RAID5 got slower puzzle

Yammer · 10 Jan 2013 at 11:07

Hi there, overclockers.

I'm using a Titan Lite (Asus P5Q Deluxe) I bought in 2008. It has the same Core2 Duo (2.66GHz) and 4GB Corsair TWIN2x memory running at 1066MHz. Windows 7 has been installed since 2009. It uses the Intel P45 Express chipset with an ICH10R controller.

The boot drive is a 500GB WD Caviar Blue, and the data is on a 3-drive RAID 5 of 500GB WD Blacks. I bechmarked the drives with HD Tune back in 2010, and again just now. The boot drive benchmark looks almost identical, but the RAID benchmark is *significantly* down.

2010 Nov

2013 Jan

That's quite a big performance hit. I am at a loss to explain this. I used the same copy of HD Tune both times, and disabled anti-virus and other obvious unnecessary background stuff.

Of course, I have to look at what has changed in the meantime. I can think of only three things:
1. I overclocked the CPU in May 2011, from 2.66GHz to 3.55GHz. RAM still runs at 1066MHz.
2. One of the RAID drives was replaced in Nov 2011.
3. I upgraded the ATI Radeon HD5670 with an HD7850 in Nov 2012.

Additionally, I removed a third drive from the ICH10R (a WD Caviar Green), but I don't see how that could slow anything down. Also, Intel RST and Chipset drivers have been updated to current versions, as and when they appeared.

Can anyone suggest what might be slowing things down before I start hacking away at BIOS settings and drivers, please?

Yammer · 11 Jan 2013 at 11:55

Well, for anyone who's interested:
this morning I altered the Ai Tweaker settings on the motherboard to bring the FSB back to defaults. I stopped short of defaulting the whole lot, because I wasn't sure what would happen to the RAID settings. So there's some doubt that I may have missed a non-default setting or two.

Anyway, there was no effect on the RAID volume benchmark. So, I'm looking at the RAID drives/settings/drivers, or the graphics card (which I'm fairly sure is irrelevant, frankly).

Please feel free to chip in with any suggestions or observations.

zakblood · 11 Jan 2013 at 12:14

no idea but that's a massive change, hwy did you replace a drive?

all drives same age?

no driver update or bios changed like a rom update etc?

Yammer · 11 Jan 2013 at 13:19

The old drive failed, and I was sent a new one by WD. I'd already read-up on WD and RAID, and made a point of asking them for a compatible Caviar Black to go in a RAID 5 set. It went in and rebuilt ok.

It's not exactly the same, but I have to assume that it's compatible, unless told otherwise. The two original drives are WD5001AALS-00E3A0 (firmware 05.01D05), and the new drive is WD5001AALS-00L3B2 (firmware 01.03B01).

Actually, typing all this out is making me think I should post in a WD forum too.

I have updated the BIOS in the past, but I'm not sure when. I suspect it was before I built the RAID, because I would have been very nervous about applying one afterwards with a working array. Put it this way, it's not running the latest BIOS now, as I don't want to risk it, and there's nothing in the changelog to suggest I need it.

I have updated the Intel Chipset and Rapid Storage Technology software at least twice since creating the array. I even updated RST yesterday (there was a new version in December), after making the above post, but it made no difference.

JonJ678 · 11 Jan 2013 at 17:33

Defrag the data? The other thing that is likely to have changed since 2009 is the amount of data on the drives.

On a sorrier note, one of the drives starting to die would also hurt the speeds. Have you got a backup of the data?

Yammer · 11 Jan 2013 at 17:53

I regularly defrag the volume.
Yes, there's probably double the data now.
I guess I need some decent software for analysing the RAID. RST doesn't really tell you much.

Yammer · 12 Jan 2013 at 12:04

Apparently, WD doesn't support desktop drives in RAID5. It's possible that their replacement drive is causing the slowdown, but unless they help, I'm not going to know.

Interestingly, looking at the figures, the max speed has dropped by almost exactly 25%, and the min speed by almost exactly 50%. It looks mathematically significant, but I have no idea why.

If I get nowhere, I'll probably end up breaking the array and trying RAID 0 or 1 instead. What a shame.

Yammer · 13 Jan 2013 at 11:01

Mystery solved! I split the RAID up and tested individually. All drives passed, but ...

HD1 (replacement):

HD2:

HD3: