Should I go from 8GB down to 4GB?

[yhelothar]

I don't exactly know how uncessary paging works or if it even happens. But I'll tell you what I do notice.
Say I have 8GB of ram, and my paging is on. Task manager might report that I have about 6GB free on average. If I disable my paging file, it'll report I have about 4GB free on average. If that doesn't seem like unnecessary paging that slows down system performance, then I would like to see your reasoning for that. Clearly, it would seem that the memory stored in HD instead of RAM being thousands of times faster can only have negative detriments on performance.

I would really like to see your reasoning on how that can possibly improve performance.

 

[eelw]

I tested this a lot back back at xp times where 4gb became affordable.

Hell, I even did it when I only had 1GB on my P4. Yes, I got low memory errors with multiple programs open. But now with 12GB on my gulftown, no low memory errors at all. Even when I'm video encoding with a VM running.

 

[Blue Shift]

I would take it out, and then you can swap the sticks every week, so each set of memory sticks gets one week on, and then one week off. They love that.

Definitely. Popping them out and in every week also gives the tabs on the slots more exercise and can keep your RAM from getting loose or even falling out.

Remember, happy RAM makes for a happy user!

 

[marlinman]

So if a win7 x64 user with 4Gb or 8Gb RAM and an SSD chooses to use virtual memory, how large should the pagefile be?

 

[Blain]

So if a win7 x64 user with 4Gb or 8Gb RAM and an SSD chooses to use virtual memory, how large should the pagefile be?

1GB min & 1GB max

 

[Nothinman]

I don't exactly know how uncessary paging works or if it even happens. But I'll tell you what I do notice.
Say I have 8GB of ram, and my paging is on. Task manager might report that I have about 6GB free on average. If I disable my paging file, it'll report I have about 4GB free on average. If that doesn't seem like unnecessary paging that slows down system performance, then I would like to see your reasoning for that. Clearly, it would seem that the memory stored in HD instead of RAM being thousands of times faster can only have negative detriments on performance.

I would really like to see your reasoning on how that can possibly improve performance.

The only indication that makes to me is that you've forced ~2G of data into memory that doesn't normally have to be there, which sounds like a net loss to me. If that's data that hasn't been referenced in hours, days, etc then it makes more sense to store it somewhere else so that data that you're actually using can be kept in memory.

So if a win7 x64 user with 4Gb or 8Gb RAM and an SSD chooses to use virtual memory, how large should the pagefile be?

There's no straight answer to that, it all depends on your usage patterns.

It all comes down to unnecessary paging. When there are no memory, OS will automatically defragment data in RAM when insufficient memory occurs. On paper, this is a low operation, but in practice, it is faster than paging itself.

I haven't looked at the VM implementation in Windows, but AFAIK there's no defragmentation going on at all. It may combine larger, contiguous segments of virtual or physical memory allocations in order to save memory tracking them, but I don't believe it actively defragments anything. That would be especially difficult because the virtual mappings should almost never match up with the physical mappings so defragging one will likely make the other worse.

I tested this a lot back back at xp times where 4gb became affordable.

Then your test results are invalid because a lot has changed since XP's release...

I have a partition just for pagefile, but I can feel the impact from using pagefile. It isn't night and day, but enough for me to disable it.

That's most likely what is called a placebo. Paging to one file as opposed to another won't be any different. The only thing you're doing is forcing Windows to keep stuff in memory that you likely don't need and removing a safety net. Both of which are bad ideas.

 

[jihe]

Running Win7 HP 64-bit SP1. I never use more than about 2.5GB of RAM according to task manager, although it does use up to another 5GB for file cache.

I suppose since it's already installed, I should just leave it.

Only makes sense on your laptop since 4g modules consume more power than 2g ones, otherwise leave them in.

 

[tweakboy]

I would take it out, and then you can swap the sticks every week, so each set of memory sticks gets one week on, and then one week off. They love that.

:thumbsup:

 

[StarTech]

I wouldn't recommend it.
I had a 1GB stick of RAM in a Dell years ago and thought it was too much. (Viruses were using half of it to attack my system.) I took it out and broke it in half so that half of the memory chips were on each side and put one half back into the computer.
It didn't work, so I must have needed the full 1GB for it to work.
Lesson learned.

You must have left in the half with the viruses! Otherwise it would have been just fine.

 

[marlinman]

There's no straight answer to that, it all depends on your usage patterns.

Am I then to infer that one can create a page file that's 'too big', in some sense?

 

[VirtualLarry]

Am I then to infer that one can create a page file that's 'too big', in some sense?

Yes. The first-level paging tables, cannot themselves be paged out to HD. So if you create a big enough VM arena, then all of your existing RAM could be consumed by just the first-level page tables, not leaving enough room for the OS kernel non-pagable portions, nevermind the pagable process spaces.

 

[Nothinman]

Yes. The first-level paging tables, cannot themselves be paged out to HD. So if you create a big enough VM arena, then all of your existing RAM could be consumed by just the first-level page tables, not leaving enough room for the OS kernel non-pagable portions, nevermind the pagable process spaces.

I don't think that's a concern since adding pagefile space doesn't add VM, the total amount of VM is locked at whatever is addressable by the CPU and kernel regardless of physical memory or pagefile space available. I think creating too much pagefile space is just a waste of disk space, it shouldn't have any other adverse affects on performance.

 

[Seero]

I haven't looked at the VM implementation in Windows, but AFAIK there's no defragmentation going on at all. It may combine larger, contiguous segments of virtual or physical memory allocations in order to save memory tracking them, but I don't believe it actively defragments anything. That would be especially difficult because the virtual mappings should almost never match up with the physical mappings so defragging one will likely make the other worse.

To simply put it, Windows Memory Management (WMM) will adjust memory usage and distribution with or without VM. This adjustment, is what I called defragment. It is very different than defragmenting HDD.

What I was trying to say is, the process of swapping in and out will have a defragment effect. However, without VM, WMM use other methods to achieve this effect.

Then your test results are invalid because a lot has changed since XP's release...

I said since XP. I have gone through vista and 7 without WM.

That's most likely what is called a placebo. Paging to one file as opposed to another won't be any different. The only thing you're doing is forcing Windows to keep stuff in memory that you likely don't need and removing a safety net. Both of which are bad ideas.

WMM doesn't keep stuffs that are not needed. VM isn't a garbage dump, it is used as soon as any process starts and VMM assigns physical memory according to statistics, and the size of VM contributes into these statistics. In other words, if there is a VM, then data from each and every process will be distributed into VM. Knowing that reading/writing off HDD is far slower than from RAM, performance increases when the size of RAM > the maximum memory usage.

Guess what, if it doesn't improve performance, then people won't do it. What prevents people from disable VM is the crashes. With proper registry tweak, most of these crashes can be avoided. Those that still occurs will go away simply by re-enabling VM without reverting the registry changes. Don't take my words, try it.