Windows Page File and Skyrim Load Times

[Kuldebar]

My Specs:

• Intel i3570k,
• Radeon HD6850 1 GB VRAM
• 8 GB System RAM, Windows 7/64 bit
• SanDisk 120 GB SSD
• 1 TB Standard HDD

 

 

I normally allow Windows to manage page files, default settings, etc. But when I do something on my PC, I usually tailor my whole PC around that something, in this case a game called Skyrim.

 

When I installed an SSD as my primary drive (OS/game is installed on it) I did turn off the page file located on the SSD initially and left the other in place on the conventional drive. Then over the course of starting to play Skyrim and tweaking, I dispensed with page files completely most likely after reading some compelling arguments supporting such a move.

 

Everything was fine.

 

Yet sometime in the period of intervening months, and after the Sheson Revolution and inherent ENBoost stability, I turned paging back on.

 

It took me a while to notice that my load times (leaving Whiterun to the open world was like 1 minute or so!) started getting longer. My Papyrus was clear, no massive scripts and no silly uGrids or ridiculous textures sizes, WTF was going on?

 

Then I thought about page files and what I had originally suspected about them. <eureka!>

 

Sure enough, it appears windows paging to disk is slowing my load times dramatically.

 

 

 

[Octopuss]

I found an interesting article I read several months ago: https://www.tweakhound.com/2011/10/10/the-windows-7-pagefile-and-running-without-one/

There are some links at the bottom that are worth reading as well.

 

I really don't see any reason to turn off page file on a SSD (with the amount of data that go in and out of it, it hardly affects performance at all).

 

Bottom line is, if you are getting 1 minute long loading times, there's a massive problem somewhere, and it's not in activated or deactivated page file.

[Kuldebar]

I found an interesting article I read several months ago: https://www.tweakhound.com/2011/10/10/the-windows-7-pagefile-and-running-without-one/

There are some links at the bottom that are worth reading as well.

 

I really don't see any reason to turn off page file on a SSD (with the amount of data that go in and out of it, it hardly affects performance at all).

 

Bottom line is, if you are getting 1 minute long loading times, there's a massive problem somewhere, and it's not in activated or deactivated page file.

Well, turning off paging resulted in load times of 20 seconds or less in the areas that were taking over a minute before. My guess is that windows was swapping data during loads when it was not truly necessary, I certainly wasn't running out of memory.

 

Page files take up a decent amount of drive space, and SSD disk space on my budget anyway is very limited, so a page file on the conventional drive just slows things down too much and completely mitigates the advantage of SSD in some respects.

 

So, when I had a page file on my SSD, I had no issue EXCEPT my drive had less available space, this is how the problem went undiscovered because just a week or so ago I turned paging off for the SSD (don't need a page file on each drive, after all).

[Octopuss]

Sometimes completely WTF things happen...

But with 8GB of RAM there shouldn't be much pagining going on at all. I mean I don't suppose you have 30+ programs running when playing Skyrim. The thing is, some paging always takes place, but it really is super small amounts of data (and especially if you have plenty of free RAM)  - I think we're talking about a few MBs at most here or something similar.

 

With 120GB SSD I can understand the capacity concern. But you can always just limit the size to 2GB or something, just to be sure. It looks like _some_ page file is better than none (there are some rare cases of programs that flat out refuse to start when a page file isn't present for example), and if you don't have crap ton of stuff running, it's usually perfectly enough.

 

I still don't get how having the file on the other disk would cause performance degradation of THIS scale. That makes no sense. I mean, even if the disk went to sleep while you were gaming and then had to spin up upon zoning, it would add maybe 10 seconds maximum, not a minute... :O_o: I guess there are no bad sectors on it? I am trying to think about all variables that affect performance I can, like old slow disk, super heavy fragmentation, disk in sleep mode, etc., and not even all of that together should be causing such loading times. Heh, not even having all the game files on that disk.

[phazer11]

Well, turning off paging resulted in load times of 20 seconds or less in the areas that were taking over a minute before. My guess is that windows was swapping data during loads when it was not truly necessary, I certainly wasn't running out of memory.

 

Page files take up a decent amount of drive space, and SSD disk space on my budget anyway is very limited, so a page file on the conventional drive just slows things down too much and completely mitigates the advantage of SSD in some respects.

 

So, when I had a page file on my SSD, I had no issue EXCEPT my drive had less available space, this is how the problem went undiscovered because just a week or so ago I turned paging off for the SSD (don't need a page file on each drive, after all).

Windows does in fact use the page file even if ou have plenty of system RAM to spare. The other problem with putting the page file on the SSD is the increased number of writes.

https://www.tomshardware.com/reviews/ssd-performance-tweak,2911.html

[Kuldebar]

No, I know what you mean though, some people have like 500 things running in the background, etc.

 

It's always been talk/debate in geekdom about how windows memory page file is/is not a performance hog. For most people it is a non-issue and a no-brainer everyone should use it, but I think in some narrowly defined situations it could and does slow things down. To be clear, none of these were infinite loading screens...just consistently longer than normal load times. I get real used to how the game plays "normally" for me...so when it deviates I start looking for a reason why.

 

That's a good idea about manually adjusting the page file size, I will probably do that so I can keep it on my SSD...and that should resolve the issue I suspect.

Windows does in fact use the page file even if ou have plenty of system RAM to spare. The other problem with putting the page file on the SSD is the increased number of writes.

https://www.tomshardware.com/reviews/ssd-performance-tweak,2911.html

That's definitely a concern, and what prompted me to originally not use paging on the SSD way back when I got it.

[Octopuss]

Writes are not a concern at all. Realistically, even with heavy load, you will replace a SSD with something faster much sooner than it would "die" (I think someone calculated it to like 20 years, or even more - not exactly sure).

Edited June 2, 2014 by Octopuss

[Kuldebar]

Writes are not a concern at all. Realistically, even with heavy load, you will replace a SSD with something faster much sooner than it would "die" (I think someone calculated it to like 20 years, or even more - not exactly sure).

There's a back and forth debate ongoing, I don't worry too much about it tbh, but I don't defrag my SSD either. :)

 

 

 

First, it limits the number of writes to an SSD. Now, there is much debate about how worried enthusiasts need to be about writing unnecessarily to solid-state storage. As you probably know, write endurance is the specification that matters most in defining the longevity of memory cells that compose an SSD. Some people say not to worry; you'll probably not see the day when your drive's cells stop storing data reliably. Others do everything in their power to minimize writes, taking no chances. Turning off System Restore helps placate that latter group. Additionally, we've seen chatter indicating that System Restore potentially degrades performance over time due to the way it interacts with the TRIM command. Check out this story for more depth on that.

[Octopuss]

I remember now what I had in mind regarding the writes.

part I:   Introducing the SSD Endurance Experiment

part II:  The SSD Endurance Experiment: 200TB update

part III: The SSD Endurance Experiment: Testing data retention at 300TB

part IV: The SSD Endurance Experiment: 500TB update

part V:  The SSD Endurance Experiment: Data retention after 600TB

 

The whole test apparently took ~7 months (while the SSDs were writing data pretty much nonstop), so this is as real as it gets.

Actually, the test might not be over yet, llol.

 

"I used to be scared of writes like you, but then I took a flash chip to the knee."

Edited June 2, 2014 by Octopuss