Recommended defragmentation program for Windows XP?
November 9, 2004 6:41 PM Subscribe
Defrag this! What defragmentation program would people recommend for Windows XP? The one which came standard with the OS or a third party?
Yes defragging helps (on FAT filesystems, not sure how well the NTFS code does prevent fragmentation), but the religious every week defragmentation routine some people have is purely a placebo, only extreme fragmentation will cause significant slowdowns.
posted by fvw at 6:59 PM on November 9, 2004
posted by fvw at 6:59 PM on November 9, 2004
Disk Keeper. Executive Software, I believe.
posted by adampsyche at 7:06 PM on November 9, 2004
posted by adampsyche at 7:06 PM on November 9, 2004
NTFS fragments somewhat more slowly than FAT, but then, so does everything. It does, however, benefit from defragmentation. Most filesystems do, if it's done intelligently and not just a dumb consolidation of free space.
The most flexible software I've seen to do this job is oodefrag. Another popular product is Diskeeper, a crippled version of which is shipped with Windows as the "built in" defrag. Diskeeper's a good choice for enterprise networks because of the management facilities, but as an actual defragmenter it's only so-so.
posted by majick at 7:06 PM on November 9, 2004
The most flexible software I've seen to do this job is oodefrag. Another popular product is Diskeeper, a crippled version of which is shipped with Windows as the "built in" defrag. Diskeeper's a good choice for enterprise networks because of the management facilities, but as an actual defragmenter it's only so-so.
posted by majick at 7:06 PM on November 9, 2004
PerfectDisk
is very good. I use it on xp pro. Read user comments on link.
posted by JohnR at 7:10 PM on November 9, 2004
is very good. I use it on xp pro. Read user comments on link.
posted by JohnR at 7:10 PM on November 9, 2004
I'd like to see some actual performance statistics pre- and post-defragging. I'm not convinced it makes a significant difference in most cases. (Executables are not loaded sequentially from the disk anyway.)
posted by mcguirk at 7:27 PM on November 9, 2004
posted by mcguirk at 7:27 PM on November 9, 2004
The only performance increase I've ever experienced from defragging is that it allows you to defrag faster.
posted by ChasFile at 7:31 PM on November 9, 2004
posted by ChasFile at 7:31 PM on November 9, 2004
I use to defrag all the time. I have never seen any difference, except the time loss to defragging. Perhaps it helps on huge hard drives, but on a smaller one I've had zero performance increase.
posted by justgary at 7:37 PM on November 9, 2004
posted by justgary at 7:37 PM on November 9, 2004
Response by poster: Are there any negatives to Norton Speed Disk?
posted by phyrewerx at 8:24 PM on November 9, 2004
posted by phyrewerx at 8:24 PM on November 9, 2004
Response by poster: And thanks for all the help mefites!
posted by phyrewerx at 8:24 PM on November 9, 2004
posted by phyrewerx at 8:24 PM on November 9, 2004
The one that comes with windows is reasonable.
All other alternatives tend to be faster.
Degragging is such a simple operation, I doubt you'll find the paid alternatives do much more than get the job done quicker.
Negatives to Norton Speed Disk: You have to install norton utilities.
posted by shepd at 9:04 PM on November 9, 2004
All other alternatives tend to be faster.
Degragging is such a simple operation, I doubt you'll find the paid alternatives do much more than get the job done quicker.
Negatives to Norton Speed Disk: You have to install norton utilities.
posted by shepd at 9:04 PM on November 9, 2004
John Muster had some interesting thoughts about this. He quoted a study that showed that due to the brief instant it takes for a hard disk's head to read one sector of data and become ready to read the next sector of data, defragmentation is actually harmful. A fully defragmented file's pieces are all in a row, and piece number one might not have cleared the head before piece number two has rotated to directly under the head. In this case, the head is still busy with piece number one just as piece number two rotates underneath it, and the platter has to spin all the way around again for piece number two to come under the head, once the head is ready. According to this theory, randomly distributed data is, overall, easier for a hard disk drive to read than completely defragmented data. In fact, if there is an optimal way to write data to a hard disk, it probably has nothing to do with keeping files in contiguous pieces, and more to do with fragmenting them according to some optimized pattern.
posted by scarabic at 10:52 PM on November 9, 2004
posted by scarabic at 10:52 PM on November 9, 2004
I'll second PerfectDisk, excellent software.
posted by TungstenChef at 11:21 PM on November 9, 2004
posted by TungstenChef at 11:21 PM on November 9, 2004
If defragging doesn't work, can the techies recommend other processes us noobs can run to keep our pc's running better? My computer is getting pretty slow and I would love a general utilities recommendation.
posted by vito90 at 12:58 AM on November 10, 2004
posted by vito90 at 12:58 AM on November 10, 2004
> If defragging doesn't work, can the techies recommend other processes
> us noobs can run to keep our pc's running better?
Just spent a day giving my daughter's pc a thorough cleanout. It's less than a year old and was getting slow as cold peanut butter.
1. Make sure you have a good amount of space left on whatever drive has your virtual memory (pagefile.sys). It can really slow a system down if you're set for, say, a 512-mb pagefile and the drive it resides on has 30 mb free.
2. Run Ad-Aware and Spybot (with up-to-date pattern files.) Spyware/adware eats up performance, besides being inherently evil due to privacy invasion.
3. Run a full antivirus system scan, using an up-to-date virus signature file.
4. Open the task list, write down the name of every running process, and look up everything you don't recognize on extremely helpful sites like this one. I found msbb.exe running; this is n-case 180search assistant, a browser hijacker that's notorious for slowing browsing to a crawl. Neither Ad-Aware nor Spybot picked it up.
5. Reduce the things in your system tray to a minimum. Most of 'em just sit there turning electricity into heat (and causing blue screens.)
6. I vote yes on defrag. If nothing else, it makes things look more orderly and merits doing the same as blowing the dust out of your case whenever you have it open even if you're not having overheating problems. I've seen the theory, described above, that fragmentation actually hurts performance, and I just can't believe that modern HDs can't read as fast as they can spin. The manufacturers do test these things under real-world conditions before they go out the door, and pick the speed that gives best throughput.
As for which software, I have to say I like Norton. I find Windoctor useful: letting programs like Ad-Aware fix the problems they find tends to leave stuff like shortcuts and registry keys that point to nothing; Windoctor (part of Norton Utilities) finds and fixes these and I consider it the finishing touch to a malware sweep. And it's not like it costs a bunch of money: I found full Norton Systemworks, including Utilities, for $6.50 new on ebay. This was the OEM version with no pretty box and no printed manual but I can live without a box and the manual was on the CD.
posted by jfuller at 3:58 AM on November 10, 2004
> us noobs can run to keep our pc's running better?
Just spent a day giving my daughter's pc a thorough cleanout. It's less than a year old and was getting slow as cold peanut butter.
1. Make sure you have a good amount of space left on whatever drive has your virtual memory (pagefile.sys). It can really slow a system down if you're set for, say, a 512-mb pagefile and the drive it resides on has 30 mb free.
2. Run Ad-Aware and Spybot (with up-to-date pattern files.) Spyware/adware eats up performance, besides being inherently evil due to privacy invasion.
3. Run a full antivirus system scan, using an up-to-date virus signature file.
4. Open the task list, write down the name of every running process, and look up everything you don't recognize on extremely helpful sites like this one. I found msbb.exe running; this is n-case 180search assistant, a browser hijacker that's notorious for slowing browsing to a crawl. Neither Ad-Aware nor Spybot picked it up.
5. Reduce the things in your system tray to a minimum. Most of 'em just sit there turning electricity into heat (and causing blue screens.)
6. I vote yes on defrag. If nothing else, it makes things look more orderly and merits doing the same as blowing the dust out of your case whenever you have it open even if you're not having overheating problems. I've seen the theory, described above, that fragmentation actually hurts performance, and I just can't believe that modern HDs can't read as fast as they can spin. The manufacturers do test these things under real-world conditions before they go out the door, and pick the speed that gives best throughput.
As for which software, I have to say I like Norton. I find Windoctor useful: letting programs like Ad-Aware fix the problems they find tends to leave stuff like shortcuts and registry keys that point to nothing; Windoctor (part of Norton Utilities) finds and fixes these and I consider it the finishing touch to a malware sweep. And it's not like it costs a bunch of money: I found full Norton Systemworks, including Utilities, for $6.50 new on ebay. This was the OEM version with no pretty box and no printed manual but I can live without a box and the manual was on the CD.
posted by jfuller at 3:58 AM on November 10, 2004
> that fragmentation actually hurts performance,
Left out the de...
posted by jfuller at 4:10 AM on November 10, 2004
Left out the de...
posted by jfuller at 4:10 AM on November 10, 2004
The built-in defrag is fine. I really don't know who needs all the scheduling and extended stuff - fragmentation really isn't that important, on NTFS and most Linux filesystems at least.
The only time defrag has ever made a noticeable difference for me was when streaming a barrowload of live video data onto a hard disc that could barely keep up. Keeping bigger available empty blocks contiguous on the disc helped in this case.
If you're not doing anything like that, once a year is fine IMO. Leave it chugging away when you go out on New Year's.
posted by BobInce at 6:38 AM on November 10, 2004
The only time defrag has ever made a noticeable difference for me was when streaming a barrowload of live video data onto a hard disc that could barely keep up. Keeping bigger available empty blocks contiguous on the disc helped in this case.
If you're not doing anything like that, once a year is fine IMO. Leave it chugging away when you go out on New Year's.
posted by BobInce at 6:38 AM on November 10, 2004
I'll second jfuller's points, especially number 4. For an easier way to check what is running, I like this program.
posted by neurodoc at 7:41 AM on November 10, 2004
posted by neurodoc at 7:41 AM on November 10, 2004
scarabic: The theory that consecutive sector arrangements would hurt performance was true. Back in the days of ST-225 MFM disks without interleaved sector numbering. That's since been fixed with proper low-level interleaving. In any case, the performance hit of seeking far overshadows even the most suboptimal rotational latency. Proper defragmentation means fewer seeks, which translates to faster performance.
Even the argument that paged executables aren't read sequentially doesn't hold much water. Since, oh, somewhere around 1989, just about all block I/O subsystems have employed a combination of both read-ahead and caching, and since the mid-90s drives themselves have been performing read-ahead.
It's not uncommon for a hard disk to slurp all or part of a track into memory as the disk goes by, on the off chance that it will be asked for.
Good defragmentation is about keeping files contiguous to reduce seeks, and keeping them as near as possible to their inode (or whatever the local filesystem equivalent is). Modern filesystems already do this, with varying degrees of success. In the case of NTFS, with very little success at all.
At this point it's indisputable that defragging a heavily fragmented NTFS filesystem will improve performance. But it's true that it isn't something you need to waste time on every week. Most users could get away with a quarterly or semiannual defrag of local volumes.
posted by majick at 8:19 AM on November 10, 2004
Even the argument that paged executables aren't read sequentially doesn't hold much water. Since, oh, somewhere around 1989, just about all block I/O subsystems have employed a combination of both read-ahead and caching, and since the mid-90s drives themselves have been performing read-ahead.
It's not uncommon for a hard disk to slurp all or part of a track into memory as the disk goes by, on the off chance that it will be asked for.
Good defragmentation is about keeping files contiguous to reduce seeks, and keeping them as near as possible to their inode (or whatever the local filesystem equivalent is). Modern filesystems already do this, with varying degrees of success. In the case of NTFS, with very little success at all.
At this point it's indisputable that defragging a heavily fragmented NTFS filesystem will improve performance. But it's true that it isn't something you need to waste time on every week. Most users could get away with a quarterly or semiannual defrag of local volumes.
posted by majick at 8:19 AM on November 10, 2004
Nice task list utilities, jfuller and neurodoc. Cheers!
posted by magullo at 10:27 AM on November 10, 2004
posted by magullo at 10:27 AM on November 10, 2004
NTFS is pretty good at keeping itself unfragmented. The only time you should have to defrag is if you repeatedly drop below say10-15% free disk space or if your free space is smaller than your windows determined swap size. One a year is probably suffiencent for most people.
To add to majick's debunking remember that where ever the disk would have to seek to once it got there the head would have to wait on average 1/2 a revolution for the data to appear. You'd be alot better off reducing the speed of your disk than depending on randomized data access.
posted by Mitheral at 10:34 AM on November 10, 2004
To add to majick's debunking remember that where ever the disk would have to seek to once it got there the head would have to wait on average 1/2 a revolution for the data to appear. You'd be alot better off reducing the speed of your disk than depending on randomized data access.
posted by Mitheral at 10:34 AM on November 10, 2004
Run Ad-Aware and Spybot (with up-to-date pattern files.)
Beware: On all three machines I've tested, attempting to grab the latest Ad-Aware 6 reference file (as of 2004.10.26) via the software update hoses the Ad-Aware install. You can fix this by manually downloading the reference file from the sidebar here and replacing the corrupted file.
posted by Danelope at 10:42 AM on November 10, 2004
Beware: On all three machines I've tested, attempting to grab the latest Ad-Aware 6 reference file (as of 2004.10.26) via the software update hoses the Ad-Aware install. You can fix this by manually downloading the reference file from the sidebar here and replacing the corrupted file.
posted by Danelope at 10:42 AM on November 10, 2004
This thread is closed to new comments.
posted by haqspan at 6:49 PM on November 9, 2004