Raptors and RAID: Game Load-Up Results

How much does RAID help game loading? – Conrad aka Vio 1

I took it upon myself to once and for all figure out if changing the stripe size in raid-0 would change the real-world load-up of games. So for all the people wondering what stripe size to set: LOOK NO FURTHER.

I also decided to see how one drive stacks up against raid-0.

Here is the background info:

Games Used:

  1. UT2004
  2. Ground Control 2
  3. Lock On
  4. Far Cry
  5. Unreal 2

Benchmark Utilities Used:

  1. Sisoftware Sandra
  2. PCMark04
  3. PCMark 2002
  4. HD Tach 2.70

System Used:

  • INTEL P4 2.8C @ 3.22¹ GHZ
  • ASUS P4S800DE DLX
  • 2 x 256 MB CORSAIR XMS DDR PC3500
  • 2 x 36.7 GB 8MB 10000 rpm WD RAPTORS (RAID-0) & 120 GB 8 MB SE WD HDD
  • ATI 9700 PRO (351/702)
  • 19″ VIEWSONIC G90FB monitor
  • SOUND BLASTER AUDIGY 2 ZS
  • THERMALRIGHT SLK900U w/ 92MM PANAFLOW

¹Normally I run at 3.6 GHz – I have it OC’d to 3.22 GHz because it’s summer here and too hot outside to overclock to 3.6 GHz.

Procedure Used:

I installed a fresh copy of windows XP Pro onto my raid-0 raptors and installed all the critical patches for Windows. I installed the newest drivers for all my system components. I then installed all the games and benchmark utilities mentioned above. I then defragged my HDD and used Norton Ghost 2003 to make an image of the drive and copied that to a separate HDD.

I then proceeded to time the load up for the games and programs listed above (restarting the computer between each game/program).

When that was done, I restarted the computer, deleted the raid setup and created a new one with the next stripe size up. I then copied the image created in Norton Ghost over to the clean HDD and restarted the whole process of timing.

I did this over and over until I tested all the stripe sizes available to me, and then I tried a single HDD.

What did I time exactly?

  • UT2004 – I timed the load up of the largest map size (DM-1on1-Desolation)
  • Ground Control 2 – I timed the load up of the tutorial 1 after the video
  • Lock On – I timed the load up of the F-15C Instant flight
  • Far Cry – I timed the load up of the first level after the video
  • Unreal 2 – I timed the load up of the first level after the video
  • Sisoftware Sandra – I timed how long it took the HDD to complete the File System Benchmark
  • PCMark04 – I timed the HDD test suite
  • PCMark 2002- I timed the HDD test

    Note: I used a hand-held stop-watch to do these timings; therefore there is a slight margin of error.

    AND NOW — THE RESULTS!


    {mospagebreak}

    Drive Load Time in Seconds

    Benchmark

    16k Stripe

    32k Stripe

    64k Stripe

    128k Stripe

    256k Stripe

    Single 36Gb Raptor

    UT2004

    12.07

    13.83

    12.12

    12.64

    13.65

    14.32

    GC 2

    11.31

    11.78

    11.81

    12.22

    12.08

    12.37

    Lock On

    30.26

    26.82

    25.96

    25.82

    26.57

    28.16

    Far Cry

    1.34.96

    1.29.36

    1.23.70

    1.27.03

    1.25.98

    1.30.42

    Unreal 2

    16.92

    16.46

    15.90

    16.29

    16.19

    16.44

    Sisoftware Sandra

    49.00

    45.68

    47.17

    46.72

    54.88

    1.21.39

    PCMark04

    3.30.74

    3.30.85

    3.42.25

    3.42.91

    3.33.15

    3.55.34

    PCMark 2002

    23.00

    21.27

    20.78

    24.92

    25.03

    35.00

    HD Tach 2.70 seek time – ms

    8.7

    8.8

    9.1

    8.9

    8.8

    8.5

    max/min/avg mb/s

    119.5/63.9/96.7

    119.4/28.5/59.3

    99.9/32.0/61/6

    91.6/41.4/59.3

    87.0/34.9/53.4

    60.0/36.4/50.1

    Conclusion

    I noticed no change in load up times across the board in the games (except for Far Cry). Stipe size made no real difference in how long it took to load the levels.

    As to how Raid-0 stacks up against a single Raptor, the only difference is in the benchmark scores – other then that, in real world use there is NO REAL IMPROVEMENT in load up times.

    Conrad aka Vio1 in OC Forum

Be the first to comment

Leave a Reply