The problem with server benchmarks is that they're very specific. They're either about well-established and typical applications (MySQL and so on) or on completely unknown, specific code. That's why many higher-end server vendors let the customer try the server with their own code before any purchase.
For typical server benchmarks, those links that were mentioned here are probably enough. Spec, TPC, and so on are pretty standard server fare.
In any case, you could try anandtech's IT Computing
articles, which cover server CPUs ranging from Intel Woodcrests to Opterons to SUN T1 servers. It's an interesting read, and it would probably be a good start. Don't take those articles as the whole truth though, because there has been some discussion as to anandtech's methods before...