How much virtual memory with 2 GB RAM ?

[Super1Bat]

Hi

1 - How much virtual memory do i need for 2 GB RAM or i let the Windows decide ?

2 - Do i need virtual memory for 2 GB RAM ? ( i think i don't )

I have Windows 7 RTM 7600

 

[her209]

1. Let Windows decide.

2. Yes.

 

[TemjinGold]

Originally posted by: Super1Bat
Hi

1 - How much virtual memory do i need for 2 GB RAM or i let the Windows decide ?

2 - Do i need virtual memory for 2 GB RAM ? ( i think i don't )

I have Windows 7 RTM 7600

What makes you think you don't?

 

[Billb2]

Run fro a few days, weeks, than check what windows set automatically. Then lock that in.

 

[Super1Bat]

Originally posted by: TemjinGold

Originally posted by: Super1Bat
Hi

1 - How much virtual memory do i need for 2 GB RAM or i let the Windows decide ?

2 - Do i need virtual memory for 2 GB RAM ? ( i think i don't )

I have Windows 7 RTM 7600

What makes you think you don't?

Some one told me

 

[TemjinGold]

Originally posted by: Super1Bat

Originally posted by: TemjinGold

Originally posted by: Super1Bat
Hi

1 - How much virtual memory do i need for 2 GB RAM or i let the Windows decide ?

2 - Do i need virtual memory for 2 GB RAM ? ( i think i don't )

I have Windows 7 RTM 7600

What makes you think you don't?

Some one told me

That some one is wrong.

 

[hanspeter]

Originally posted by: TemjinGoldThat some one is wrong.

It depends on how you define wrong and what the needs are of the system.

 

[Nothinman]

Originally posted by: hanspeter

Originally posted by: TemjinGoldThat some one is wrong.

It depends on how you define wrong and what the needs are of the system.

Perhaps, but the only time you should consider going without a pagefile (which isn't the same thing as virtual memory) is when you've got a very strictly controlled workload and can guarantee that memory usage will never get above a certain highwater mark. Which pretty much just leaves embedded systems.

 

[hanspeter]

With 512MB RAM and and 1.5GB page file, you can commit as much as you can with 2GB RAM and no page file. It all depends on the workload.

 

[Nothinman]

With 512MB RAM and and 1.5GB page file, you can commit as much as you can with 2GB RAM and no page file. It all depends on the workload.

Yea, the Windows commit charge limit would be the same but having no pagefile leaves you with no net in case something goes wrong. Can you honestly say that you can guarantee with 100% certainty that the combined memory usage of all of the processes and kernel on your PC will never go above the 2G mark?

 

[TemjinGold]

Bottom line is, the fact that the OP needed to ask someone means s/he is most likely much better off letting Windows decide.

 

[Emulex]

let windows decide the amount, then take that amount and make it fixed at that size.

too much won't necessarily hurt unless you go gung ho and make a big fragmented file. defrag first if you don't have an offline defrag tool.

 

[Nothinman]

let windows decide the amount, then take that amount and make it fixed at that size.

No, the pagefile isn't accessed in large, sequential blocks so having so making it a fixed size gains you virtually nothing. Just leave it Windows managed.

 

[hanspeter]

There was that static size argument again. Making it static is equivalent to installing the same amount of physical ram and having no page file. That was my point earlier.

 

[Nothinman]

There was that static size argument again. Making it static is equivalent to installing the same amount of physical ram and having no page file. That was my point earlier.

But it's not because the pagefile isn't directly accessible like memory. From the user's perspective, sure it'll mean that the OS will let them have more things running concurrently, however it'll be hundreds or thousands of times slower because the OS has to page things back in from the pagefile into memory in order to use them.

 

[hanspeter]

That wasn't what I ment. It was about having no page file. People seem to accept that RAM + a static page file is okay, but having huge amount of ram and no page file is bad news.

 

[Nothinman]

That wasn't what I ment. It was about having no page file. People seem to accept that RAM + a static page file is okay, but having huge amount of ram and no page file is bad news.

And they're right. Having a static page file is still better than having none.

 

[hanspeter]

You missed the point. 8GB RAM and no pagefile gives you the same commit limit as 4GB RAM and 4GB page file. The first one people screams about, the latter they are okay with.

 

[Nothinman]

You missed the point. 8GB RAM and no pagefile gives you the same commit limit as 4GB RAM and 4GB page file. The first one people screams about, the latter they are okay with.

If your point was just to complain about people, then move along, we already know that people are dumb.

You should have enough memory to handle your workload, period. If you're trying to work with a dataset that needs 8G of memory and you think that you can do it with 4G physical memory and a 4G pagefile then you're stupid. It might end up working but it won't work well and it'll take you many magnitudes of time longer. And you should still have a pagefile regardless of the amount of memory installed because it gives you a safety net should you exceed your expected workload and it gives the OS a place to store data you haven't used so that memory can be used for other, better things like filesystem caching and other processes.

 

[hanspeter]

I am not sure we are talking the same language. Would 2GB RAM and a static 6GB page file be okay with you? Would that 6GB page file be enough of a safety net?

 

[lopri]

I usually set fixed-size pagefiles at the beginning partition of a different physical disk once OS installation finishes. Haven't run into a problem yet.

 

[lopri]

OK I realize I didn't follow the thread after reading the first few posts. Please disregard my above post.

 

[Emulex]

i was merely saying a static page file is not likely to fragment like a dynamic one. i assume you realize the speed of your pc will suck if you are actually needing to come near your limit of pagefile allocation (given that it is reasonably sized).

 

[Nothinman]

i was merely saying a static page file is not likely to fragment like a dynamic one.

The only time a pagefile may fragment is when it grows and that's very unlikely unless you set the initial size way too small and frequently exceed the amount of physical memory that you've got. And fragmentation of the pagefile won't hurt performance much, if at all, anyway because pagefile reads/writes are small and random.