I know this thread is old, but I just thought I'd post the solution in case anyone else runs into this issue.
This issue occurs on nforce boards utilizing the HPET chip, running kernels built with AMD_X64 prior to 2.6.23 and running Asterisk module zt_dummy. I'm not sure what broken code exists in the zt_dummy module, but it causes a lockup of seemingly random devices (read NIC, SATA, IDE, etc.) after an arbitrary number of RTC errors.
Using a kernel that is compiled with HPET_EMULATE_RTC=y option will solve your issue (this config option is enabled by default in all build architectures except AMD_X64, in which it was disabled by accident). Additionally, removing the zt_dummy module will work (but also break Asterisk). On some boards that have both RTC and HPET chips you can disable the HPET chip in the bios, although I don't recommend this on anything that will be playing media. The HPET is useful for smooth playback of processor intensive media, such as h264 content.
I'm not entirely sure if these boards emulate RTC for older operating systems, like Windows XP or if they have both an RTC and HPET. If you disable ACPI that should also force the system to use the RTC or emulate it at least.