[OT] virtual memory
 

i've got a new hard drive. should i put virtual memory on it, to?

Re: [OT] virtual memory
 

You usually want to put virtual memory on the drive with the most free space. If the new drive has more free space, go for it.

Re: [OT] virtual memory
 

I think you should put your vm on the fastest drive. 99% of the time, the slowest and most limiting component on a computer is the harddrive. Speed it up, access it less, or access a faster one and you have a faster system.

JM $0.02

Slick.

Re: [OT] virtual memory
 

If you have enough RAM you can try to disable virtual memory altogether.
If you get out of memory errors or warnings you can enable it.

If you can work without virtual memory it is prefarable, as it will speed up the computer some.
If you must use VM try to set it to a fixed size, the scaling of it up and down also takes some time, and if it is a fixed size you can do a boot defrag on it, if it's dynamic it will fragmantize.

[ July 31, 2003, 06:54: Message edited by: Ruatha ]

Re: [OT] virtual memory
 

The virtual memory should be fixed in size, do not let the page file alter its size dynamically. This causes it to be fragmented all over the partition.

If you have a second drive on a second controller, moving the page file to that drive will speed performance a little. It lets the system page at the same time that the system is accessing data on the system drive. Well, almost at the same time http://forum.shrapnelgames.com/images/icons/icon7.gif Moving it to your SCSI stripe volume could be interesting :0

With a gig of memory, a large page file will slow the system down. Above a gig, the memory addressing load will start to slow system performance, and windows has a way of grabbing every bite of ram that it can latch onto.

As a rule, the page file should be 1.5 times the size of the installed ram. But I never found any reason for this. Big page files take longer to access, so IMHO keep it small.

No page file can cause problems. Some apps are written to use the page file for some data. This was done to free up system ram back in the day, but it still populates some code. Seti is an app that pages by default. IMHO 128megs is a ball park lower limit.

All things ram are done in 32’s, 32-64-128-256-512-1024. Try to work the page file size in the same increments. It eases the addressing load on the CPU.

Re: [OT] virtual memory
 

<font size="2" face="Verdana, Helvetica, sans-serif">Still, if you don't use proggies that needs VM it's better to have none, as it then won't access the HD all the time.

<font size="2" face="Verdana, Helvetica, sans-serif">This I seriously question!
What should be the reason for the CPU to have "easier adress load" with 33554432 Bytes (32M) or 32M-some (ex 30M)?
Every adress between the 24 and the 32M range uses the same adress length (actually, above 16M it uses the same registers IIRC).
And a modern PC is made for large adress lenghts, so it should make no significant change to the internal workload for the computer what size the page file is, the only difference is in the drive handling. (And as I've stated far below in fragmantazion, aka also drive issues)
Right?
Enlighten me.

[ August 01, 2003, 07:24: Message edited by: Ruatha ]

Re: [OT] virtual memory
 

i read 2.5.

what's an SCSI?

and remember, when you open up a bunch of tabs, reload.

[ August 01, 2003, 07:27: Message edited by: narf poit chez BOOM ]

Re: [OT] virtual memory
 

<font size="2" face="Verdana, Helvetica, sans-serif">Small Computer Systems Interface.

A interface for harddrives and other peripheals, instead of the AT/IDE bus and for example the parallel port.
Can be used for a various array of peripherals but is mainly used for disks.
Previously RAID was exlusive for SCSI (Redundant Array of Inexpensive Disks) but now you can do that in IDE with a RAID controller aswell.
The SCSI interface is smarter than IDE and has some inbuilt logic circuits that made it faster in shuffling data and SCSI disks where generally faster and larger than their ATA counterparts.
The differences are lessening all the time.

Re: [OT] virtual memory
 

The configuration of virtual memory even depends on your Version of windows. i think windows xp can even manage multiple page-files.

On rule to speed up things is: if you have two drives, put them on different controllers. Put your resource-consuming application to the drive where the page file is NOT. So using the page file and reloading software components won't interfere.

It is not a good solution to turn of the virtual memory. If you run multiple applications, your system will soon need it. Btw: Windows usually writes first the data to disk, that appears to be less important( e.g. idle apps etc )

<font size="2" face="Verdana, Helvetica, sans-serif">I think the cpu runs in protected mode anyways. So every address is 32-Bit. There are no more "far" and "short" jumps. the cpu accesses each (even)adress at the same speed.

[ August 01, 2003, 08:37: Message edited by: JoeViterbo ]

Re: [OT] virtual memory
 

I ran along time with 512M ram and no page file.
It's just when I started using many programs at once that I needed the VM, and then you get an error message telling you soo.
So there's no danger in trying to disable it and see if you can live without it.

I use XP an dhas most unnessacary services disabled or started when needed (Started automaticly then but actually in the services folder it says "Manual" http://forum.shrapnelgames.com/images/icons/icon7.gif )

And yes, in XP you can configure one page file for each drive if you wish.

[ August 01, 2003, 08:54: Message edited by: Ruatha ]

Re: [OT] virtual memory
 

I run my comp with 512 mb RAM and no pagefile under Win XP fine too. I find that the only apps that need VM are resource-intensive 3-D games and graphics applications.

Re: [OT] virtual memory
 

<font size="2" face="Verdana, Helvetica, sans-serif">Still, if you don't use proggies that needs VM it's better to have none, as it then won't access the HD all the time.

Quote:

All things ram are done in 32’s, 32-64-128-256-512-1024. Try to work the page file size in the same increments. It eases the addressing load on the CPU.

<font size="2" face="Verdana, Helvetica, sans-serif">This I seriously question!
What should be the reason for the CPU to have "easier adress load" with 33554432 Bytes (32M) or 32M-some (ex 30M)?
Every adress between the 24 and the 32M range uses the same adress length (actually, above 16M it uses the same registers IIRC).
And a modern PC is made for large adress lenghts, so it should make no significant change to the internal workload for the computer what size the page file is, the only difference is in the drive handling. (And as I've stated far below in fragmantazion, aka also drive issues)
Right?
Enlighten me.</font><hr /></blockquote><font size="2" face="Verdana, Helvetica, sans-serif">32 bit memory buss.

Re: [OT] virtual memory
 

<font size="2" face="Verdana, Helvetica, sans-serif">I think the cpu runs in protected mode anyways. So every address is 32-Bit. There are no more "far" and "short" jumps. the cpu accesses each (even)adress at the same speed.</font><hr /></blockquote><font size="2" face="Verdana, Helvetica, sans-serif">XP and 2K can both manage more than one page file.

The reason for x32 page file size goes back to the original x86 code, and the function of the data buss. Take a look a the default page file size that windows sets. It will vary with the amount of ram installed, but it will always devide by 32.

A bench mark trick is to run a small page file. Just enough to meet the needs of the mark being run. Some times it will be 1meg or even 1/2 a meg. If you vary from x32, you will see it in the results.

Re: [OT] virtual memory
 

Ok Thermo I see your point but it's still misguided.
If you change your page file to 32 M it doesn't matter if it 's anywhere between 16M+1 byte and 32M, it still has the same adress range, so having 32M doesn't make it go any faster than having 30M!
It's not some magic number, it's simply logic.
If you change from 32M to 32M+1 byte it can have some insignificant affact, hardly noticable I'd think but I'd agree that you might see it with a benchmark, haven't tested though, bit from say 24M to 29M or 32M it wouldn't have any aeffect, i e 32M page file is not faster then 24M pagefile.

The reason windows et it to "size dividable with 32" is that it want's the maximum amount possible if it's going above "size divided by 32" +1 byte, but youas a thinking human might hav ereasons not to set it that way.
And I've seen "not even dividable by 32" sizes if the free part of the disk doesn't allow it.
See?

[ August 02, 2003, 08:45: Message edited by: Ruatha ]

Re: [OT] virtual memory
 

Not exactly, but also true. As systems have gotten faster, page file virtual addressing has gotten more transparent, the load is a very small percentage of CPU time now. But the system still expects to see a size based on Base2 binary, starting with a 1 and then being all zeros. It will write each page as a 4kb page or slice of data. A simple way of seeing this is to enter 1 and a 0 into your electric slide rule, then look at the result. Repeat the process adding a zero each time. The results will be well known numbers.

Here is a link to the best place I know to get plain English answers on this stuff.

http://aumha.org/win5/a/xpvm.php

If you need more details, then we suffer the Chinese horror of the 1000 white papers 

Re: [OT] virtual memory
 

That was a good link!

Re: [OT] virtual memory
 

Glad you liked it http://forum.shrapnelgames.com/images/icons/icon7.gif

In the long gone days of win98, this guy was the keeper of the grail for those of us who couldn’t deal with the white papers.