• First the most obvious, make sure your BIOS version is current and as close as possible on revision number on every node of a cluster.  The majority of issues can be avoided by getting the bios revision as close as possible (if not identical) on all physical hosts.
• Make sure all BIOS settings are consistent across all nodes of a cluster.  Human error can get you here, even if you think all of the settings are the same, check again.
• If you are using HP Blades with on-board FC HBA’s make sure to leave them enabled or else you may not be able to install ESX as described in this article .
• Enable the "No Execute Memory Protection" in the BIOS of all nodes or vMotion and EVC may not work .
• Select "Windows 2000" as the operating system in the BIOS.  Except for the DL 740 and DL760 which should select "Linux" as recommended in this article .
• Set the "MPS Table Mode" to "Full Table APIC" as described in the previous article .
• Be cautious of interrupt sharing and balance devices across interrupts so that you do not run into performance issues like those described here .  There’s also an updated article for ESX 3.5 which describes it here .
• On AMD Opteron systems the BIOS gives an option in the memory settings for "Node Interleaving".  The workloads you will be running will dictate how this is set.  This article can help you determine the proper setting.
• Enable "Intel Virtualization Technology" if available under processor options.
• If you do not have some of the settings above, they may be for a different version of your BIOS or for different hardware.
• Lastly, if you update your BIOS for any reason, go back and check ALL settings.  Occasionally manufacturers will reset the defaults for settings and an upgreat of the BIOS will change yours.

As always, if there are updates to the ones I’ve named above or if you know of additional ones, please post in the comments section.