Optimal setting for swap file question

The best size of the pagefile is such that it accommodates the peak commit charge plus a comfortable safety margin. Other than that it really makes very little difference. All recommendations regarding pagefile size were designed to provide that. There is no specific size that will provide optimum performance.

The commit limit is RAM size plus pagefile size, minus a small overhead. If the commit charge hits the limit and the pagefile cannot be enlarged bad things will happen, such as application failures. Some applications can handle this gracefully, others cannot. This does not imply that the pagefile will have much usage, it just has to be that size to provide for the necessary commit limit.

The simplest way to have an adequate pagefile size is to set it to system managed. Contrary to many sources, this is not inefficient. Enlarging the pagefile does not require writing any data to the file, only updating some of the files metadata.

From a performance standpoint the pagefile should be on an SSD. Typical pagefile usage are an almost perfect match for SSD performance characteristics. For a conventional drive they could hardly be more wrong.

by doing some tests the computer goes faster / smoother overall with initial 8192 and maximum 12288.

thanks everyone for input.
Gabrio

Hi there
please understand the difference between a page file and a swap file.

a page represents a block of memory that has to be written to some type of non RAM storage device because there's not enough available RAM - what system does is look at currently running processes and sees what memory (real RAM) allocated to that application isn't being currently used -- it then writes those to the paging device and when the application needs those agian it reads the device and writes those back into RAM so the application can continue.

It's hideously more complicated than that but essentially that's what the process does

Swapping is another issue entirely -- an image of the entire application / user space will be written to HDD -- a much lengthier process -- this is usually only done on Multi User systems such as Linux where a server may have 100's of users logged on concurrently but only say 1 or 2 active.

AFAIK windows 10 (a single user system -- not dealing with server versions here) doesn't use swapping - no reason why it should.

The best policy though with paging is always to have sufficient RAM for what you want to do, don't attempt to run too many processes at once and let the system manage it --the algorithms are very complex - tampering with it unless you have a good understanding of OS Internals is likely to yield inoperative or inferior performing systems.

Some very specialized applications might go bonkers when Windows sets the paging but that's probably because these are applications bypassing the Windows API and trying to deal with the actual hardware -- some older gaming / simulation things e.g flight / driving simulators etc to get max performance out of older machines - but these days that should be a no no. Dedicated GPU's will manage intensive gaming apps.

Cheers
jimbo

Steve C said:

I'm talking about the virtual memory / paging file size

Hi there
agreed but some posts mentioned swap which is a totally different issue --just trying to clarify -- your stuff is fine.

Cheers
jimbo