I already tried 5000ms, no difference.
The hardware of our old cluster wasn't that much slower. 4 socket HP DL580G7, HP Lefthand iSCSI storage, connected to Cisco Nexus 10ge switches. The new cluster consists of IBM Flex System (Blade) with x240 nodes, EN4093 network switches in passthru mode connected to the uplink Cisco Nexus switches, FC3171 san switches in passthru mode connected to the datacenter san switches. And EMC VPLEX with HP EVAs as storage. The fast storage seems not to be the only reason, if I simply svMotion a VM to an old storage system, the problem is still there. Either I move the whole VM to the old cluster or I've to limit cpu to eg. 300 Mhz too.