Sata raid not quite performing up to snuff.

[theTomunist]

So, i'm running a RAID 0 setup, 2 of them actually. One is with 2 40 gig drives stripped on a hacked Promise Ultra 100 card, modified to be a Promise Fasttrack 100. They work beautiful. The problem is with my other raid. I have 2 80 gig SATA drives stripped on the Silicon Image Raid headers on my motherboard, an Asus A7N8X Deluxe. This would be a pretty sweet setup, except that the SATA raid is slow as, well, that blue haired lady driving in front of you. Here is a screenshot taken from HD Tach to show you exactly how crappy its running.http://www.tech.mtu.edu/~tlnichol/ScreenShot004.jpg

So, i'm looking for suggestions, other than cover it in napalm and watch it blaze. Though that would be fun. Hmm...

[E4g1e]

Those B@stiges at $i1ic0n 1m4g3 really screwed you over, eh? It sux0rz to be like dat. I feel ya, dawg. :(

[defakto]

You do realize that the raw performance of that is up to par, fi you don't believe me run HD tach on a single drive that you have in teh array and see what 1 drive benches at.

[E4g1e]

Oh, I had the exact same controller on my NF7-S. Single-drive performance gave me no problems other than the slightly high CPU utilization result of nearly 20%. But when it comes to RAID 0 performance on any of the current Silicon Image SATA controllers, fuhgeddaboudit!

Overall, a terrible software-based RAID implementation from Silicon Image. Other software-based RAID 0 implementations also show up-and-down cycles on the HD Tach graph - but not this extreme.

[theTomunist]

Firstly: raw performance is not up to par, it was never that jumpy when i had a single disk on there and the average speed was actually decent, not as fast as i would have hoped but quite a bit faster than it is now
Secondly: how do you figure its a software raid? Unless i'm a complete ass i think its a hardware raid.

[E4g1e]

Secondly: how do you figure its a software raid? Unless i'm a complete ass i think its a hardware raid.
Actually, the Si3112r's RAID is not fully hardware-based. All the hardware (or the BIOS) RAID settings do is to tell the software drivers in Windows to use that alogarithm. And the software does most of that work.

[theTomunist]

I'd like to know where you happened upon this software raid information. It doesn't seem to click with me. If it does rely on a system driver, than why can i use the raid without windows loaded? I can't find any data saying that it is a software raid, and even if it is, looking at other peoples benchmarks of the same chipset, my setup gets spanked.

[Nitrogen]

I've seen a similar problem with my RAID. I noticed it seemed sluggish at one point and upon HDTaching saw that jumpiness. That's what originally brought this all to light (I'm friends with theTom). I knew when I set up my array (dual WD800JB's on a VIA controller onboard my Asus P4P800) that it benched a lot smoother than that.

So this crap started occuring right after I swapped my components to a new case. I went through hell diagnosing why it was happening. It finally came down to two screws in one of my drives. If it was screwed in completely, for some reason my performance went to shit.

I e-mailed Western Digital because I was thinking maybe it was a manufacturing defect. Here are the comparison images I sent them, since they're useful here as well.

Screw Tightened (http://personal.cmich.edu/magri1aj/dump/ScrewTight.png)
Screw Loosened (http://personal.cmich.edu/magri1aj/dump/ScrewLoose.png)

Now, we've been trying to diagnose this, but to no avail. Doesn't seem to be the same problem, but the general effect is the same. We don't know why. Guess I'm offering this as evidence to what we think the RAID should be capable of--we don't think his is hitting its full potential.

Having all that out of the way, good suggestions so far. Keep 'em coming. It may be a case of 'the more obscure it seems, the better'. This seems to be one of those really wacky problems.

[theTomunist]

I should probably mention that the raid has 2 different drives on it. Why? because i'm a dumbass and ordered the wrong one when i got the second drive. So the first drive is a Seagate ST380013AS 80GB, 7200 RPM, 8MB buffer, the other is a Samsung sp0812c, 7200 RPM, 8MB buffer. I didn't send the second one back because at first i didn't have them in a raid anyway, and when i did decide to do that i figured they'd work ok with each other, having very similar specs. maybe i was wrong. dammit.

[theTomunist]

Problem resolved. It turned out that one of the drives was slowly dying, which really messed with the raid performance. Since it was replaced its been working like a dream. And as a followup it works just as well and with as little cpu usage as my PCI raid controller, so i think rumors of it being a software raid are unfounded. Especially since i can use the raid out of windows.