Which drive for pagefile?

[dparm]

Here's the deal: I have an external USB hard drive (7200rpm 8MB cache Western Digital) hooked up to my laptop for additional storage. I've installed all my games to it, and put some other backups on there as well.

Would it be faster to have Windows use a pagefile off the USB drive, or the internal drive? I currently have it on the internal drive because I was thinking that if you have the pagefile on the same physical drive as the games, you'd slow yourself down. The "rule" is to keep the pagefile on a separate drive from your System and/or most commonly used apps.

Thoughts?

[abudhu]

Keep it on the internal. Interally it will transfer data much faster than the USB 2. And yes as a rule of thumb you usually want your PF to be on a seperate partion. I have my PF working off my 2nd internal rather than my first which is my OS. You could also partition your internal drive up, and set one of the partitions as the PF.

There was a way to calculate if you need PF...let me find it:

CTRL+ALT+DELETE = Taskmanger
Click the Performance Tab:
Look at the Commit Charge: Look at the PEAK
Now take that number Peak * 1.25.
Take your total Ram and subtract it from that.

So.

(Peak*1.25)-Total Ram.

If you get a + number thats the number your PF should be. This is the BARE minimum it needs. So its wise to give yourself some comfort level and + 300 on.
Negative means you dont Really need it.

I always get a negative number but I use PF, because I noticed a negative impact form turning it off. Others have seen postive impacts from having it off.

Also its very WISE to not allow your PF to grow. As your PF grows it fragments, like all good things do, and your normal Disk Defrag has no way of defraging it. You will have to download a program to defrag your PF. If you do not allow it to grow you decrease the amount of time it takes to Fragment. I have mine set at 2000-2000.

[dparm]

Thanks.

I'm familiar with correct sizing. Mine is set to a decent level now. I don't let the system manage it, and I have a pagefile defragmenting utility that keeps it in one piece. I too know of that formula and get a negative number, but still use a pagefile.

I'll keep it on the internal like I had before. The internal is a 5400RPM IBM, BTW.

(FWIW, completely disabling the pagefile is a very bad idea. Some programs need swap memory REGARDLESS, and I know that Windows itself doesn't like having a 0MB pagefile.)

There is an interesting setting in Windows 2000 & XP that can force the entire Windows kernel to load into RAM rather than a swap file...it was originally intended for servers but can be a big performance boost on systems with 1GB+ RAM. I experimented with this but found it too unstable, especially with the ATI drivers (a known problem).

[jasonk]

Putting a pagefile on a separate partition on the same physical disk yeilds no improvement, and having it on a separate disk only helps if the disk is on a different channel (since IDE can only read from one IDE device per channel at a time).

The pagefile.sys virtual memory architecture came from VMS, and is designed around being on the same disk as applications and OS, so the benefit of even having it on a separate disk on a separate channel is minimal at best.

Just set it to something static (so windows won't resize it with the chance of it being fragemented) and let it be.

[Unregistered]

Mine is set at a static size. I'll just leave it on the internal since I need every ounce of speed I can get while gaming (which has all its files on the external).

Just curious, what kind of benchmarking utilities would be useful to test this? There was no change in 3DMark scores.

[dparm]

Dang, forgot to log in on that last post.




Mine is set at a static size. I'll just leave it on the internal since I need every ounce of speed I can get while gaming (which has all its files on the external).

Just curious, what kind of benchmarking utilities would be useful to test this? There was no change in 3DMark scores.

[jasonk]

The best test would be something like a content creation benchmark or office application benchmark that you run with some of your RAM removed (down to ~256) so that windows is forced to page during the benchmark.

[dparm]

Yeah but aren't there benchmarking utilities out there? Most of the big computer magazines/websites always have these bar charts...

[jasonk]

Yeah but aren't there benchmarking utilities out there? Most of the big computer magazines/websites always have these bar charts...

Yes, there are... I was referring to BAPCOs stuff in my previous post..

(Which apparently cost $400 :eek: )

You could probably... find... them... somewhere....

Ziff Davis makes Winbench... looks like you can get it free here...
http://www.veritest.com/benchmarks/winbench/default.asp

[dparm]

Winbench, that's what I was thinking of.

I know Systemworks 2004 had PerfTest, a similar utility.