Do you have the Page file (Virtual Memory) on or off?

[ranakor]

Page file is off but meh, 256GB of ram is a nice buffer

 

[Cerb]

There are good reasons to disable the PF, IMO, since its "safety net" functionality is can turn a Core i7 into a 6x86, and the only option you're left with is to turn the PC off. Or, you can get more RAM and disable the PF, at which point misbehaving programs will do what they aught to have done anyway, and crash, leaving everything else alone and running smoothly.

All SSD guides I've read also suggest disabling it 1) to minimize space

Get a bigger SSD and set it to a fixed size.

2) to minimize writes to the SSD

If you have an SSD where that's a problem, you either need more RAM, or to realize that old Indilnix and JMicron controllers were complete garbage, and upgrade.

PS: so do you guys suggest I set it to 1024 MB Max / Min even though I have 16 GB RAM?

I suggest you at least read Pushing the Limits of Windows, and only after it makes sense, make any such decisions on your own (you will also realize that Windows has often made poor default choices wrt PF size, too ). It's probably the most concise description of all of this you will get. The short answer is that your total RAM+PF size needs to be smaller than your peak commit, at any time.

 

[Carfax83]

I've had my page file effectively disabled for as long as I can remember. Even when I ran Windows 98se I had it disabled, same with XP, Vista, Windows 7 and now Windows 8.

As long as you have enough memory, there shouldn't be a problem. I've heard that some programs "require" a page file regardless of how much memory you have, but if that's the case, I've never used one.

 

[theevilsharpie]

Or, you can get more RAM and disable the PF, at which point misbehaving programs will do what they aught to have done anyway, and crash, leaving everything else alone and running smoothly.

You assume that the misbehaving application is the one that will crash in a low-memory situation. That's not a given.

 

[Cerb]

You assume that the misbehaving application is the one that will crash in a low-memory situation. That's not a given.

No, I do not assume that. I can observe that the OS does a good job at killing the right one. When it doesn't...well, TBH, I've yet to have that happen in Windows. On the rare occasions where an application has died due to lack of memory, it has always been the right one killed in Windows. OOM killing is something done way better in Windows than Linux. In Linux, the web browser is typically the first to go, then it will kill the out of control process that it should have killed in the first place.

However, if you have so little RAM that it's not an obvious case of, "this process is being a prick," then you simply cannot reasonably disable the page file, and how much RAM might be too little is highly dependent upon how you use your PC. These days, FI, I have to be a little careful, and do things like only run 1 VM at a time, as I consider upgrading.

The thing is, there aren't other good options. Windows does not offer an option to not use the PF as additional RAM swap space, nor any way to tune how much it may overcommit RAM (TMK, Windows treats RAM+PF as a simple total, and will not truly overcommit, like Linux will). So, it's no PF, and dealing with those limitations for the sake of time, or shutting the entire OS down rudely, as it is reading and writing to the drive.

Now, IMO, what it should do is allow the operations going on with the PF to do so in the background, so that only that process gets slowed down. But, it's not smart like that (nor is Linux), instead freezing everything.

 

[HeXen]

I understand that, but part of life is weighing cost vs need. If you can't afford what you need to do something right then you shouldn't get it until you can.

Apps that remove parts of Windows that MS never intended to be removed aren't a good idea either. I've seen reports of odd things failing because they don't understand the inter-dependencies and MS just assumes certain things exist in a Windows install. Like disabling the pagefile, you may not have had any issues so far but when you do it'll be a PITA to fix. And all because you cheaped out on an SSD? Not worth it to me.



Yes, an understanding of how virtual memory works.

Been doing it for years since the first Nlite Beta's. But then again, I know what i'm doing and don't have issues. No one who uses such is stupid enough to not test out each ISO in a VM first.
BTW, my first SSD was originally $800 for 64gb's. Even my current was $300 as I chose speed over quantity. But you shouldn't assume as to the reason why I customize ISO's cause it sure isn't for the size like it once was. Just pointing out that I do anything but cheap.

 

[Nothinman]

Yes, but we are talking physical files here, aren't we? There's only one pagefile...

We are definitely talking physical files. Virtually all I/O to/from disk is done via paging and virtual memory, the pagefile is really only used for memory pages that have been modified and have no other backingstore on disk. Disabling the pagefile removes 1 source of I/O activity and just adds more memory pressure since the kernel doesn't have secondary storage for those pages any longer.

There are good reasons to disable the PF, IMO, since its "safety net" functionality is can turn a Core i7 into a 6x86, and the only option you're left with is to turn the PC off.

If paging activity causes that much of a performance drop you need to look at your storage. And besides, in a situation that would cause that much pagefile activity there's also going to be tons of I/O to other locations on disk so you're not really helping much by removing just 1 of those locations.

No one who uses such is stupid enough to not test out each ISO in a VM first.

Right. You shouldn't assume people are smart enough for things like that.

BTW, my first SSD was originally $800 for 64gb's. Even my current was $300 as I chose speed over quantity. But you shouldn't assume as to the reason why I customize ISO's cause it sure isn't for the size like it once was. Just pointing out that I do anything but cheap.

You're the one that pointed out price was probably the issue that Sheep221 got one that was too small. I don't see how your willingness to waste $800 on an anemic drive is relevant to him and his situation.

 

[Cerb]

If paging activity causes that much of a performance drop you need to look at your storage. And besides, in a situation that would cause that much pagefile activity there's also going to be tons of I/O to other locations on disk so you're not really helping much by removing just 1 of those locations.

The last time it happened, just after setting up my first post-beta Win 7 install, no, there was no other I/O activity. It was a memory leak of a single application. After a few hours, I cut the power. As soon as I doubled the RAM, off went the PF. It is wrong that the OS allows that to happen. What about other applications that might need some CPU and IO time?

I could look at my storage all day long. It will just sit there, so I don't. Currently, the fastest NV storage is still thousands to tens of thousands of times slower than RAM, and having to use it like that cuts performance much more than even that, just to spending so much time in syscalls and page walks.

The correct behavior is for the OS to not allow the interface to be non-responsive, regardless of what a single user process may want to do (currently, Windows does not have IO priority implemented sufficiently well to implement this, TMK). Anything hogging IO for any reason should be given lowest priority over others. Being responsive to my input is the 2nd most important thing it needs to do (not wildly corrupting data and hard crashing, like in the 9x/FAT days, being 1st).

Currently, disabling the PF is is the only guaranteed method of doing that, and even after going up by more than 100%, RAM is still cheap. A good CPU costs as much as 32GB, FI, and I can count on zero fingers the number of times I've had less money in my RAM than in my CPU. That is a big change from the past, when RAM was too expensive to buy more of than was necessary.

 

[HeXen]

You're the one that pointed out price was probably the issue that Sheep221 got one that was too small. I don't see how your willingness to waste $800 on an anemic drive is relevant to him and his situation.

Cause you said I cheaped out on an SSD. Duh. If you were referring to him then type the reference.