I don't know the source of your problem, but while looking it up on the web, I found out about the interesting command addr2line, which might be able to locate the routine that causes the panic. http://fixunix.com/kernel/396075-re-...eption-0e.html
I've never used addr2line and I don't understand why it worked in that example. Does the syntax that was used ("vmlinuz") depend on being booted successfully with the particular kernel being analyzed? Could you use a longer name (like "vmlinuz-3.3.5-2.fc16.x86_64") to analyze an error in a kernel that won't boot? Must the kernel be compiled with debugging options for this command to work? Perhaps another forum member can explain this command.
The gist of the example is that the early panic with error:
PANIC: early exception 0e rip 10:ffffffff80224b4e error 0 cr2 ffffffffff5fb030
was diagnosed (apparently by someone who could boot with the kernel in question) by the command
addr2line -Cfe vmlinux ffffffff80224b4e
which produced output