parition & drive hookup ?s First, GREAT article on VM, I learned a lot! I have three physical drives. The first drive (c) is in 2 partitions, 38mb fat32 (unmounted?) and the rest 55+gb NTFS. Second drive is all one 20gb partition and the third is an external drive all one partition. Three questions: 1. My computer (a Dell) originally came with the C drive in the 2 partitions. The small 38mb paritition does not show up as a drive with a letter and I only can see it with some drive utilities. What is this partition for and what can I do with it? It is not shown as a usable drive in XP's VM settings. Since it is the first partition on the physical drive, it is using up valuable outer tracks of the drive. Would I be better off reformatting the whole drive as NTFS and reinstalling windows? The second physical drive (D) has one partition, but it is a slower 5400rpm compared the the C drive's 7200rpm. If I read your article correct, it would be a faster page file if I put a permanent one on the D drive if the two hard drives were hooked up to separate IDE channels even though the D drive is slower in RPMs, correct? 2. How do you physically hook up the PATA hard drives and optical drives to separate IDE channels (I have two)? That is, I want a master hard drive and slave dvd-rom on one IDE channel and another master hd and slave cd-rw on the secondary IDE channel. Every computer I have owned makes it difficult to do this. The cable is only long enough between connections to reach the next physical drive below it. Is there a special IDE drive cable that will reach a few extra inches to plug into a different drive? I could try to move the hard drives physically closer to the optical drives, but they wouldn't be secured to any part of the chassis. 3. I've followed your suggestions and used Diskeeper in the boot defrag method. My page file on both hard drives was not moved at all. What steps can I take to optimize the movement of the page file towards the outer edges of the drive? A reformat/install of windows with Diskeeper ran immediately following? Or something else? Thanks! pd
Thanks. Usually, companies like Dell create a Recovery Partition that contains installation files that can restore a corrupted operating system or partition into a factory-installed condition. However, I'm not entirely sure about that small 38MB partition. Perhaps it just contains the restoration utilities and not the installation files of the operating system itself. It is up to you whether you want to retain that partition or not. However, since I'm personally familiar about installing Windows directly from the CD, I do not need such recovery features. Therefore, I would use FDISK to delete all partitions and redo the partitions. However, if you are not comfortable about installing Windows by yourself, you should not delete that partition. Well, in the article, I was only looking from the standpoint of using two hard disks of the same speeds. But generally, putting the paging file on a separate hard disk on its own IDE channel will yield better performance, even if the hard disk is a slower hard disk. However, if you want to hedge your chances, you can create two paging files - one on each drive. Windows will automatically use the fastest one (in its context). Err.. Well, the only way to do that would be to use longer IDE cables or move the drives closer together. There are specially-shielded cables that are longer than usual. For example - http://www.rojakpot.com/showarticle.aspx?artno=134 The problem with newer disk defragmenting software is that they have their own internal algorithm that determines if the paging file should be moved at all. However, that's not such a bad thing at all. I will be explaining it in an upcoming update. But for now, the best way to get the paging file close to the outer tracks would be to format the partition, install Windows and create the paging file of your choice. This automatically creates a contiguous paging file that is as close to the outer tracks as is possible.
Ok, I misunderstood. You did use the word "peak". What was confusing is you said to run all apps and play games and look at the "peak" number. Well that's possible to do that opening a bunch of apps and then looking at the PF usage number. I didn't think that was possible to do that with a game cause when you exited the game, the PF number would immediately go down. I suppose Alt-Tab the game to desktop would work but I figured the number would drop then too. I also assumed (and still do) that the PF usage number was including the physical ram in the system. For example, right after booting up, my PF usage is around 200 megs. I believe that number is showing how much ram is currently being used, not how much swap file is being used. My understanding of the swap is that it's not used UNTIL all physical ram is used. So how could that 200 megs of swap be used when I have a gig of ram? If I set the swap to 0, at bootup the PF number is still going to read the same 200 megs. Now the Commit Charge Peak will always show the maximum number that was attained until a reboot. That's where I got confused. I thought what I said in my first post was also a really good way to judge what the size of the swap needs to be. If the Commit Charge Peak number goes higher than the ram in the system, then the difference will give a good indication of what the swap needs to be (adding some for a comfortable cushion of course). So if I got this stuff all wrong, please clarify because I thought I understood what these numbers mean. If I don't, I would like to know.
Mike89, I presume you're on Win2k as Win2k's task manager does not display the same items as in WinXP. You could instead use Win2K's Performance in the Administrative Tools It's basically similar to the guide's tip for Win9x on how to get the peak VM use. 1) Open up the Performance tool. Control Panel -> Administrative Tools -> Performance 2) See screenshot 1 attached, click on the + button to add items to the system monitor 3) See screenshot 2 attached, select Process from the Performance objecy drop down list. And then select both "Page File Bytes" and "Page File Bytes Peak" and for the instance column, select "_Total". Click on the "Add" button then "Close" 4) Minimize that window, or just leave it as it is. Run all your apps and etc until you're satisfied. Then go back to the Performance Monitor and press CTRL+R to get the report view. There you should see the peak Pagefile use as the attached screenshot3.
Oh. ok, my bad. But the steps i laid out above would also work for WinXP for the true peak Pagefile usage value. Remember that you would still need to pad up about 50-100mb on top of the peak usage recorded to be safe
That's okay. Maybe I wasn't specific enough. I already made corrections to the page. I personally prefer using the Page File Usage meter and graph as a quick and easy way to measure peak page file usage. That's because all I need to do is load up the memory-intensive apps that I usually use and check the peak page file usage. When it comes to games like Doom 3, you don't actually have to Alt-Tab to check the page file usage. Even if you do, the page file usage will NOT drop. There is no reason for the game to release those paged out data. In any case, you can set the Update Speed of Task Manager to Low. By the time you quit the game, the chart will still show the peak page file usage. That's why I say that for all practical purposes, using the Task Manager is an effective way to check on peak page file usage. Actually, Windows will automatically page-out data, even if it still has a lot of memory. That has already been covered in the guide. That doesn't mean that the data is running off the paging file. Windows just pre-emptively pages out data. That's why the page file will appear to be in use even after a fresh boot. The problem with that is ... Windows will always use as much memory as it can for the disk cache. That's one thing. The other thing is that Windows will pre-emptively page out data, whether there is plenty of memory or not. This allows Windows to quickly release memory if it runs out of memory. So, no matter whether you have lots of memory or not, the commit charge will always be high. IMHO, that's why it's not a good indicator of real page file usage.
Hmm, I guess I'm going to have to rethink my thinking. I had always assumed that the commit charge peak reading was the maximum amount of ram that was used. Taking Vampire for example. After exiting the game and checking the commit charge peak number (I routinely do this after playing a game thinking I was looking at how much ram the game was using) it read approx 1.3 gigs. That is the very first time I had seen that number go over 1 gig. I assumed from this that I had finally played a game that used more memory than I had (1 gig). I didn't like this predicament thinking this game was having to tap the hard drive for additional memory. (It sounds like from what you're saying is that all games are going to tap the hard drive at some point no matter how much ram you have. I still have a problem with that concept). Anyway I went and upped my system to 1.5 gigs of ram. Went back in and played Vampire and this time it played better than with the 1 gig (smoother like the hard drive was no longer being hit).
Hi again Adrian, Hello to the new posters in this thread, i've read all of it I was tring to test the ConservativeSwapfileUsage feature on windows me with 8MB/7200 HD and only 64MB SDRAM and the stopNTExecutive feature on a Winxp system with 7200 non8MB cache drive but with 1024DDR400. My conclusions are that the ConservativeSwapfileUsage on windows Me tends to be unstable, it is very hard to determine how much to set the swapfile. For example, Monitoring the system monitor's page file in use after I start the system will be 0MB but after opening a small program (sometimes even the notepad with a small textfile) it will jump to 96MB. Therefore I made the minimum to 128MB and the maximum to 512MB. computer is for surfing the internet and uses IE5.5sp2. With the XP computer I noticed that the PF Usage in the Task Manager's Performance tab is actually the PF allocation and not the real usage. For example when I set the minimum to 128MB the total real pagefile size is 128MB, PF Usage in Task manager(Allocation) is 174MB and the real usage (with stopping NTExecutive on) is 24MB (Peak 36MB). I'm still testing my optimal minimum page file settings (right now it is set to minimum 128 maximum 480 -no third-party defragment software here so PF isn't in the outer tracks). For monitoring I'm using a freeware version of OuterTracks CacheMan that is the same like the registered version for my purposes. www.outertech.com There are more freeware software for defragmenting pagefile (although didn't find anything for moving to outertracks) and defragmenting the registry. WinXP's SP2 had improved performance, I read that it is something to do with an uneffective file management feature turned off and there is a parth to set it off also in SP1. I also found out that Norton Utilities will move the page file to the outer tracks on win9x/me but not on winxp. Diskeeper came with a free upgrade to version 9 and it WILL move the PF to the outertracks in WinXP but not in 9x/me. I'm not sure but I think Norton Utilities's algorithm for file placement is better than diskeeper's so for winxp if it's possible I think better using diskeeper for moving the PF file on boottime and after it finishes, using NU for defragmenting the drive. Anyhow I have 3 questions: 1) is it effective to use the real page file usage like in the Microsoft Management Console or in the cacheman software i mentioned (not the allocation called "PF usage" in the task manager) for guidelines to the minimum pagefile settings while NTexecutive is stopped? 2) Did anyone had any more stable swapfile on 9x/me systems with conservativeSwapFile on? 3) Although pagefile readouts will be faster on the outertracks, they won't have to move to the outertracks if they are seeking data or software execution files in the data tracks (more close to the middle). Won't it be optimal for the page file to stay Hi again Adrian, I was tring to test the ConservativeSwapfileUsage feature on windows me with 8MB/7200 HD and only 64MB SDRAM and the stopNTExecutive feature on a Winxp system with 7200 non8MB cache drive but with 1024DDR400. My conclusions are that the ConservativeSwapfileUsage on windows Me tends to be unstable, it is very hard to determine how much to set the swapfile. For example, Monitoring the system monitor's page file in use after I start the system will be 0MB but after opening a small program (sometimes even the notepad with a small textfile) it will jump to 96MB. Therefore I made the minimum to 128MB and the maximum to 512MB. computer is for surfing the internet and uses IE5.5sp2. With the XP computer I noticed that the PF Usage in the Task Manager's Performance tab is actually the PF allocation and not the real usage. For example when I set the minimum to 128MB the total real pagefile size is 128MB, PF Usage in Task manager(Allocation) is 174MB and the real usage (with stopping NTExecutive on) is 24MB (Peak 36MB). I'm still testing my optimal minimum page file settings (right now it is set to minimum 128 maximum 480 -no third-party defragment software here so PF isn't in the outer tracks). For monitoring I'm using a freeware version of OuterTracks CacheMan that is the same like the registered version for my purposes. www.outertracks.com There are more freeware software for defragmenting pagefile (although didn't find anything for moving to outertracks) and defragmenting the registry. WinXP's SP2 had improved performance, I read that it is something to do with an uneffective file management feature turned off and there is a parth to set it off also in SP1. I also found out that Norton Utilities will move the page file to the outer tracks on win9x/me but not on winxp. Diskeeper came with a free upgrade to version 9 and it WILL move the PF to the outertracks in WinXP but not in 9x/me. I'm not sure but I think Norton Utilities's algorithm for file placement is better than diskeeper's so for winxp if it's possible I think better using diskeeper for moving the PF file on boottime and after it finishes, using NU for defragmenting the drive. Anyhow I have 3 questions: 1) is it effective to use the real page file usage like in the Microsoft Management Console or in the cacheman software i mentioned (not the allocation called "PF usage" in the task manager) for guidelines to the minimum pagefile settings while NTexecutive is stopped? 2) Did anyone had any more stable swapfile on 9x/me systems with conservativeSwapFile on? 3) Although pagefile readouts will be faster on the outertracks, they won't have to move to the outertracks if they are seeking data or software execution files in the data tracks (more close to the middle). Won't it be optimal for the page file to stay still and defragmented like it is after a fresh installation of winxp? I'm sure that NU algorithm somewhat solves the excutables problem but the data can still be in the end of the disk (i hope I'm making any sense. Another question that I guess most of you won't have the answer from Hi again Adrian, I was tring to test the ConservativeSwapfileUsage feature on windows me with 8MB/7200 HD and only 64MB SDRAM and the stopNTExecutive feature on a Winxp system with 7200 non8MB cache drive but with 1024DDR400. My conclusions are that the ConservativeSwapfileUsage on windows Me tends to be unstable, it is very hard to determine how much to set the swapfile. For example, Monitoring the system monitor's page file in use after I start the system will be 0MB but after opening a small program (sometimes even the notepad with a small textfile) it will jump to 96MB. Therefore I made the minimum to 128MB and the maximum to 512MB. computer is for surfing the internet and uses IE5.5sp2. With the XP computer I noticed that the PF Usage in the Task Manager's Performance tab is actually the PF allocation and not the real usage. For example when I set the minimum to 128MB the total real pagefile size is 128MB, PF Usage in Task manager(Allocation) is 174MB and the real usage (with stopping NTExecutive on) is 24MB (Peak 36MB). I'm still testing my optimal minimum page file settings (right now it is set to minimum 128 maximum 480 -no third-party defragment software here so PF isn't in the outer tracks). For monitoring I'm using a freeware version of OuterTracks CacheMan that is the same like the registered version for my purposes. www.outertracks.com There are more freeware software for defragmenting pagefile (although didn't find anything for moving to outertracks) and defragmenting the registry. WinXP's SP2 had improved performance, I read that it is something to do with an uneffective file management feature turned off and there is a parth to set it off also in SP1. I also found out that Norton Utilities will move the page file to the outer tracks on win9x/me but not on winxp. Diskeeper came with a free upgrade to version 9 and it WILL move the PF to the outertracks in WinXP but not in 9x/me. I'm not sure but I think Norton Utilities's algorithm for file placement is better than diskeeper's so for winxp if it's possible I think better using diskeeper for moving the PF file on boottime and after it finishes, using NU for defragmenting the drive. Anyhow I have 3 questions: 1) is it effective to use the real page file usage like in the Microsoft Management Console or in the cacheman software i mentioned (not the allocation called "PF usage" in the task manager) for guidelines to the minimum pagefile settings while NTexecutive is stopped? 2) Did anyone had any more stable swapfile on 9x/me systems with conservativeSwapFile on? 3) Although pagefile readouts will be faster on the outertracks, they won't have to move to the outertracks if they are seeking data or software execution files in the data tracks (more close to the middle). Won't it be optimal for the page file to stay Hi again Adrian, I was tring to test the ConservativeSwapfileUsage feature on windows me with 8MB/7200 HD and only 64MB SDRAM and the stopNTExecutive feature on a Winxp system with 7200 non8MB cache drive but with 1024DDR400. My conclusions are that the ConservativeSwapfileUsage on windows Me tends to be unstable, it is very hard to determine how much to set the swapfile. For example, Monitoring the system monitor's page file in use after I start the system will be 0MB but after opening a small program (sometimes even the notepad with a small textfile) it will jump to 96MB. Therefore I made the minimum to 128MB and the maximum to 512MB. computer is for surfing the internet and uses IE5.5sp2. With the XP computer I noticed that the PF Usage in the Task Manager's Performance tab is actually the PF allocation and not the real usage. For example when I set the minimum to 128MB the total real pagefile size is 128MB, PF Usage in Task manager(Allocation) is 174MB and the real usage (with stopping NTExecutive on) is 24MB (Peak 36MB). I'm still testing my optimal minimum page file settings (right now it is set to minimum 128 maximum 480 -no third-party defragment software here so PF isn't in the outer tracks). For monitoring I'm using a freeware version of OuterTracks CacheMan that is the same like the registered version for my purposes. www.outertracks.com There are more freeware software for defragmenting pagefile (although didn't find anything for moving to outertracks) and defragmenting the registry. WinXP's SP2 had improved performance, I read that it is something to do with an uneffective file management feature turned off and there is a parth to set it off also in SP1. I also found out that Norton Utilities will move the page file to the outer tracks on win9x/me but not on winxp. Diskeeper came with a free upgrade to version 9 and it WILL move the PF to the outertracks in WinXP but not in 9x/me. I'm not sure but I think Norton Utilities's algorithm for file placement is better than diskeeper's so for winxp if it's possible I think better using diskeeper for moving the PF file on boottime and after it finishes, using NU for defragmenting the drive. Anyhow I have 3 questions: 1) is it effective to use the real page file usage like in the Microsoft Management Console or in the cacheman software i mentioned (not the allocation called "PF usage" in the task manager) for guidelines to the minimum pagefile settings while NTexecutive is stopped? 2) Did anyone had any more stable swapfile on 9x/me systems with conservativeSwapFile on? 3) Although pagefile readouts will be faster on the outertracks, they won't have to move to the outertracks if they are seeking data or software execution files in the data tracks (more close to the middle). Won't it be optimal for the page file to stay still and defragmented like it is after a fresh installation of winxp? I'm sure that NU algorithm somewhat solves the excutables problem but the data can still be in the end of the disk (i hope I'm making any sense. Another that I guess most of you won't have the expiriance refers to working with the computer as a music producing/editing/recording studio. Usually for this application there is a need for minimum 2 HDs. 1 for OS and software and one for the audio files (same goes for film editing). From what i read and from my logic I think the best solution for 2 HDs is to put the pagefile in the OS drive and leave the other one for audio files (for editing/mixing it's less significant but for recording it is). Can anyone confirm my opinion? Thank you very much and sorry for the long posting! -Noam :Edit: Tried to think about it again and I guess ConservativeSwapFile isn't meant for 64MB RAM.. Also,. I saw on the IamNotaGeek performance guide the key LargeSystemCache for loading system kernel into RAM. is it worth a look? more keys that I found maybe useful - SecondLevelDataCache, IdleEnable-USB polling Interval, AlwaysUnloadDLL, EnablePrefetcher=3, Enable UDMA 66, IRQ8Priority, Win32PrioritySeparation=38
XP not obeying specified VM settings I don't know if anyone else has noticed this (or if it's just my system)..Anyhow here's what I've discovered. I've got a 120GB HD partitioned in two - primary/boot partition is 30GB, second partition is the rest (90GB). I've got 768MB memory and am running XP with SP2 installed. I've disabled the pagefile on the second partition (D: ) and set the pagefile on my first partition (C: ) to an initial size of 512MB and a max. size of 1024MB. Windows lists a 'recommended' size of 1150MB. Here's the problem - on reboot, the Virtual Memory properties page displays that the pagefile is currently allocated at 1151MB despite my specified settings - the file itself is ~1.8GB. If I disable the pagefile all together and then reboot, the VM properties page displays an allocation of 0MB (although it's interesting to note that taskmanger lists a PF usage but no pagefile.sys file is found on the system...). If I then specify the VM settings as above, the system accepts the changes without requiring a reboot - checking the pagefile.sys size and everything is as it should be. However, upon a reboot, the problem reappears - pagefile is allocated at 1151MB. If I move the pagefile to the D: partition, it works as it should on every boot. If I set up the pagefile on both partitions (setting a min./max. of 2MB on D: and C: as specified above) the problem reappears. Finally, if I set it such that the page file on c: is 512/1024 and d: is 128/128, then all works fine - note that the total 'max' between both partitions matches the 'recommended' size Windows specifies - it appears that if you specify less than what Windows would like, you run into the problems I've mentioned. Thinking I've figured out the problem, I go ahead and disable the page file on D: and set C: to 512/1152. Reboot and find that the page file is back at 1151MB in size initially... Can anybody confirm the same problem on their end? Finally, has anyone used Perfect Disk 7.0 for defragmenting? Also, Norton's Speed Disk appears to be able to put the pagefile at the very begining of the drive - just be sure to specify that pagefile.sys should be placed first under the 'Files First' option.
Mister E... : Do check out this article here The Stubborn Paging File Guide! http://www.rojakpot.com/showarticle.aspx?artno=163
Are you using NTFS or are you using FAT 32? If using FAT32 then you are better off limiting the boot partition to 8 GB. You can install a complete copy of Windows XP in the 8 GB and it will be very fast due to the 512 byte allocation units. However, as far as the problem you described. I never experienced this myself. I have a 256MB paging file and never had any problems when moving it from one partition to the other.
peaz - that did the trick - didn't see that article. I'm running NAV 2004 - in the configuration under AutoProtect->Advanced I turned off "Load Auto-Protect during system boot (recommended)" and that solved the problem (auto-protect still loads without a problem). Bitpower - I've got NTFS here - thanks for the input though!
Glad to be of help in any way BTW... welcome to ARP's forum too!!! Hope you enjoy it here and hang around our forums too!
Woah, Protools_Opearator! That's a really long post! I can't think properly right now. Still sick. But let me address what I can tonight. About PF Usage and monitoring true paging file usage, well, IMHO what you saw is because Windows XP automatically uses more than 128MB of virtual memory at start up. So, with a minimum paging file of 128MB, Windows XP automatically starts off with 128MB. For my setup of 512MB minimum, Windows XP boots up with a paging file usage of almost 200MB. And this changes as I load up my software, etc. About ConservativeSwapFile with Windows ME, I'm not entirely sure if it supports it. If I'm not wrong, that's officially supported only with Windows 98, although people have stated that it works with Windows ME too. OT a little, Windows ME really sucks. I would really suggest downgrading to Windows 98 or moving up to Windows XP. As for your last question.. I think you are referring to less seeks if the paging file is situated in the middle of the partition, right? I'm not entirely sure what you are referring to.
Another Question... Here's another question: With Windows 98/ME, there was a tweak which I'm sure most of you are aware of - adding the line "ConservativeSwapfileUsage=1" to your system.ini file under the [386enh] section. This essentially told Windows to use RAM more and the swapfile less. From what I've seen so far, it appears that this setting should have no effect in Windows XP. However, for the sake of experimentation, I added the line to my system.ini file and compared the Pagefile usage as reported by Task Manager. I openned the exact same apps/windows with the line added and with it missing and compared (rebooting each time). With the line added, my PF Usuage hit ~200MB; with the line removed, PF Usuage hit ~230MB. Also, from a fresh boot, no apps loaded (other than the same startups) - with the line 145MB, without 150MB... Now, I'd have to do more extensive testing/comparisons to fully determine whether or not my findings are accurate. Does anyone else want to try and compare notes (or do you have a definitive answer)?
Yep - after doing some more testing, that appears correct (i.e. no effect on XP). Also, doing some more digging turned up the following info regarding the option under Win98 - it appears that it doesn't do what most think/thought it does/did... http://support.microsoft.com/default.aspx?scid=kb;en-us;223294
Yup, that's what the PageFile_Call_Async_Manager service does. But it appears that disabling it actually reduced page file usage. This is based on anecdotal evidence, of course.
Adrian, Excellent guide! I just have one question though. You mention that "large cluster sizes improve the performance of the file system", and that the cluster size should be set to 4Kb minimum, but is there an upper limit at which you gain little in going any larger? ie. Does it make any sence to set the cluster size of the partition that the page file is on to 64Kb for NTFS or 256Kb for FAT? (assuming the partition is dedicated to the swap file and has no other data or programs on it) thanks, Bill
