I have the same board a user in a blow thread has. The Gigabyte 8KNXP-Ultra Rev 2. It has an onboard u320 controller. This controller operates on a seperate PCI bus which is rated over 500 megabytes/sec (so, I am not limited by the standard PCI 33 Mhz bus).
I have 3 Maxtor Atlas IV 36gig drives. They are 80 pin hard drives, and are being converted to 68 pin by EXPENSIVE adapters rated for u320 (like $35 a piece). So, I am not using cheap ebay junk. I am using an LVD/SE cable, and the cable is properly terminated.
I Think I am affected by the XP bug, so tell me what you think
1) When I had a 2 drive RAID 0 stripe with the 3 drive set as a spare, I was getting the following numbers:
on the RAID 0 array I was averaging about 90 megabytes/sec read, bursting to 110.
on the SINGLE SPARE, I was averaging about 60 megabytes/sec read.
2) When I created a 3 drive stripe set, my numbers did not change for the stripe. Still averaging about 90 megabytes/sec.
3) I have used a lot of different tools. One of the tools I saw very weird results in was DiskSpeed32. On some boots (when I had a 2 drive stripe), the thing was showing 3.5 ms access time, and average transfers of 180 megabytes/sec. That just seemed bogus to me. Yet on other boots, I was seeing the 90 megabytes/sec (which seemed to be confirmed from other programs. I thought this was crap, until I had PCMark HD benchmarks vary from 2000 to 3500. Maybe this is nothing, maybe it is something.
So, does it look like the XP problem, and that I need to use 2000, or is something else going on?