Hard Drive Free Space Shrunk After RAM Install?

[ZeR04U]

Hello,

I just installed 16GB of RAM (up from 6GB) and my hard drive space shrunk by the exact 10GB. Is that normal? I seem to recall it is but just can't remember.

Hibernation is powered off and there is no hibersys.fil which I believe would be the only culprits so left scratching my head.

EDIT: NVM, it set my paging file size to 16GB (must've been prior at the 6 or something). Anyway, should I disable page file or what's a good range with the memory I have? Right now it is system allocated and using 16GB.

 

[Elixer]

You can make it a fixed size of 1GB (set both min/max to 1024)
Some programs still need that, no matter how much system RAM you have.

 

[ZeR04U]

Thanks Elixer, this is what I was reading on other sites as well.

I read some guides but sometimes I like hearing it from users and forums can better explain what's going on. Is this the right thinking?

When I launch an app the hard drive transfers the app to RAM which access it much faster. Down the road if I am doing something RAM intensive, it writes it to a page file and frees up my RAM to do more pressing matters. This would obviously be important if I had a small amount of RAM so I could dedicate that RAM to gaming, benchmarking, etc. However, if I had a large amount of RAM, there's little chance that the RAM will be completely full (as in the case of 16gb) so it's much better to keep a majority written to the RAM. Obviously I'd want to have SOME page file just in case it got completely used up at some point. Completely disabling could cause system instability where everything is reading from the hard drive.

However, for users with excess and/or large amounts of RAM, it's better to save the extra space on your HD for games/apps/etc. if you're somewhat limited on space (I only have an 80gb SSD at the moment, don't feel like installing and formatting a new one) considering that your RAM will rarely be fully utilized even under heavy load.

That said, if your RAM stores the data for later use, when does it get released? When the app itself is closed?


For example, I have firefox and various programs open utilizing a chunk of RAM. I'm also running some other programs. Then I launch a game. At this point only 11GB of my RAM is utilized. I close the game, and let's assume now 3GB of my RAM is utilized (game took 8GB let's say). Now opening back up those apps (various programs/firefox/whatever) is now much faster then if I had my RAM tapped out and utilizing the page file to transfer it back to RAM.

Is that the right thought? And what programs still need it?


Anyway, sorry for the long text, like to read up and try to understand before just posting "What is it?" so any education from an experienced techy would be appreciated.

 

[Elixer]

Well, the more RAM your system has, the more things it can keep in memory at the same time.
That memory is then free'ed when the program exits--well, usually, if the program is badly written, it can linger on, but that is way outside the scope of this.

Suffice it to say, that anything that is memory intensive, will love the extra overhead of being able to use 16GB.

For most people, that is overkill, since most everything they do will never hit over 4GB.
If you are into video, then you want gobs of memory, and 16GB is the starting point.

 

[sxr7171]

Windows will do that to page file. And for people running SSDs with large amounts of RAM it's a very big waste. I set mine to 2GB. Might test 1GB. Could maybe get away with 512MB.

And yes, a gaming PC will never use 16GB. I don't think I've even used 8GB ever.

 

[code65536]

Anyway, should I disable page file or what's a good range with the memory I have? Right now it is system allocated and using 16GB.

All depends on your peak memory commit. If your physical RAM is enough to cover the peak commit plus a few GB of extra headroom, you're fine with disabling it completely. Some people will say that disabling the page file will cause problems, but that's incorrect and has been incorrect since, eh, the 90's. The only thing you lose if you disable the page file are crashdumps from kernel bugchecks (BSODs) (this is because there's no guarantee that the file system driver is even functional when the kernel crashes, so crashdumps are written blindly to disk, and the only safe place to do something that is the page file).


The more RAM you have, the smaller page file you need. So it seems counter-intuitive for Windows to scale the page file up as the RAM increases. But it does make sense from an OS design standpoint--setting the page file such that RAM plus page file equals a target means that it needs to somehow predetermine a target, and what may have been a reasonable target when the OS was released might no longer be years down the road. Whereas if it assumes that people will increase the RAM as memory requirements for applications increase, then the page file will scale up along with rising application requirements too. That breaks when the user increases the RAM to a point where the it can meet all the memory requirements without help from a page file, but it's a safer assumption to make than a pre-ordained target, and it's not the kind of problem that affects the vast majority of mainstream users.

 

[VirtualLarry]

The more RAM you have, the smaller page file you need. So it seems counter-intuitive for Windows to scale the page file up as the RAM increases.

This is not true. A pagefile is a performance optimization, for the common case. It therefore makes perfect sense that the pagefile size increases, as your RAM size increases. A pagefile is a place to put stale memory pages, in order to free up RAM, for more performance-oriented uses, like immediate foreground application need, or file-caching optimizations.

But it does make sense from an OS design standpoint--setting the page file such that RAM plus page file equals a target means that it needs to somehow predetermine a target, and what may have been a reasonable target when the OS was released might no longer be years down the road. Whereas if it assumes that people will increase the RAM as memory requirements for applications increase, then the page file will scale up along with rising application requirements too. That breaks when the user increases the RAM to a point where the it can meet all the memory requirements without help from a page file, but it's a safer assumption to make than a pre-ordained target, and it's not the kind of problem that affects the vast majority of mainstream users.

Total active working-set should still be lower than total RAM installed, otherwise, a pagefile shows its ugly side: in the worst case, a pagefile can start thrashing the disk, causing a major bottleneck (less so with SSD, but still a bottleneck).

 

[code65536]

This is not true. A pagefile is a performance optimization, for the common case. It therefore makes perfect sense that the pagefile size increases, as your RAM size increases. A pagefile is a place to put stale memory pages, in order to free up RAM, for more performance-oriented uses, like immediate foreground application need, or file-caching optimizations.

No, a page file is an enabler in cases where physical RAM is insufficient. There is no performance to be gained from paging out stale memory--it's actually quite costly as it incurs disk I/O. If the immediate foreground application cannot fit into the memory without having some of it be paged out, then it would crash without that page file. A page file prevents that crash, thus enabling the program to run, at the cost of performance. Even as a way to reallocate memory to serve as a disk cache, it makes dubious sense, as you're incurring a write cost (which is more expensive than reads) and a future read cost to potentially save on some other read costs.

 

[tonyfreak215]

I've disabled my page file since I built a PC with 4 gigs of ram. (back on Win XP)
iTunes is the only program that has ever given me issues.

 

[VirtualLarry]

There is no performance to be gained from paging out stale memory--it's actually quite costly as it incurs disk I/O.

Again, not true. You should study NT's VM design. Especially the part about pre-emptive paging. Again, it's a performance optimization in the common case.