*Initial Stats:*
These results are as of the initial Boot with an active wireless connection to my home network. I have considered the machine ‘idle’ as there were no applications running outside of the Windows Task Manager to deliver these statistics (and the active wireless connection with no activity).
C: Drive - 2,632,749,056 – Used (2.45GB)
5% - CPU Consumption at Idle
237020 Total RAM
99200 – 99600 at idle
80600 – System Cache at idle
PF Usage – 111MB
Kernel Memory 19084
Paged 15032
NonPaged 4052
*Installed Tune-Up Utilities 2004:*
Installed a registered full version of the Tune-Up Utilities 2004 and before adjusting any settings I took the following readings using TUU to gather the following machine statistics:
Processor: Transmeta ™ Crusoe™ Processor TM5800 99…
L2 Cache: 512
Memory: 231 MB (62% used)
Mainboard: OQO ZEPTO
Monitor: Default Monitor
Video Card: OQO Lynx3DM
Mode: 800x480 with 24bit color depth
Drive: Toshiba MK2004GAL
Initial Boot Timings
I was going to attempt to log and track boot times. However, this task quickly became to time consuming as to do it right would require several measurements relating to cold boot vs warm boot, and where you drawn the line of when you consider the machine booted (i.e. when logon appears? When start menu appears? When all services have fully started?, etc.) I will leave that to another person for another day. I believe that the optimizations I am about to make are going to have a positive impact and to create my subjective baseline I have elected to us a single boot time that represents the following:
Cold boot from a cold machine
No logon screen (straight into the machine) (not secure I know)
All services to be started must be started and settled and all indications of activity have ceased. (subjective I realize, but the best you are gong to get from me, because to do this really right would take more time and energy than I am going to expend, sorry).
From hitting the power button until a full boot and services started and settled took one (1) minute fifteen (15) seconds. I am sure mileage may vary but this is my timings using a stopwatch.
Customization & Optimization of XP Installation
I wrote this while I was actually doing these times so these are the exact order that I did them in.
1. Run Windows Update – Custom Install Option and didn’t find any updates to install. I opted not to install the three optional items including the .Net stuff as of now.
2. Using TUU I started with a registry defrag. TUU reported that it could reduce the registry by 17% or a little over three megabytes (3,256K). Decided that an improvement here would certainly help the machines performance so went ahead and let it rip.
Results were interesting:
Memory Available grew to 120200 – 127400 at idle
System cache grew to 102524 at idle
PF Usage dropped to 97 MB
Kernel Memory dropped to 15092
Paged dropped to 11768
NonPaged dropped to 3324
These results are very positive, but a little odd nonetheless. Lets keep moving.
3. To speed up overall disk performance I decide to turn-off the indexing service on the HD. To disable the Indexing Service go into "My Computer", right-click on all your hard drive partitions one at a time, left-click "Properties". Uncheck "Allow Indexing Service to index this disk for fast file searching". Select "Apply changes to subfolders and files". If any files cannot be updated select "Ignore All". The constant indexing of files actually slows down system performance. Indexing searching performance benefits would only be noticeable on a file server not a home PC. Seems like an obvious candidate for easy optimization.
4. To further realize benefits I elect to actually remove the indexing service from the machine. I just don’t see the value for the cost to overhead. While I am at it, I don’t use MSN as my ISP so I go ahead and nuke those services as well. While I am in there I notice that Outlook Express and Windows Messenger can go as well. Go to "Start", "Settings", "Control Panel", "Add or Remove Programs", select "Add/Remove Windows Components", uncheck the “Indexing Service”, “MSN Explorer”, “Outlook Express” and “Windows Messenger” then select "Next" and "Finished.”
5. Setting a static size to your virtual memory file (page file) will reduce system overhead. XP will no longer have to dynamically manage the size of the file and it will help with fragmentation on your drive. When setting the static size of the file the generally accepted rule of thumb is to have this set to around one-and-a-half the size of your real memory. HOWEVER, any file larger than 512MB can and frequently will slow your machine down. Given that the OQO has less than 256MB real memory available setting this to a static file size of either 384 or 512 seems about right. As I would prefer to be told that I am out of memory rather than face additional overhead I opt for 384MB as both my minimum and maximum page file (virtual memory) size and thus make XP keep it static in size. To do this I went to "Start", "Settings", "Control Panel", "System", "Advanced" tab, in the "Performance" section select "Settings", "Advanced" tab, in the "Virtual Memory" section select "Change", change the "Initial Size" and "Maximum Size" to the same size then select "Set" and "OK".
6. My next optimization may not be too popular with many people. Since the OQO does not have a speaker (usable by XP) to system sound effects I elect to save the computer cycles used by the OS to play sounds to the sound device when it would be very rare that I would be able to hear them anyway. This adjustment will only affect system sounds (error beeps, boot songs, etc.) By leaving the sound devices and services alone when I perform an activity where I am likely to have headphones on (watch a movie, listen to music, play a game, etc.) the sound will work just fine. I just won’t be spending precious system resources to play beeps and sounds on the OQO that I will not likely be able to hear. To disable all Windows XP task sounds go to "Start", "Settings", "Control Panel", "Sounds and Audio Devices", select the "Sounds" tab, under "Sound Scheme" select "No Sounds". I saved this a Sound Scheme ‘None’ just so as to make it clear to myself later.
7. Feeling pretty good at this point, I decide that it is time to install some more software before I begin further optimizations. I have always great results with Executive Software’s Diskeeper software for optimizing NTFS file systems. So, now seems like a good time to get that installed. I purchased several licenses for 7.0 quite a while ago. There is a newer version 9.0 now out, but I have not seen anything substantially new enough to make me upgrade. After install and a bug-fix upgrade I run Diskeeper which replaces the default Windows Defrag utility. I select to do a boot defragmentation and check the following options: Directory Consolidation, Run CHKDSK before defragmentation, Reduce paging file fragmentation (now that we have it static), Defragment the MFT, Pause after defragmentation. I elect to do all of these things on the next reboot and of course then I reboot. The three pass CHKDSK goes clean and it reboots followed by the deep defragmentation process. All six stages of defragmentation complete without a hitch. The 6th stage took the longest, as it was the defragmentation of the paging file. After the reboot, I perform a standard defragmentation with Diskeeper.
8. Feels like time to run another boot timing and look at operating environment. Using the same criteria as the initial boot timing, overall boot time decreased by just over 12 full seconds. Environment stats from Task Manager have improved over the initial optimized reboot:
Available memory grew to: 127000 - 155400
System cache decreased to 86280
PF Usage dropped to 94.2 MB
Kernel Memory dropped to: 14680
Paged dropped to: 10516
NonPaged dropped to: 3164
While I believe that these are good results, this is far from scientific. All the key numbers moved in the right direction by substantial enough amounts to have made the effort appear worthwhile to me. It also goes to show that using the Task Manager to evaluate differences is problematic, as XP dynamically changes it’s optimizations based on the environment it detects.
While I will likely continue to work on optimization these changes seem like a really good start and I will now divert my attentions towards installing my applications and visual goodies to make my OQO my ‘own’. I will post another article on what I do to skin the OQO with minimal performance impact for as dramatic as effect as possible. After all something as cool as the OQO should not just be functional it should look really cool while doing it.
Guess you can tell I will NOT be running windows classic UI with an 8 bit color depth to optimize performance!
I hope some of this was useful to someone here. Special thanks go to *Captain* for the introduction to the TUU suite. Also, some of the techniques were leveraged from the following outstanding website link and so all credit for them should go there.
Cheers!
These results are as of the initial Boot with an active wireless connection to my home network. I have considered the machine ‘idle’ as there were no applications running outside of the Windows Task Manager to deliver these statistics (and the active wireless connection with no activity).
C: Drive - 2,632,749,056 – Used (2.45GB)
5% - CPU Consumption at Idle
237020 Total RAM
99200 – 99600 at idle
80600 – System Cache at idle
PF Usage – 111MB
Kernel Memory 19084
Paged 15032
NonPaged 4052
*Installed Tune-Up Utilities 2004:*
Installed a registered full version of the Tune-Up Utilities 2004 and before adjusting any settings I took the following readings using TUU to gather the following machine statistics:
Processor: Transmeta ™ Crusoe™ Processor TM5800 99…
L2 Cache: 512
Memory: 231 MB (62% used)
Mainboard: OQO ZEPTO
Monitor: Default Monitor
Video Card: OQO Lynx3DM
Mode: 800x480 with 24bit color depth
Drive: Toshiba MK2004GAL
Initial Boot Timings
I was going to attempt to log and track boot times. However, this task quickly became to time consuming as to do it right would require several measurements relating to cold boot vs warm boot, and where you drawn the line of when you consider the machine booted (i.e. when logon appears? When start menu appears? When all services have fully started?, etc.) I will leave that to another person for another day. I believe that the optimizations I am about to make are going to have a positive impact and to create my subjective baseline I have elected to us a single boot time that represents the following:
Cold boot from a cold machine
No logon screen (straight into the machine) (not secure I know)
All services to be started must be started and settled and all indications of activity have ceased. (subjective I realize, but the best you are gong to get from me, because to do this really right would take more time and energy than I am going to expend, sorry).
From hitting the power button until a full boot and services started and settled took one (1) minute fifteen (15) seconds. I am sure mileage may vary but this is my timings using a stopwatch.
Customization & Optimization of XP Installation
I wrote this while I was actually doing these times so these are the exact order that I did them in.
1. Run Windows Update – Custom Install Option and didn’t find any updates to install. I opted not to install the three optional items including the .Net stuff as of now.
2. Using TUU I started with a registry defrag. TUU reported that it could reduce the registry by 17% or a little over three megabytes (3,256K). Decided that an improvement here would certainly help the machines performance so went ahead and let it rip.
Results were interesting:
Memory Available grew to 120200 – 127400 at idle
System cache grew to 102524 at idle
PF Usage dropped to 97 MB
Kernel Memory dropped to 15092
Paged dropped to 11768
NonPaged dropped to 3324
These results are very positive, but a little odd nonetheless. Lets keep moving.
3. To speed up overall disk performance I decide to turn-off the indexing service on the HD. To disable the Indexing Service go into "My Computer", right-click on all your hard drive partitions one at a time, left-click "Properties". Uncheck "Allow Indexing Service to index this disk for fast file searching". Select "Apply changes to subfolders and files". If any files cannot be updated select "Ignore All". The constant indexing of files actually slows down system performance. Indexing searching performance benefits would only be noticeable on a file server not a home PC. Seems like an obvious candidate for easy optimization.
4. To further realize benefits I elect to actually remove the indexing service from the machine. I just don’t see the value for the cost to overhead. While I am at it, I don’t use MSN as my ISP so I go ahead and nuke those services as well. While I am in there I notice that Outlook Express and Windows Messenger can go as well. Go to "Start", "Settings", "Control Panel", "Add or Remove Programs", select "Add/Remove Windows Components", uncheck the “Indexing Service”, “MSN Explorer”, “Outlook Express” and “Windows Messenger” then select "Next" and "Finished.”
5. Setting a static size to your virtual memory file (page file) will reduce system overhead. XP will no longer have to dynamically manage the size of the file and it will help with fragmentation on your drive. When setting the static size of the file the generally accepted rule of thumb is to have this set to around one-and-a-half the size of your real memory. HOWEVER, any file larger than 512MB can and frequently will slow your machine down. Given that the OQO has less than 256MB real memory available setting this to a static file size of either 384 or 512 seems about right. As I would prefer to be told that I am out of memory rather than face additional overhead I opt for 384MB as both my minimum and maximum page file (virtual memory) size and thus make XP keep it static in size. To do this I went to "Start", "Settings", "Control Panel", "System", "Advanced" tab, in the "Performance" section select "Settings", "Advanced" tab, in the "Virtual Memory" section select "Change", change the "Initial Size" and "Maximum Size" to the same size then select "Set" and "OK".
6. My next optimization may not be too popular with many people. Since the OQO does not have a speaker (usable by XP) to system sound effects I elect to save the computer cycles used by the OS to play sounds to the sound device when it would be very rare that I would be able to hear them anyway. This adjustment will only affect system sounds (error beeps, boot songs, etc.) By leaving the sound devices and services alone when I perform an activity where I am likely to have headphones on (watch a movie, listen to music, play a game, etc.) the sound will work just fine. I just won’t be spending precious system resources to play beeps and sounds on the OQO that I will not likely be able to hear. To disable all Windows XP task sounds go to "Start", "Settings", "Control Panel", "Sounds and Audio Devices", select the "Sounds" tab, under "Sound Scheme" select "No Sounds". I saved this a Sound Scheme ‘None’ just so as to make it clear to myself later.
7. Feeling pretty good at this point, I decide that it is time to install some more software before I begin further optimizations. I have always great results with Executive Software’s Diskeeper software for optimizing NTFS file systems. So, now seems like a good time to get that installed. I purchased several licenses for 7.0 quite a while ago. There is a newer version 9.0 now out, but I have not seen anything substantially new enough to make me upgrade. After install and a bug-fix upgrade I run Diskeeper which replaces the default Windows Defrag utility. I select to do a boot defragmentation and check the following options: Directory Consolidation, Run CHKDSK before defragmentation, Reduce paging file fragmentation (now that we have it static), Defragment the MFT, Pause after defragmentation. I elect to do all of these things on the next reboot and of course then I reboot. The three pass CHKDSK goes clean and it reboots followed by the deep defragmentation process. All six stages of defragmentation complete without a hitch. The 6th stage took the longest, as it was the defragmentation of the paging file. After the reboot, I perform a standard defragmentation with Diskeeper.
8. Feels like time to run another boot timing and look at operating environment. Using the same criteria as the initial boot timing, overall boot time decreased by just over 12 full seconds. Environment stats from Task Manager have improved over the initial optimized reboot:
Available memory grew to: 127000 - 155400
System cache decreased to 86280
PF Usage dropped to 94.2 MB
Kernel Memory dropped to: 14680
Paged dropped to: 10516
NonPaged dropped to: 3164
While I believe that these are good results, this is far from scientific. All the key numbers moved in the right direction by substantial enough amounts to have made the effort appear worthwhile to me. It also goes to show that using the Task Manager to evaluate differences is problematic, as XP dynamically changes it’s optimizations based on the environment it detects.
While I will likely continue to work on optimization these changes seem like a really good start and I will now divert my attentions towards installing my applications and visual goodies to make my OQO my ‘own’. I will post another article on what I do to skin the OQO with minimal performance impact for as dramatic as effect as possible. After all something as cool as the OQO should not just be functional it should look really cool while doing it.
Guess you can tell I will NOT be running windows classic UI with an 8 bit color depth to optimize performance!
I hope some of this was useful to someone here. Special thanks go to *Captain* for the introduction to the TUU suite. Also, some of the techniques were leveraged from the following outstanding website link and so all credit for them should go there.
Cheers!