I'll miss the introduction and proceed to explain my problem.
I have two WS-C3750G-24TS catalyst switches in a stack. Couple months ago I found out an issue with them and two days ago understood with my friends at work what exactly the problem is. Of course we use them as a layer 3 devices and they are responsible for many bgp sessions. When I try to simply ping a host from one vlan to another, where should come the routing process in the stack, there is no problem if the vlan member (host) is on the same switch. If the traffic have to move through the stack cables to the other stack member appears high latency, sometimes big losses of packets. If I try to ping a host on the same vlan but in another stack member, there's no problem and the traffic is great - no issue. One of my friends said it should be a hardware issue of the platform. The stack is build with the desktop routing template. The software is C3750-ADVIPSERVICESK9-M, Version 12.2(25)SEE, RELEASE SOFTWARE (fc2). A single server connected to the second switch (the emptier, low used) did run up the stack cpu usage by more than 20%. Please, provide some light in the tunnel. Is there any possible way to fix it? Thanks in advance!