I have been comparing 32 vs 64 bit flavours of Ubuntu 12.04 LTS across multiple datacenters and have found that the 32-bit version are consistently up to 40% slower in UnixBench testing. Approximate 32-bit score (mean across multiple datacenters): ~700 Approximate 64-bit score (mean across multiple datacenters): ~1100 I can't find this discussed anywhere else, but it seems 64-bit is the only way to go if you are interested in performance. (As opposed to slightly increased memory usage.)