Same here... When I had this problem for the first time, I
replaced the eeprom/clock chip and the system behaved well.
Then, after a few months, it died and never came back. The
initial hiccup may or may not be related to the final demise.

The clock chips is cheapest to get from an old PC mobo, these
sell for next to nothing at the local 2nd hand weekend market
(in Australia and surely elsewhere too).

My led doco, collected from var emails, says
 ======================
   LED          Significance
   3210 7654

   1xxx xxxx    System can't write to EEPROM on system module
   0xxx xxxx    System can write to EEPROM on system module

   x1xx xxxx *  maximum baud rate is 9600
   x0xx xxxx *  maximum baud rate is 19200

   xx1x xxxx    32MB module in bank 0
   xx0x xxxx    8MB module in bank 0

   xxx1 xxxx    reset is enabled and halt disabled
   xxx0 xxxx    halt is enabled and reset disabled

   x1xx 1111    initial state at power-up, no code has executed
   x1xx 1110    entered ROM, some instructions have executed
   x1xx 1101    stack tested and set
   x1xx 1100    NXM taken
   x1xx 1011    security initialisation
   x1xx 1010    DZ initialisation
   x1xx 1001    exception tests
   x1xx 1000    FPU test
   x1xx 0111    write buffer test
   x1xx 0101    data cache tag test
   x1xx 0110    instruction cache tag test
   x1xx 0100    TLB test
   x1xx 0011    data cache test
   x1xx 0010    instruction cache test
   x1xx 0001    data cache i-stream test
   x1xx 0000    instruction cache i-stream test

   x0xx 1111    memory bitmap test
   x0xx 1110    memory address test
   x0xx 1101    memory data test
   x0xx 1100    memory moving inversion test
   x0xx 1011    memory data shorts test
   x0xx 1010    RTC tests
   x0xx 1001    DZ tests
   x0xx 1000    SII tests
   x0xx 0111    LANCE tests
   x0xx 0101    EEPROM test
   x0xx 0110    option card test
   x0xx 0100    reserved
   x0xx 0011    console I/O mode
   x0xx 0010    primary boot operation
   x0xx 0001    secondary boot operation
   x0xx 0000    operating system running

* The function of LED 2 during normal operation.  LED 2 has a different
  significance during the diagnostic tests.
Source: Simon Burge
 ======================
from the DECstation 5000 Hardware Operator's Guide:(think this is 
true for 5100 also)

o=light is off 
x=light is on

oooo oooo
=
If the workstation does not turn on,  device or module is not 
connected properly or is broken

If the workstation turns on, all components of the workstation have 
been initialized, and the workstation is ready for use

xxxx xxxx
=
Workstation hardware initialization could not begin

ooxx oxxx
=
The firmware could not calculate the cache size.

ooxx xxxo
=
The firmware could not calibrate the millisecond delay loop.

ooxx xxox
=
The firmware could not start the power-up versus reset code.

ooxx oxxo
=
Memory bank 0 could not be initialized

ooxx oxox
=
The firmware could not initialize the IO ASIC to a known state

ooxx xxxx
=
The ROM in the system module or in one of the option modules could 
not be initialized

ooxx xoxx
=
The system console programs could not be initialized

oooo ooxo
ooox ooxo
ooxo ooxo
ooxx ooxo
=
The system module or one of the option modules could not be 
initialized. The lights on the left show the slot number of the 
module. This table shows slot numbers 0 (off off off off) to 
3 (off off on on)

oooo ooxx
ooox ooxx
ooxo ooxx
ooxx ooxx
=
The workstation was initialized, but there is a problem in the system 
module or one of the option modules. the lights on the left show the 
slot number of the module with the problem. this table shows slot 
numbers 0 (off off off off) to 3 (off off on on).

oooo ooox
ooox ooox
ooxo ooox
ooxx ooox
=
The workstation software could not be loaded. the lights on the left 
show the slot number of the module where the software is stored. this 
table shows slot numbers 0 (off off off off) to 3 (off off on on).

Source: Daniel Meinhold
 ======================

Lord Isildur wrote:
> 
> i have a 5k200 board that does the same thing. after swapping a lot of
> memory in and out, i decided that the board finally died. :(
> 
> On Sat, 18 Oct 2003, Urban Boquist wrote:
> 
> > Hi all,
[description of dead mobo trimmed]

--
Eyal Lebedinsky (eyal@eyal.emu.id.au) <http://samba.org/eyal/>