The thing I am looking for most right now maximum throughput in one direction only, not bidirectional.  The thing i am trying to measue is the cpu load that the networking part of my setup is taking up. Right now, with a 100megabit connextion, the cpu load time 14%. I want this to be as close as possible to 0%, so the cpu can spend its time doing my algorithm on the data in memory.  In reality, i just want around 400mega bits, on one connection (not 4), that simulates a mock internet in the future, where gigabit lans will be a norm.  
So basically, I need to aquire the following things:
2 NIC's with its own processor, maximum cpu offload capabilities
CAT 5e cabling
Switch
Software on the servers. Im guessing WIN 2K is no good? I would like to stick to it as much as possible, but a viable option to change my software for a new OS.
So please point me in the right direction for more info, and exact componnents for the list above.  
thanks again. i appreciate all your help.