Onyx), and my system is snappier, and hasen't given me a single error since.

Unregistered.

Serephucus wrote: »

If you do end up putting browsers and the like on your SSD, there's a very handy tweak that'll store FF's temp cache in RAM instead of on a drive.

FF address bar - about:config, and confirm that you know what you're doing.

find browser.cache.disk.enable - set to false.
right click, new, int, broswer.cache.memory.capacity, and set to whatever you want in KB. I suggest at least 128MB.

The disadvantage with this obviously is that your RAM usage for FF will go up more than usual, but since most comps no-a-days have 4GB+, who gives a crap?

Oh, I've also been running without a pagefile since I got my SSD (32GB Onyx), and my system is snappier, and hasen't given me a single error since.

I did this last year, and the increase in RAM usage is not noticeable.

marco_polo

Serephucus wrote: »

If you do end up putting browsers and the like on your SSD, there's a very handy tweak that'll store FF's temp cache in RAM instead of on a drive.

FF address bar - about:config, and confirm that you know what you're doing.

find browser.cache.disk.enable - set to false.
right click, new, int, broswer.cache.memory.capacity, and set to whatever you want in KB. I suggest at least 128MB.

The disadvantage with this obviously is that your RAM usage for FF will go up more than usual, but since most comps no-a-days have 4GB+, who gives a crap?

Oh, I've also been running without a pagefile since I got my SSD (32GB Onyx), and my system is snappier, and hasen't given me a single error since.

For sure disabling it is not a big deal if you have 4Gb and never get anyway near to full RAM usage at any time.

However by disabling it you are actually reducing your usable ram as processes often reserve more memory that they actuallly need. Because windows will only ever permit processes to reserve as much memory as it can physically provide, (this is called the commit limit which the sum of physical RAM + pagefile size), with 4Gb and no page file you are limiting your commit limit to exacly 4GB.

Looking at my task manager (performance tab) I have 1.44 Gb of physical memory in use, but my commit size is 1.98 GB. With no page file the total usable memory I have left would be just over 2 GB , but because I have a pagefile my total usable memory is still 2.6 GB (Note my commit limit is 8GB 4Gb RAM + 4Gb page file).

With the page file in use Windows can use the pagefile to backup memory which is reserved by processes but is not actually being used (and may never be), keeping physical RAM free for memory that is actually needed. However with no page file that reserved but currently unused memory must be backed up with physical RAM.

Pretend I had only 2GB of memory and no page file with my current commit size of 1.98 GB I would not be able to even open firefox without getting a "System is running low on virtual memory" message even though only 1.44 Gb or my RAM is actually in physical use.

Unregistered. wrote: »

I did this last year, and the increase in RAM usage is not noticeable.

This is because processes work with a virtual address space and are unaware of whether this actual maps to physical or virtual memory. Because of this level of abstraction firefox will behave the exact same regardless of whether there is a page file or not.

Unregistered.

marco_polo wrote: »

This is because processes work with a virtual address space and are unaware of whether this actual maps to physical or virtual memory. Because of this level of abstraction firefox will behave the exact same regardless of whether there is a page file or not.

Aye, but i was referring to the hack of moving the browser cache to main memory. I think the basis for it is that the cache is written to often, therefore it is kept in physical memory.

marco_polo

Unregistered. wrote: »

Aye, but i was referring to the hack of moving the browser cache to main memory. I think the basis for it is that the cache is written to often, therefore it is kept in physical memory.

Misinterpreted what you meant, right you are

Moon54

So the long and short of this is, keep your pagefile, am I right?

I recently moved the pagefile from my SSD to my Samsung hard drive and I
notice my OS drive now has 8GB more free space!

That's gotta be good thing, right?

marco_polo

Moon54 wrote: »

So the long and short of this is, keep your pagefile, am I right?

I recently moved the pagefile from my SSD to my Samsung hard drive and I
notice my OS drive now has 8GB more free space!

That's gotta be good thing, right?

If someone has disabled their pagefile and has never had a single glitch who am I to say they are wrong . I have just never seen any real evidence of a performance benefit to disabling it, and from my (relatively superficial) understanding of how memory management works in Windows nor would I expect there to be.

On the other hand wanting to maximise space or bring the number writes down to an absolute minimum on the SSD are fairly reasonable reasons for moving it off the SSD to a HHD (or even disable it).

Unless you specifically had the size set to 8GB though that doesn't sound like it is all attributable to the pagefile.

Pimp Ninja

Muchos Gracias for all the responses on this.
Ive done some further digging regarding the pagefile, and came across an awesome thread on the OCZ site iteself.

The specific part about pagefiles

Page/Swap File Sizing and Placement

3 options available..
Let the OS manage it,
Turn it Off,
Manage it yourself.
- Initially, let your OS manage it.. then after some monitoring, manage it yourself.


In Windows, turning it off does not mean it will never create one.
If it needs to, Windows it will create a temporary one of it's own choosing.
Moving it to slower storage will result in it's functions being slower.

In Win 7 put it on your SSD.
Most Win 7 operations are reads, most writes being sequential.
These are underpinning arguments to eventually manage the Page File.
Don't let it manage your install.

So, once I get home this evening 3 gig pagefile on my SSD :-D