An ESXi 5.5 Update 2 host might fail when using the HyperTransport feature on an AMD server with 1TB or more RAM. On such servers the platform firmware marks the HyperTransport address range as being reserved in the memory map.
This issue happens when the operating system (ESXi) maps this address range as being present in the page tables.
The symptom of this issue is that the server might unexpectedly stop responding or reboot. Furthermore, the platform might report an Uncorrectable Machine Check Exceptions (UMCEs) in the Integrated Management log file similar to the following log:
Critical","CPU","##/##/2014 15:41","##/##/2014 15:41","1","Uncorrectable Machine Check Exception (Board 0, Processor 3, APIC ID 0x00000060, Bank 0x00000004, Status 0xF6000000'00070F0F, Address 0x00000050'61EA3B28, Misc 0x00000000'00000000)",
Mode of failure: Unexpectedly reboot. IML displays UMCE occurred.