Very poor iSCSI performance with HP DL160 G6 and ESX 4.0

[xav89]

I am experiencing very poor iSCSI performance with HP DL160 G6 servers and ESX 4.0 and would welcome help on this matter.

I have setup iSCSI on 3 * ESX 4.0 hosts.

Each have a corresponding single target and single LUN on the SAN. The LUNs are formatted as fmfs3 types.

Each host is connected to the QNAP SAN in the same way via a Gigabit switch.

All NICs operate at 1Gb.

To test performance I am using vmkfstools to clone a disk from the local store to the LUN

On the first host (a DL140 G3) performance is good at 48MB/s.

On the 2 DL160 performance is very poor and barely achieve 2MB/s

When cloning, the DL160 servers also repeatedly log the following in /var/log/wmkwarning:
servername vmkernel: 0:05:23:53.214 cpu2:4308)WARNING: NMP: nmp_DeviceRetryCommand: Device "naa.xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx": awaiting fast path state update for failover with I/O blocked. No prior reservation exists on the device.
servername vmkernel: 0:05:23:54.195 cpu3:4210)WARNING: NMP: nmp_DeviceAttemptFailover: Retry world failover device "naa.xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" - issuing command 0x4100051a20c0

I have made sure that the servers are up-to-date:
DL160 G6 Configuration:
ESX 4.0.0 build-702116
2 * HP NC362i Gb nics (firmware 1.7.2, driver 1.3.19.12.1)
1 * Smart Array P212 Controller (firmware 5.14)

SAN Server Configuration:
QNAP Ts-459u-rp (firmware 3.6.1 Build 0302T)

I have searched everywhere for an answer.
I have also tried the "esxcfg-advcfg -s 1 /VMFS3/FailVolumeOpenIfAPD" as suggested on some posts.
All to no avail.

Any help / advise would be appreciated!