The controller is the AAA-UDMA, a four-channel UDMA/66 RAID controller supporting up to four hard drives in RAID 0, RAID 1, RAID 0/1, and RAID 5 configurations.
Acquiring four identical yet new ATA drives with which to test this controller was no simple task, but fortunately Computer Geeks Discount Outlet was generous enough to provide us with four 46.1GB UDMA/100 Maxtor DiamondMax 60 drives. Although those willing to pay more for higher performance should probably use IBM DeskStar 75GXPs or other new 7200RPM disks, these 5400RPM drives offer respectable speed and high capacity for not that much money. So, how did the controller fare in the benchmark tests? First, let's look at the loadout of the test system:
Intel Pentium !!! ("Katmai") 450
Abit BX6 R2.02 (BIOS revision QR) motherboard w/Intel 440BX chipset
128MB PC100 CAS2 SDRAM (1 DIMM)
Diamond Viper V770 Ultra AGP video card (TNT2 Ultra using Detonator 3 drivers)
Diamond Monster Sound MX300 PCI sound card (Aureal Vortex 2 chip)
3Com 3C905B-TX PCI Ethernet adapter
Primary drive: 22-gig IBM Deskstar 22GXP, master on primary UDMA/33 controller
CD-ROM: NEC 4x, master on secondary UDMA/33 controller
Windows 2000 Professional SP1 Test drives: 4 46.1GB UDMA/100 Maxtor DiamondMax 60s (Model #94610H6)
Note that, for some reason, Intel's IOMeter wouldn't let me select any RAID arrays created with this controller for testing, and that's why it doesn't appear in any of these results. With all striped arrays, I used the default 64KB stripe size. Although the controller supports single hard drives, you have to initialize them as RAID 0 arrays for the controller to handle them through the CI/O software. This is what I did for a baseline test. Here are the results:
As you can see, the drive performs respectably on its own. There is surely a little bit of overhead for the controller since it's pretending that this one drive is an array, so it's likely that the drive would be faster on its own on a non-RAID controller.
>
Next in the lineup is a RAID 0 array with two drives. Here are the results for that configuration:
This is not what I expected. Although this array
has higher write speeds than the single drive, its read speeds are slower. With
a two-drive RAID 0, I figured, given the single-drive performance, max read
speeds would be in the 40MB/s range. That's not the case, and it doesn't make
much sense. Of course, with computer hardware, theoretical performance rarely
translates to real performance, but I expected at least 150% that of the single
drive in all categories except CPU usage.
RSS Newsletter
