Symptoms of overload on Southbridge <-> Northbridge

[watermodem]

This question is not truly a kernel question but it has kernel 
attributes so I decided to ask it here.

I have some measurement situations (using P-3 and P-4) machines where 
the Southbridge to Northbridge connection could get saturated and maybe 
even hit the bandwidth limits.  These machines are usually older 
uni-processor Dells with one flavor or another of Intel chipset.

What sort of hardware symptoms can I expect to see when the bus overloads?

(I am not that worried if the OS crashes but am worried about possible 
hardware damage - heat, out of sequence drive commands etc...)

The machines will have 1 gigabit ethernet taking a full sniffer port 
load from a loaded CISCO 6509 switch and many more 100Mbit ethernets, 
several sniffing and the rest communicating. (Adaptec 4 port units) I 
know that is an insane load but we have hundreds of labs many being near 
idle most of the time.  The mainly near idle labs could use the cheaper 
equipment. It would be nice if there was a way to sense nearing overload 
and dynamically "throttle down". Would any of the /proc provided 
statistics give me the clues I am looking for?

Thanks...