This really depends on what you mean by 'speed'.
If you care about latency and nothing else (for example, you have mostly CPU-bound workloads), take the reciprocal of the rated operating frequency for the module and multiply it by the CAS latency to get the actual time it takes for a single controller access sequence. The lower that number, the lower your overall latency, and the 'faster' your RAM is. This is invariant of the rest of the system, barring cases of the firmware configuring non-standard timings for the RAM or the system needing to underclock the RAM for some reason.
Beyond that, you really need to just benchmark things and not try to consider the RAM in isolation. Actual memory bandwidth for example (which is what I suspect you actually care about when you're asking about 'speed') is too dependent on the specifics of the rest of the system to be measured in isolation like you seem to want to do.