View Full Version : Allocation Size Tweak-Somone wanna test this?
http://www.tweakxp.com/article37042.aspx
Anyone have a big ass HD that they can spare some space on? Like to try making their allocation size to stuff greater than 4kb? The advantage is speed. Disadvantage is more space is wasted, just curious how it will go though.
BTW I've read a page at Microsoft.com and this stuff is legit.
Bigger than 4kb? Whatever for?
yeah an Allocation Cluster Size bigger than 4KB, the advantage is performance gains, the disadvantage is more space wasted, but I'm still interested in the performance gains, esspecially if u have alot of room to waste.
BrutalDrew
07-21-05, 05:24 PM
Stop reading tweaks. Windows XP is for the most part is tweaked out of the box.
Stop reading tweaks. Windows XP is for the most part is tweaked out of the box.
And leave everything default and never ever try to squeeze out any more performance out of anything you buy... Ohhh wait this is an overclockers message board. :p Go ahead and tweak away......lol
Stop reading tweaks. Windows XP is for the most part is tweaked out of the box.
First off this isn't really a tweak, it's a MAJOR part of Windows and the NTFS partition, i shouldn't have called it a tweak but w/e, it doens't even have to do w/ the registry or anything like that
yeah seriously Brutal, stop bashing tweaks and show us some proof that they don't work, huh? I respect you as a knoweledgeable member of this forum but still, I don't support bashing like that w/ out proof, so no offense.
I always find i have to tweak any operating system to my needs really :confused:
If anyone has some benchmarks they want me to run, and suggests some files to use, I have a brand spankin new 200GB drive coming to me on Tuesday :)
I'll try a few sizes if you want, and do a little write up. Might be interesting.
I always find i have to tweak any operating system to my needs really :confused:
If anyone has some benchmarks they want me to run, and suggests some files to use, I have a brand spankin new 200GB drive coming to me on Tuesday :)
I'll try a few sizes if you want, and do a little write up. Might be interesting.
That would be EXCELLENT David! I guess the normal benchmarks like: 3DMark01SE, 3DMark03, 3DMark05, Super PI (1.4 mod), PCMark05, and Aquamark3 would be good ideas to start w/, I'm not sure if anything other than those would be real nessesary.
I'd also be rather interested in seeing benchmarks for the HD itself, like it's transfer speeds, the app is called HDTach.
I REALLY appreciate you stepping up to volunteer for this, thx!
I know with RAID 0 Arrays it's best to have 16k strip and set the Windows allocation size to 32k.
The main point is to have to set the windows Allocation size to the strip size set in the RAID BIOS times the number of drives.
Now doing this will make smaller files take up more space than they should. At 4k allocation sizes every 1k file is 4k so if it's set to 64k your overhead is huge.
What I will do:
- Create one 200GB partition on my new drive, with various allocation sizes: 4kB, 8kB, 16kB and 32kB.
- Install the following benchmarks on the partition and run them from the partition. OS is on another drive.
1) 3DMark2001SE
2) 3DMark2003
3) PCMark2004
4) Sandra
5) any others?
I will compare the following factors:
- Overall space actually used
- Percentage wastage
- Speed
Looking at Tues or Wed for this.
^^ Sounds like a good test. I suggest some Real world apps too. Antivirus and defragging would be nice as well. I know those things really stress on harddrive access.
Hold on David, how will running benchmarks from that other partition make it just like having the OS on that partition? I think I'm just confused, but will there be an OS on that 200GB parition, and ur just viewing from the outside OS or something?
I'd like to see how transfer speeds compare using HDTach.
Hold on David, how will running benchmarks from that other partition make it just like having the OS on that partition? I think I'm just confused, but will there be an OS on that 200GB parition, and ur just viewing from the outside OS or something?
I'd like to see how transfer speeds compare using HDTach.
OS is on another drive.
did ya get the drive David? It's thursday now, no thread. No hurry, just wondering.
Albigger
07-28-05, 11:50 PM
If you're benching transfer rates and the like in addition to HDTach you could also try ATTO or diskspeed I believe. ATTO tests different file sizes I believe but won't report random access times I don't think.
I benched it using an IDE ATA100 Seagate drive, used HDTach and Sandra. I am in the process of writing an article but here are the benchies anyway:
http://www.oc-rev.com/forums/images/hosted/Screenshot512.png
512 bytes:
http://www.oc-rev.com/forums/images/hosted/hdtach512.png
http://www.oc-rev.com/forums/images/hosted/sandra512.png
1024 bytes:
http://www.oc-rev.com/forums/images/hosted/hdtach1024.png
http://www.oc-rev.com/forums/images/hosted/sandra1024.png
2048 bytes:
http://www.oc-rev.com/forums/images/hosted/hdtach2048.png
http://www.oc-rev.com/forums/images/hosted/sandra2048.png
4096 bytes:
http://www.oc-rev.com/forums/images/hosted/hdtach4096.png
http://www.oc-rev.com/forums/images/hosted/sandra4096.png
8192 bytes:
http://www.oc-rev.com/forums/images/hosted/hdtach8192.png
http://www.oc-rev.com/forums/images/hosted/sandra8192.png
klingens
07-29-05, 06:43 AM
Your benches are useless since your benchmark tool is not suited to measure what you want to know
You want to know more about filesystem performance not raw disk performance. HD Tach measures the latter howeve. It won't use files on the disk, instead it reads directly from disk sectors, bypassing the filesystem. What you want is something like bonnie++ for benchmarking.
klingens
07-29-05, 06:45 AM
PS: unless you have a very very very special case, you never notice any improvement by changing clustersizes. Your results will be below the margin of error in 95%+ of all cases, so don't bother.
^^ He's right David. You're using Benchmarks that reflect RAW performance. Try some video editing benches.
Wow I just realized I never went back to see these results, thanks David.
This is interesting. I've always been interested in tweaking with windows, and just like David and jcq122, i've always found the need to tweak my xp installation.
Lookin forward to these results david.
BrutalDrew
01-03-06, 08:00 AM
And leave everything default and never ever try to squeeze out any more performance out of anything you buy... Ohhh wait this is an overclockers message board. Go ahead and tweak away......lol
Overclocking increases performance. Tweaking Windows, for the most part, does not. Tweaking is passe. It was OK with W98 and those crappy OSs, but with XP other than simply customizing the looks there is nothing much to do.
Overclocking increases performance. Tweaking Windows, for the most part, does not. Tweaking is passe. It was OK with W98 and those crappy OSs, but with XP other than simply customizing the looks there is nothing much to do.
I wouldn't even call this a tweak, it's much bigger than that, it's messing w/ the actual file system and file sizes.
futura2001
01-04-06, 09:50 PM
Overclocking increases performance. Tweaking Windows, for the most part, does not. Tweaking is passe. It was OK with W98 and those crappy OSs, but with XP other than simply customizing the looks there is nothing much to do.
Ummm, no. There is a lot that can be done to Windows to improve performance. For instance, explorer.exe is very bloated and contains resources that haven't been used since previous versions and resources that quite frankly are never used. By removing as much of the bloat, I was able to lower the memory usage of explorer.exe from 23mb to just over 2mb. This may not seem like a lot in an age where 1024mb of ram is the standard, but this combined with a multitude of other tweaks was able to lower my memory consumption at startup from 150-200mb on a fresh install with nothing but all current windows updates installed, to 70-90mb when tweaked out.
Furthermore, allocation size tweaking does have its benefits. I do a lot of digital art in addition to my day job, and when drawing a poster sized 300dpi image, file sizes can quite easily exceed the gigabyte and above range. As you can imagine, at the default 4kb size, these files get fragmented exceedingly quickly; I've seen such files have six digits of fragments. So, setting a volume to a 32 or 64kb size and using this solely for storing massive files results in a performance increase and a lower level of fragmentation. And by using the volume only for larger files, I don't notice 64kb one way or the other.
threeme2189
01-05-06, 02:19 PM
interesting....
Ummm, no. There is a lot that can be done to Windows to improve performance. For instance, explorer.exe is very bloated and contains resources that haven't been used since previous versions and resources that quite frankly are never used. By removing as much of the bloat, I was able to lower the memory usage of explorer.exe from 23mb to just over 2mb. This may not seem like a lot in an age where 1024mb of ram is the standard, but this combined with a multitude of other tweaks was able to lower my memory consumption at startup from 150-200mb on a fresh install with nothing but all current windows updates installed, to 70-90mb when tweaked out.
Furthermore, allocation size tweaking does have its benefits. I do a lot of digital art in addition to my day job, and when drawing a poster sized 300dpi image, file sizes can quite easily exceed the gigabyte and above range. As you can imagine, at the default 4kb size, these files get fragmented exceedingly quickly; I've seen such files have six digits of fragments. So, setting a volume to a 32 or 64kb size and using this solely for storing massive files results in a performance increase and a lower level of fragmentation. And by using the volume only for larger files, I don't notice 64kb one way or the other.
Wow thanks for that excellent example and information futura, didn't know most of that.
BTW, where can I find these tweaks for slimming down explorer.exe? I've never heard of that before, thanks.
BrutalDrew
01-05-06, 03:53 PM
Ummm, no. There is a lot that can be done to Windows to improve performance. For instance, explorer.exe is very bloated and contains resources that haven't been used since previous versions and resources that quite frankly are never used. By removing as much of the bloat, I was able to lower the memory usage of explorer.exe from 23mb to just over 2mb. This may not seem like a lot in an age where 1024mb of ram is the standard, but this combined with a multitude of other tweaks was able to lower my memory consumption at startup from 150-200mb on a fresh install with nothing but all current windows updates installed, to 70-90mb when tweaked out.
Real-world performance is what is important, not the amount of memory allocated at initial boot up. Disabling services lowers the amount of memory used, but does that mean it improves performance? No, it does not because memory will be reclaimed as needed. Also, RAM is so cheap nowadays you would be better off buying more RAM then wasting your time tweaking Windows. If you have 256MB of RAM it may help a little, but it still won't negate the fact that you need more RAM.
There is not many things you can do to Windows XP to actually increase real-world performance.
futura2001
01-05-06, 03:58 PM
To slim explorer down, first you are going to have to download a useful program by the name of ResHacker. It allows you to view and edit resources on unencrypted files. Make a backup of explorer.exe, open it up in ResHacker, and you should see a long list of available folders. First off, virtually all the icon groups can be deleted as Windows pulls icons from Shell32.dll and not explorer. Most of the bitmaps graphics can be removed, although keep the ones pertaining to your Windows. There is a lot to lobotomize, and I am at work right now so I can't look at my lists and files, but removing the icons and bitmaps is about 2/3 of the work right there...
As for actually getting Windows to run your edited file once you get done modifying it, you have to disable Windows File Protection, which is a bit more complicated. The easiest way to do it is to download the XPlite or 2Klite demos. They won't give you all the options, and I don't recommend doing anything else with the program as removing Windows components through their wizard has an uncanny habit of breaking things.
Finally, you have to replace explorer.exe. Easiest and safest way to accomplish this is, if you have two OSes on one system or a knoppix CD, to restart and boot from the other OS, make a backup of the unedited explorer.exe and replace the system file with your edited version. This has the benefit of making it relatively easy to fix what you have done if for some reason you messed it up and your system refuses to boot.
Two other options are to either use a program called Replacer to replace your system files, or in the case of explorer.exe, shut down explorer.exe from the taskmanager, replace the file, and restart explorer.
Mind you, it is pretty easy to break explorer if you aren't sure what you are doing, so always make a backup, and use method one if you can.
Did you notice any gains yourself? How much memory did/does the system have?
futura2001
01-05-06, 07:16 PM
The system started up faster after doing this. Not that I care, I try to restart once a week whether I need to or not. I keep my restarts on the same circadian clock I shave by...
At the time, I was drawing a 9600x11200 pixel image with 1 gig of ram total in my system. Taskman was showing 900mb+ of memory usage, most of it going to Painter 7. So yeah, by disconnecting myself from the internet and shutting down all extraneous processes and programs, I was just barely able to keep the file in physical memory at all times. Definitely helped in that case, although this is a very isolated and anecdotal case. Since then, it hasn't made much of an impact. Mainly because after trimming Windows down to the minimal, I added a bunch of custom graphics and whiz-bang directly into the OS itself (somewhere in the desktop screenshots I have a post with my overmodded 2k) and bumped that memory usage back up to the 100-150 mark...
Although, at PAX'05 I brought out the minimalized system files/settings because the Pogo Linux people were bragging about Linux memory usage on their boxes and I just happened to be standing there, so when they said something along the lines of "only 96mb memory usage, I'd like to see XP do that" I just had to show them a fully functional Win2k box running at under 80mb...
vBulletin® v3.8.7, Copyright ©2000-2012, vBulletin Solutions, Inc.