eepro100 fifo buffer overruns

We've got 3 webservers, all set up basically identically, all serving the same content, all working the same way, all with identical hardware, or as close as humanly possible.

Two work fine.

One is getting constant FIFO buffer overruns on one of its ethernet interfaces:

[root@ ipv4]# netstat -in
Kernel Interface table
Iface MTU Met RX-OK RX-ERR RX-DRP RX-OVR TX-OK TX-ERR TX-DRP TX-OVR Flg
eth0 1500 0 85083085 0 0 0 112494992 0 0 1 BRU
eth1 1500 0 279693040 0 0 583863 51712373 0 0 1 BRU
lo 3924 0 13381 0 0 0 13381 0 0 0 LRU


Now, it's still working ok, I suppose, but that interface is experiencing packet-loss and other problems that bother me.

I compared it to one of the other boxes, and its general stats look the same, except without the damn overruns.

Anyone have any ideas what could be causing it and why? That interface is used strictly for connecting to the DB, it's a back-end interface. It has ~40 connections open to various listeners, which is the same number as one of the machines that isn't having problems.

Any thoughts, comments, ideas, or financial contributions will be gladly accepted.