I think L3 is much faster; perhaps 1/3rd of main memory, assuming the line is av...

I think L3 is much faster; perhaps 1/3rd of main memory, assuming the line is available and not in another core. Here are some numbers for L3 cache, from Intel (probably specific to the 5500 series)[1]:

  L3 CACHE hit, line unshared ~40 cycles 
  L3 CACHE hit, shared line in another core ~65 cycles 
  L3 CACHE hit, modified in another core ~75 cycles 
  remote L3 CACHE ~100-300 cycles 
  Local Dram ~60 ns 
  Remote Dram ~100 ns

60ns at 2.4GHz is ~144 cycles, right?

1: http://software.intel.com/sites/products/collateral/hpc/vtun...