What are the performance differences between the various lines of Cisco Catalysts? Google found noting useful.
Are there any benchmarks or real-life experiences how the C65xx, C49xx, C37xx and C35xx perform? Is the packet latency in switching mode for all series the same?
Edit: I am thinking mainly about 1GBit switching with <20 servers, although mixed unicast and multicast with different packet sizes.
One interesting doc is Cisco's switching performance pdf.
It's quite basic as it only compares the switches forwarding/routing capacity and their fabric speeds, and there is no informations about buffers/features/..., but it's extensive regarding the switches models (from 1900xl to 76xx sup720).
Its router equivalent is here.
Warning though, it's Cisco's official doc, so be warned that the numbers showed here may differ from actual situation.
Edit:
Take also a look at Cisco Switch Guide.
It's really hard to give a comparison of performance based on a chassis - To use a car analogy it's like saying you have a Mustang but not telling us the horsepower of what is under the hood. The modules you install in each chassis will substantially impact the performance you see.
Taking the Cisco 45xx series (just because I grabbed that page first), you can have four different supervisor engines with substantially different performance between the SupV10 and the Sup7E. The Cisco website will have full details on each chassis (and all supported modules), which you can peruse at your leisure.
I would agree with voretaq that the performance of the modular switches really depends on the modules installed. Generally they are demonstrably better than their 1U cousins though.
In terms of performance I just think of a 3750 as being a 3560 with the ability to stack. There isn't much to chose between a 4948 and a 3750 either. It boils down to which features you need to use rather than being able to say X is better than Y.
I have heard anecdotal evidence where people have said that the 4948 is superior to the 3750 & 3560 for iSCSI applications but I haven't personally seen any evidence of this. Network World have quite a few real world tests of switches if you haven't already come across those;