Virtual memory question

[ingeborgdot]

Do you recommend adjusting virtual memory? If so, how do you best recommend doing it?

 

[thegorx]

sure why not

the easiest way is to adjust it through windows then benchmark or test an app
to see if it runs any better
probably not but we never learn by sitting on our hands

you can always set it back the way it was despite any warning to the contrary

 

[KoolDrew]

Virtual Memory and the pagefile are two different things. For size I recommend the initial to be at least 4x actual PF usage when running your most intensive applications, and the max at least 2x that number.

 

[Fresh Daemon]

Windows is actually pretty good at running the swapfile by itself, especially for desktop/gaming useage. I would leave it alone. I have benchmarked various swap configurations and found no noticeable performance gains. The only scenario in which I think swapfile tuning would make sense is for a server.

 

[AlucardX]

just leave it alone, this topic has been beat to death over the years. or set the min=max to about 1.5xTotal_RAM.. i do the latter so my page file is always one contiguous chunk on the hd.

 

[Nothinman]

or set the min=max to about 1.5xTotal_RAM.. i do the latter so my page file is always one contiguous chunk on the hd.

Don't do that, it's a waste of disk space and buys you nothing.

 

[bsobel]

Originally posted by: AlucardX
just leave it alone, this topic has been beat to death over the years. or set the min=max to about 1.5xTotal_RAM.. i do the latter so my page file is always one contiguous chunk on the hd.

The page file isn't read serially, so having it contigous generally doesn't matter.
Bill

 

[TGS]

Here it is guys...

FIN

Edit: The real answer to the page file quandry is to just put more RAM into the system. If you are doing extensive paging, you need more RAM, period.

Changing the Page File settings will not give you tangile benefits, other than the use of a particular space of a volume. The real benefit to be had with changing the default configuration would be to move it to a seperate volume to lessen the IO contention between a Windows OS or Application volume.

 

[Nothinman]

If you are doing extensive paging, you need more RAM, period.

Not true at all. Paging is a necessary operation, it's how data is read/written to/from disk. Every time you start a binary hard faults are incurred while the binary is paged into memory from disk and soft faults are incurred while shared libraries are mapped into the processes address space. This notebook has an average of 247.11 faults/s and it's got more than enough memory for what I have running, in fact no swap space is used at all right now.

 

[bsobel]

Originally posted by: Nothinman

If you are doing extensive paging, you need more RAM, period.

Not true at all. Paging is a necessary operation, it's how data is read/written to/from disk. Every time you start a binary hard faults are incurred while the binary is paged into memory from disk and soft faults are incurred while shared libraries are mapped into the processes address space. This notebook has an average of 247.11 faults/s and it's got more than enough memory for what I have running, in fact no swap space is used at all right now.

I took his comment as, if your actually using the swap file (not swapping from binary backed images) the best solution is more memory. He's right, your being, well yourself (big )

Bill

 

[KoolDrew]

or set the min=max to about 1.5xTotal_RAM..

All this does is remove that "safety net," and having something that large on your OS partition is unnecessary and could increase seek times to other files on the same partition.

i do the latter so my page file is always one contiguous chunk on the hd.

Unless the pagefile is extremely fragmented it will not effect performance in any tangible way. This is because Windows will never read or write more then 64KB to the pagefile and they are almost never in sequential 64KB chunks. Sp after one such buffer the heads will have move, regardless if the pagefile is fragmented or not. There would also be IOs to many other files, which is why fragmentation of individual files means very little.

 

[Nothinman]

I took his comment as, if your actually using the swap file (not swapping from binary backed images) the best solution is more memory. He's right, your being, well yourself

I know, in the common case he's right. But I like being pedantic on this particular subject so many people think they know what they're talking about but in reality they don't even know what all of the terms mean. And if someone decided to poke around in perfmon or something and count the paging activity they would get much higher numbers than they expected and worst case scenario end up buying memory for no real reason.