View Full Version : 36gb raptor 6,6,6,6,6,6 partition stratigy
pulsarnx
08-09-04, 04:03 AM
so i have a 36gb raptor for a while now, wanna get more proformance.. heres what im thinking....
if i partition it into 6 equal parts.. for rounding sake lets say 6 - 6gb parts. (i know that the 36gb isnt really 36gb)
so ...
size 6gb - 6gb - 6gb - 6gb - 6gb - 6gb... hope you can follow...
partition 1 2 3 4 5 6
now knowing that the last part (6) is closest to to the inner tracks of the single platter in the 36gb raptor... if i were to install windows to that partition and then merger the other 5 parts, would i get faster proformance in boot up and basic windows operations ?
lemmi know what you think guys
pulsarnx@shaw.ca
Jaybo
I wouldn't mind seeing some tests done. Why don't you give it a try. I just formatted both my rigs with a single 36bg Raptor each, i would have tested it if i read this earlier, but i already installed everything, and that takes me forever and a day to do, 15 games isn't fun to install 2 times through.
You can partition it all you want but it doesn't work that way. It doesn't splice it like that. Even tho the partitions seem to be in order they are organized randomly, you won't really know which is on the outer edge.
SavageBasher
08-09-04, 04:38 PM
also, the fastest area of the hard drive is the outer edge, so the very first things installed on a drive should be the fastest.
Restorer
08-09-04, 09:40 PM
You can partition it all you want but it doesn't work that way. It doesn't splice it like that. Even tho the partitions seem to be in order they are organized randomly, you won't really know which is on the outer edge.
I disagree. If you've ever partitioned a drive with fdisk, you'll know that it shows partitions as block intervals. Usually the zero block is on the inside of the platter. So the fastest partition would logically be the last one. However, most drives have multiple platters, and I don't know how they handle partitions.
I'd like to see some proof that certain partitions are faster than others. I'm still under the belief the way an OS sets up partitions is different than how it's physically done on a drive.
Restorer
08-09-04, 10:07 PM
I'd love to see some proof of this too. Tell you what, as soon as my new drive comes in the mail, I'll test it this way and we'll see.
PunkRawk911
08-09-04, 10:24 PM
You want your OS on the first partition not on the last. The first partition is on the outer tracks of the drive, which have the fastest transfer rates.
Maybe you could format each of them and which ever one has the shortest time is the winner? :beer:
Hi, PunkRaw911 has it right. The first partition is on the outer edge and is the faster one. ... and very easy to test through any of the "common" STR test programs such as sandra, ATTO,...
Regards
FTC
Restorer
08-10-04, 07:55 PM
Well, maybe I was thinking of floppies or CDs when I said the first partition is on the inside. But I'm stil going to test on my 250GB drive, just to see how multi-platter drives do it. :)
Hi, you will not find suprises with multi-platter disks. Still the first partition goes to the outer edge, and takes as needed cylinders... (i.e for all available platters) as needed and beginning there (outer edge)...
You can partition it all you want but it doesn't work that way. It doesn't splice it like that. Even tho the partitions seem to be in order they are organized randomly, you won't really know which is on the outer edge.
Yeah I was just about to say that... it doesn't work like that.
ajrettke
08-10-04, 11:16 PM
Doesn't windows automaticaly move files that are accessed the most to the outer edge anyways? Seems to me you'd lose performance since games you play or programs you often use would be forced in further towards the center by using later partitions...whereas with just elaving it big and letting windows set them to the outside would be the fastest....anyone able to comment on windows ability to do this? I know I've heard that windows does this, but not sure how effecient it is.
I'm definatly interested in your results Restorer! I too have heard that partitions are physically placed somewhat randomly, against common logic that they should just go by the cylinder.
JigPu
shadowdr
08-11-04, 07:34 AM
Only the defrag utility in XP has the ability to move the files forward near the beginning of the drive.With all other OS's you must use a third party defrag tool to achieve this.To receive the most benefit from these utilities you must also set a static page file so that windows will not reserve empty space to expand the page file at the beginning(or where ever it decides to place it)of the drive.This is the only way that Norton's speed disk will move the page file to the beginning of the partition,because the files there are not expandable.Normally system files follow the page file,.dll's and shared files followed by applications,uninstall files,documents.
This is the reason it takes so long for the first defrag using third party software.Windows defrag before XP merely tried to keep fragmented files close to the original file location,probably because the slower drives of yesterday took forever to defrag.The best graphical representation of how files are treated is the Win 2000 defrag interface.Fragmented files are thrown randomly toward the end of the disk and then repacked according to the MFT.
This is one reason that partitioning strategies have been widely used before XP so that data could be accessed as quickly as possible.To see another Graphical representation of how your data is placed on the drive you can use Seqoiaview (http://www.win.tue.nl/sequoiaview/) .
This is my understanding of how defrag utilities work from all that I have read previously and how data is arranged on the physical disk.It may not be completely technically correct however.
Restorer
08-13-04, 01:44 AM
I got my 250 GB drive this morning and hooked it into my server, then found out my chipset is one revision behind being able to do 48-bit LBA addressing in Linux - so only 137 GB are recognized. Even so, I ran a few benchmarks, and even though I could only partition about half the drive, I'm already seeing big differences.
I made one 2.7 gig partition in the beginning of the drive, and another 2.5 gig partition as far to the end as I could go with LBA48. I formatted both partitions as ext2 for simplicity's sake (ReiserFS would probably be a bit faster overall, but that won't matter in comparing relative speeds). I benchmarked each partition three times with bonnie++. All the benchmarks agreed - the first partition reads 10 MB/s faster than the last (middle) one; 59112 K/s versus 48811 K/s. The write speeds were the same at around 32 MB/s. I also made some partitions between those and tested, and the read speed transitions smoothly.
So it appears that multi-platter drives work the same as a single-platter drive would be expected to work. I'll post explicit benchmark results as soon as I get a new ATA card that can handle this drive.
Restorer
08-14-04, 04:33 AM
Final results are in!
Today I put the 250 GB in my fast computer and ran more benchmarks. This time, with no chip restrictions or other bottlenecks, I ran the entire disk in quarters. Results and a writeup are at http://mattsmac.ath.cx/benchmarks/index.php.
It looks like this particular drive is especially affected by this kind of thing. As expected, seek time and OS-limited I/O (character files etc.) are largely unaffected, but sequential read/write goes down to half at the inner end of the drive. So pulsarnx, I think it would be a very good idea to put Windows in the first partition on its own.
On a sidenote, I'm impressed with my drive. I don't know if it's correct, but bonnie++ reports 3.3-3.9 ms seek times. It's advertised as "less than 9.3 ms."
vBulletin® v3.8.7, Copyright ©2000-2012, vBulletin Solutions, Inc.