8 x 512MB FBD-533 on 4 channels:
PHORONIX copy: 2014.77
PHORONIX Scale: 2016.36
PHORONIX Add: 1985.11
For comparison, here is results from my dual 5160 (3GHz)
4x2GB FBD-667 using 4 channels on 5000X:
ramspeed 2.4.1 UP compiled for i686:
INTEGER BatchRun Copy: 2414.22 Mb/s
INTEGER BatchRun Scale: 2334.57 Mb/s
INTEGER BatchRun Add: 2122.86 Mb/s
Here are results using 2 and 4 cores, respectively:
ramsmp 3.3.1 SMP compiled for i686 -p2:
INTEGER BatchRun Copy: 3990.67 Mb/s
INTEGER BatchRun Scale: 3884.67 Mb/s
INTEGER BatchRun Add: 3661.25 Mb/s
ramsmp 3.3.1 SMP compiled for i686 -p4:
INTEGER BatchRun Copy: 3985.57 Mb/s
INTEGER BatchRun Scale: 3942.61 Mb/s
INTEGER BatchRun Add: 3503.89 Mb/s
The tech docs for my S5000XVN say "Peak theoretical memory data bandwidth is 6.4GB/s with DDR2-533 and 8.0GB/s with DDR2-667."
It would be interesting to see if the quad cores can get any closer to that.
Also, notice how 2 cores nearly doubled the throughput, but 4 cores didn't do anything. So, I would say your results are cpu-limited and you aren't giving an accurate picture of maximum performance.