Question About IDE Raid

[cbman]

Hey Guys... Long Time No Speak

I finally got rid of my old system... and built a new one... but I'm having a bit of a quandry...

I got a Gigabyte GA-7N400 Pro2 with all the goodies and I'm using the GigaRaid controller with two Samsung Spinpoint SP0802N 80GB 7200 RPM hard drives as my Raid 0 Boot Drive... now also in the system I have a single 80GB WD Caviar 800BB 7200RPM and a Maxtor D740X 80GB 7200 RPM Hard Drive...

(As an aside the First Samsung Drive By Itself Preformed Better than the Maxtor and about the same as the WD on the old system)

Through a bunch of testing it seems as though the Raid drive performs slower than the two Single Drives alone. This is my first time using RAID on my own system and I'm stuck as to what might be going wrong... if you guys need numbers let me know...

As for my new system... its pretty damn sweet.

AMD Barton 2500+@2GHz
1.5 GB DDR400 (Running In Dual Channel)
Gigabyte GA-7N400 Pro2
MSI GeForce FX5600@5800 Speed
2 x Samsung 80GB 7200RPM
1 x Maxtor D740X 80GB 7200RPM
1 x WD800BB 80GB 7200RPM
LiteON 52x32x52 CDRW
Asus 16x DVD
LG 4040B DVD-RW
SoundBlaster Audigy 2

And all the extra crap that goes along with it... (I love my new job)

[dZeus]

hey cbman, long time no see indeed...

RAID0, aka striping, only improves your sustained throughput rate (STR). Normal desktop use is not really limited by the STR much, and maybe that the RAID controller uses the CPU enough to notice the difference in workstation environment.

Examples of use that does benefit from higher STR is copying of large files, and video editing.

Also keep in mind that running in RAID0 increases the risk of losing all your data on both disks exponentially over running with 1 drive with the size of the RAID0 volume. I don't think RAID0 is worth that risk, unless if you really use your system in such a way that higher STR gives a lot of benefit.

[Ribbit]

I don't know much (actually anything) about your RAID controller, but I'm very tempted to place the blame there. I know that in the past there have been RAID controllers with abysmal performance (Highpoint).

And I'd like to emphasise what dZeus said - you'd really be surprised how little difference RAID0 makes in real life, and how few cases there are where it does make a difference. I wouldn't bother with it.

[cbman]

Hey Thanks For The Replys... But I'm actually doing some Video Work For My Little Brother and on the side and am working with some pretty large files... (Most bigger than a gig)

If anyone can find someone who's benched the GigaRaid controller online anywhere... if they could post a link that would be great...

As it is here are some HDTach 2.70 jpg's of all my drives Performances... just for referance... maybe that would help.

First Is The Samsung Raid

[cbman]

Next Is The Single Maxtor

[cbman]

Then The Single Western Digital

[cbman]

Last is the Single Samsung Drive *Note this was with HDTach 2.61 but the results seems to coincide with the 2.70 version pretty well*

[cbman]

Oh and I found out that the raid controller used for ATA is the

ITE IT8212

I'll do some searching to see if I can find an add in card that uses this chip and the reviews on it

[DentyCracker]

Nice burst rate there on the maxtor. Something doesn't seem right wth the raid. I was thinking that you might want to set the raid as the capture drive and boot from one of the others, but Doc Mordrid is probably in a better position to give you advice on that

[cbman]

Thanks... Yeah the Maxtor has a pretty nice burst rate... but HDTach 2.70 rates it lower than the 2.61 version did... anyways... did some digging... the ITE controller has the newest Drivers and here are the specs on the device itself

http://www.ite.com.tw/pc/brief_it8212f.htm

Overall it doesn't look that bad... but I did a check in sisoft 2004 and this is the info it gave me... sorry about size of the pic but I wanted to get it all on the screen.

[dZeus]

don't use HDTach or Sisoft sandra for measuring HDD performance. They're well-known to give rather meaningless results.

I think that it's best to use the ATTO drivebench, though even that is not perfectly comparable to real-world performance.

[rugger]

I think I know why your raid is performing so poorly. It may have something to do with PCI performance.

If your PCI bus cannot keep up with the drives, then the drives will be unable to read the sectors as they pass under the drive heads. This will end up forcing the drives to make extra revolutions to read sectors and serious STR performance problems result from that.

With 2 modern drives, the STR rate of your RAID array would range between 120 and 80megabytes per second, which many PCI chipsets simply can't handle.

Given that your burst rate on your RAID array was only 70megabytes per sec, this does sound like the problem. I had the same problem with bandwidth when I connected a 7200.7 drive to an old LX motherboard and ATA-33.

I would simply run them as 2 separate drives.

One more stupid question, do you have both drives on the ITE raid controller their own channels ... ie, are the both on different cables?

[cbman]

Hey

Both Drives Are Masters On Their Own Channel... as for the RAID saturating the PCI bus... it very well could be the problem if that's how they have it hooked up... I'll check.

Oh and when I get home from work I'll see about running ATTO

[rugger]

The ITE controller, even if it isn't a PCI card, is hanging off the PCI bus, because it isn't intergrated into the southbridge.

There is no way to improve performance if the PCI bus is saturated, all you can do is disable RAID and use the drives individually.

[cbman]

Well I still want to do some more research but so far that looks like what the problem is... that might also account for the spiking in the data transfer rate in the raid benchmark

I also have a PCI IDE Raid card sitting at home... I might plug that into the system and hook the Raid Drives up to it to see if it saturates the bus that way as well... if it does then that for sure is where the problem lies... if not... its off to more digging for information.