slower HHD scores with RAID 0 than non RAID??!!

[kanuuker]

I recently did a clean install of Windows XP Pro on a WD1200JB and in benchmarking I got a HDD score of 1079 in PCMark2002. Just yesterday I received 2 more WD1200JB drives and decided to do another install but using 2 drives in a RAID 0 setup. I then ran PCMark2002 again to see how much quicker my HDD score would be. Turns out that not only is it not quicker, but it's a fair bit slower. My HDD score was 679! I ran the test again and got a similar score, so I booted back into the single hd installation and ran the test again. My score was 1067!

What gives? Why is my score significantly slower on a RAID 0 setup than on a single drive setup? All the drives are the same model. I am running the single drive on the 1st IDE channel and am running the RAIDed drives on separate channels on IDE 3 and 4. I am using a Gigabyte GA-7VAXP Ultra board (KT400) with the Promise MB Fastrack RAID controller.

 

[kanuuker]

Never mind, Dumb *** at work

God, I'm so stupid sometimes. All I had to do was install the RAID driver. Now my score is better, something like 1115 (I didn't write it down). However, how much of an increase should I expect to see from RAID 0? I thought it would be more.

 

[K1ll1nT1m3]

I would think it will be almost twice what you got with a single drive . I was getting over 2000 in PCMark2002 with a SATA RAID0. I was getting around 1200 with my single Maxtor SATA. RAID0 is really best if you move alot of big files or install OSs alot.

 

[Deathknight]

What stripe size did you specify when you created the array? Stripe size can have a significant impact on performance

 

[kanuuker]

I just used the auto setup in the BIOS/RAID setup. It gave me a couple of options: mirror, and stripe. Within stripe it gave me options of desktop, av editing, and server. I chose desktop. It didn't say what cluster size it used, I assumed 64k. The main reason I want the speed of raid is for games, but it's nice for overall system speed too so I figured that desktop would be my best option.

 

[stan03]

i think you should try 16K

 

[kanuuker]

I went in and checked and it is 64k size clusters. I also ran pcmark again and my score 1057 - roughly the same scores as a single drive. I don't get it.

When I click on my drive properties it shows up as 240gb (120 x 2), so it's working. What about master/slave set up? Since I have them on separate channels I have them both set as single masters (in fact I had to do it that way, it wouldn't work when I had one set as master and the other set as slave).

Also, when I installed windows I did the quick format. What's the difference between the quick and the regular and could they affect my performance?

 

[CrashOveride]

You might want to try other benchmarks, perhaps that one does a certain test combination that does not show the perks of RAID 0 well? (or maybe not, just a thought )

 

[Sonny]

Go to THIS THREAD & post your scores using ATTO, way more reliable than Sandra or PCMark for RAID/HDD testing.

 

[kanuuker]

Okay, I ran ATTO w/ length at 32 and depth at 10. I ran it on both my single drive setup and my raid setup. I've never used ATTO before but the results seem curious to me. Here's a sample.

Single Drive
Sample Size Read Write
2 1654 1721
32 17867 16457
64 33058 27280
512 45140 43389
1024 47259 44640

RAID 0
2 15909 34829
32 43406 41028
64 42744 52663
512 43862 49104
1024 44640 49657

As you can see, there is a huge difference with the sample at 64mb and under, but over that there isn't much difference. I seem to hit a wall over 64mb.

I wasn't able to check the results on the link you sent me as none of the graphics or links worked (I'm using a fresh install of windows and haven't set anything up yet. Also I normally use Opera so I don't know the IE settings, if that's even the case).

Anyway, how do my results read to you? Are they normal?


edit>> ****, now that I've posted this reply it appears that my 'table' didn't line up. Just note that there are 3 columns, 1st is sample size, second is write, third is read.