ASM Performance CALIBRATE_IO varies according to disk count on DG

Hi, i have a question regarding CALIBRATE_IO package.

    I'm running some tests on a new rac installation, 11.2.0.4, 2 nodes, a diskgroup with 6 200GB disks.

The thing is, i run calibrate_io on those disks and i get 1800 IOPS, which is bad, because for the application i have to deploy it should be somewhere around 4000 IOPS.

We get from the storage admin that those are the fastest disks on the storage device, so we try creating a DG with more, smaller disks, thinking maybe if we have more striping, performance will be better. So, I create a DG with 20 95GB disks and IOPS goes up to 3300. But then, I drop all disks but 2, and IOPS goes up to 5700 ( 2 95GB disks). And finally after seeing this i go back to the original DG with 6 disks and leave it with 2, and IOPS spike from the original 1800 to 3200.