Thanks for answering,
yes, I've inspected the socket (not very scrupulously), but systen-board stopped working while nobody touched it, except switching AC plug - I just move server from table to the closet, reconnect Eth/PS2/VGA, install AC plug, and server did not boot to BIOS (hangs on x98 post-code).
As for the POST sequence, as you can see in my previous post, code are going not in sequence, e.g. 0xB0h to 0x13h then again 0x50h (I posted video: http://www.cs.vsu.ru/tmp/123.avi ).
What is important for me to understand: if I see POST code 0x98h (Resetting the mouse). does it mean the mouse reset has been passed, and the problem is with next test? If so, than what is the next test? Numerically, the next test, according the table, is 0xB0 "Resetting fixed media device" , but looking on "jumping" codes I've decoded (see prev. message) I'm in doubt.
I already removed board, found no screws or other metal items underneath. BMC is running under Linux, it's a pity, that one can not just connect terminal emulator to serial and dump detailed test results instead of peering on LEDs.
My plans are to clean the system-board and to give it a next try. Then, if time permits, to check various supply voltages, in particular will try to check voltage regulator if it is easy to access. Because all started when I change CPU to Xeon L5335 (Low-Voltage and this changed the way VR works for years before...