Unsolved
This post is more than 5 years old
8 Posts
0
5021
June 23rd, 2004 13:00
RAID 0 file chunk size
I've recently installed RAID 0 on my 8300. The only thing I can't decide on is proper file chunk size for optimum performance. I've set up RAID with both a 64k and a 128k file chunk because most of what I've read reccomends this. I typaclly use my system for gaming, internet, etc. My benchmarks for 3D03 and for Aquamark are both very close with both settings. But the confusing part is when I benchmark my hard disks. With Sandra's test it seems that get a much higher score with the 128k chunk size (89mb/s as opposed to the 64k which is only76mb/s). The 128k also wins out on Passmark's Performance Test V5.0 with disk read at 74.1mb/s as opposed to the 64k at only 42.4. The write gave a similar result with the 128k writing at 89.2mb/s and the 64k at only 62.2mb/s. The oddball is the HDTach test. This test gives me an average transfer of 64.2mb/s at 64k and only 51.2 with 128k. Which way is the best way to go? Do these tests mean anything in real world applications? Is one test more reliable testin RAID 0 than the others? Any advice,please.
No Events found!
-WYSIWYG-
85 Posts
0
June 23rd, 2004 14:00
TJGIII,
dude that was a huge post. I learned a little about chunk sizes when working on linux drives. Though a little different, the net result is about the same.
Don't think about what performance you are going to get, but rather what the application of the computer will be. If you are going to be using (very) large file sizes, i would stick with the 128k chunk. If you are storing large amounts of information over a large number of smaller files, then stick with the smaller chunk size. This has been a good rule for me.
Hope this helps.
^_^ Cheers!
TJGIII
8 Posts
0
June 23rd, 2004 14:00
-WYSIWYG-
85 Posts
0
June 23rd, 2004 15:00
I'll put it this way .... I've never seen the HDTach Test. The others are more gaming oriented. So I can't really compare them. If you have access to Sandra rather than 3D mark 03 or AquaMark, you can target it to more business like applications.
I also google'd for some performance specs. nothing hard or factual. I would just try and find a benchmark that best represents realworld application and use that. I've gone as far as to run the actual app for some testing. Sometimes you get ... "WHAT?? No Difference!!" (ie noticable difference), and other times you get a more obvious winner.