Architecting a vSphere Compute Platform : Scalability and Designing Physical Resources : 5.11 ESXi Host Server Advanced BIOS Settings
   
5.11 ESXi Host Server Advanced BIOS Settings
When configuring the compute node system BIOS, maintaining consistent settings that conform to the manufacturer’s recommendations is the starting point for any design.
The default hardware BIOS settings provided out-of-the-box on servers might not always be the optimal choice to maximize performance when the ESXi hypervisor is configured.
When designing the configuration of a new server farm, BIOS settings to consider include the following points (again, the manufacturer's recommended settings for ESXi is always be your starting point):
Verify you are running the latest version of the BIOS available from the manufacturer for that system.
Verify that the BIOS is configured to enable all populated processor sockets as active and to enable all cores in each socket.
Enable “Turbo Boost” in the BIOS, if supported by the processors.
Verify that hyperthreading is enabled in the BIOS for processors that support this technology.
Verify that that any hardware-assisted virtualization features such as VT-x, AMD-V, EPT, and RVI, are enabled.
Disable any devices you will not be using such as serial, USB, or network ports.
If the BIOS allows for the memory-scrubbing rate to be configured, VMware recommends leaving it at the manufacturer’s default setting.
If VMware vSphere Fault Tolerance is employed in the design, disable CPU power-saving modes, such as Intel SpeedStep and AMD PowerNow!, which can reduce the CPU speed.
ESXi has the ability to manage the power capabilities of the servers in the software, and as such, reduce power consumption when the host is not being fully utilized. VMware recommends configuring the BIOS settings to allow ESXi the most flexibility in using the power management features offered by the hardware, and configure your power-management settings from within ESXi. While these features can provide a reduction in power consumption with little or no impact on performance, for workloads that are sensitive to I/O latency, overall performance could be impacted.
Other "C-states" deeper than C1/C1E (that is, C3 and C6) allow for further power savings, though with an increased chance of performance impact. Again, it depends on the design requirements as to whether or not these features are employed, but an ESXi host can take advantage of advanced processor technologies such as Intel Enhanced SpeedStep or AMD PowerNow! to adapt the frequency of the physical CPUs based on the service provider’s actual needs.