Connection watchdog failure of 3 plc from 6

Following is the problem history, for the records...
We have a network of 6 S5 PLC's model 948, each one with is own CP1430TF
interface in a Industrial Ethernet H1 network(ISO Only, layer 4). Originally
this PLC's had COROS visualization system, that in 2000 were changed to
WinCC V5.0 SP1. This change implemented the visualization using the S5
Ethernet Layer 4 Interface included in the WinCC software. There were
basically tree OS stations each one with a CP1413 interface, with the same
visualization each one communicating to the same 6 S5 PLC's, the only
changes were needed in the communications between the PLC's, because the
FETCH/WRITE function were used in a ISO protocol environment, where almost
all is based in MAC addresses. So each application and CP1430TF was
programmed with the correct parameters (FETCH/WRITE PDU Sizes, names etc. to
match the settings in the Cp1430 at the S5 side. The system has been running
OK.
By today, the PC hardware experienced faults so It must be changed, also to
keep up with current technology it was decided to use a new PC with a CP1613
and migrate the app to WinCC V6.0 SP4. We had no trouble or error migrating
the application but instead we have a communication problem, it happens that
the application can communicate consistently only with tree of the six
plc's, the other tree keep up giving disconnection errors, also when the app
connects to these plc's it only shows ZERO values ( 0.0 in all visualization
boxes), and basically it keeps connecting and disconnecting from these tree
plcs. The other tree plcs give no errors all the time and give all the
values ok.

We have made the following test to date:

1.- We check the plc system parameters configuration of the Winccv6 app
against the same running in WinCC5, checking all is the same. In fact, the
other two OS's are running OK without faults. Also we check the
communications function type (fetch/write), pdu size and name and hex
against the CPs1430TF configurations, and also be sure the MAC address of
the new CP matches the old one because in the CP1430 this mac address is a
parameter in the fetch/write functions. All looks OK.

2.- We deleted one by one the s5 plcs that are communicating OK to see if
one of these reacts to this, without any result. We end with the tree plcs
still not communicating, the original ones.

3.- We change the plc configuration in the wincc v6 app and the mac address
of the Cp1613 to one of the OS in WinCC v5 that was operating, and
disconnect the OS with wincc v5 from the net and connected the new OS with
wincc v6 with the new network and plc configuration, obtaining exactly the
same results.

4.- We changed again and reconnect the OS with wincc v5 and build an
application in the new OS in Wincc v6 from scratch and put in only 2 s5
plcs, each with only one tag configured. From the two plcs, we observe the
same result: the one that was communicating OK keeps doing OK and the one
failing keeps failing.

5.- We apply HF1 for the SimaticNet 2005 version and repeat the tests with
same results.

We are appending the full diagnostic directory so you can check the log and
trace files.

The next step I am taking now is to replace the 1613 card in the PC with an old model ( the one I put is an CP1613 A2, the newest one), I already got it and I will make he test this week. The only difference in the manual says that the A2one does not work in WinNT4 and the other is OK.


Thank in advance for any help you can provide.

Chilamacj