ICH9R Raid 5 Actually Very Good

[Owls]

For home and small server use it's actually better than most cheap PCI/PCI-e add on cards (under $300 of course).

Specs:
Q6600
Gigabyte P35C-DS3R
4x2GB DDR2 800
5x Seagate 160GB SATAII
1x Seagate 160GB PATA
Antec HE 550 Power Supply
Antec Sonata II case


Raid 5 with 16KB Stripe
Bench

 

[imported_wired247]

Why did you choose such a small stripe size?

I used this as a rudimentary guide when setting up my raid array
http://www.tomshardware.com/20...ling_charts/page7.html

Also how 13% CPU usage is a tad high, but if you aren't doing anything else CPU intensive it should be fine.

 

[Owls]

Originally posted by: wired247
Why did you choose such a small stripe size?

I used this as a rudimentary guide when setting up my raid array
http://www.tomshardware.com/20...ling_charts/page7.html

Also how 13% CPU usage is a tad high, but if you aren't doing anything else CPU intensive it should be fine.

The thing I've learned with RAID controllers is that certain stripe sizes are not good for all RAID controllers. ICH9R in particular had problems with slow writes using a 64kb or 128kb stripe with 4-5 hard drives. Many have reported that a smaller stripe with the ICH9R works best. When dealing with ICH9R I follow this rule: 3 drives = 64kb, 4 drives = 32kb, 5 drives = 16kb

As far as the CPU is concerned I got about 11% with HD Tach's Long Bench (Read and Write) and considering I have a Q6600 it's not a big deal, my primary goal of this server is to run my own CS: Source, DOD: Source, and maybe a TF2 server along with my Exchange 2007/Web server (I'm running Win2k3 x64)

 

[Rubycon]

Is this for multi user (file server) application?

If so, HD Tach really tells nothing that you need to know. Use IOMeter Text and run these tests on each stripe size to find the best value.

 

[Owls]

Running the IOMeter now.