UT2003 Folder Copying:

We start of with recording the time it took the copy a folder TO the Maxtor drive(s) from the system drive, and afterwards copying that folder back FROM the Maxtor drive(s) to the system drive!
Copying TO : write speed test
Copying FROM: read speed test


Detailed view of the folder that was copied




First thing you notice is the almost zero difference between the 2mb & 8mb versions when they are used raid-less. RAID 1 demands more from the drives when copying to them, hence the increase in time while copying to the RAID1 array, but it does show a small improvement when reading (copying from). The 8Mb cache makes the drive less erratic between the read/write tests, while the 2Mb version takes the lead for the read tests, but comes up with the slowest time when "copying to".
The 1k stripe size shows no increase or decrease, for neither disks, but the 4k SS seems to be the sweet spot for this kind of disk activity, both 8mb and 2mb HD's decrease their times for copying files with the 8mb having a comfortable lead over the 2mb version.
Increasing the SS makes the difference between READ/WRITE smaller, the largest one(512K SS) even has the WRITE test outperform the read test on the 2mb version, while the 8mb levels out for the 2 tests and performs equally good when coping TO or FROM.
Software RAID0 works very good for the Maxtor 8Mb setup, while it causes a serious performance impact on the 2mb one.

In short, conclusion for this test:

RAID 1 slows downs writing speeds , and shows a small increase in reading tests

RAID 0 - 1k shows almost no difference between the none-raid setup

RAID 0 - 4k is the sweet spot for the 8Mb version

RAID 0 - 16&64k has +/- same read speeds as the none-raid, but it gives a considerable speed boost for the write speed

RAID 0 - 512k makes the 8mb setup perform equally fast for read/write, while it speeds up the writing of the 2mb version more then the read speed.

Software RAID 0 gives very good results for the 8mb version, 2mb sees a small increase in write speed but a larger setback for the read test.

UT2003 Map Loading times:

We started an UT2003 map & timed how long it took before it loaded, then we exit the map & relaunch it from the menu, this gives us the 2Nd result: Buffered Load



The 8Mb version doesn't show any clear sign of improvement when using RAID this makes me assume that in this case most data required for UT2003 maps can fit in those 8Mb.
Using RAID 1 gives an increase in read speed, and this is shown in the graph. A 1k stripe size increases the CPU load, therefor it takes longer for the game to load. The software RAID 0 is also pretty CPU intensive and it shows the same symptoms as with the 1k SS. Overall it has only a very small advantage to use RAID 0 in this game with the 2mb Maxtor.