Re: more trouble with Sun Blade 1000

On 2009-07-05, glennklockwood <glennklockwood@xxxxxxxxx> wrote:
On Jul 5, 11:05 am, glennklockwood <glennklockw...@xxxxxxxxx> wrote:
On Jul 5, 2:36 am, Andreas Wacknitz <a.wackn...@xxxxxx> wrote:

[ ... ]

Blade 1000/2000 CPU modules are very sensitive. Have you used the
provided tool to attach them? In my experience your kind of problems is
caused by badly fixed CPUs. You should try to remove them and insert
them again (one by one).


As an update, it in fact was an issue with a CPU module.  In this
case, apparently one of the 900MHz processors I bought was bad.  I did
try re-seating and re-torquing each module using the cylindrical
torque wrench provided in the SB2k's, and I found that the system
would not boot with one specific module in either CPU0 or CPU1.  It's
strange that the module did work the first few times after it arrived
but broke very soon after.  The guy who sold it to me must've cracked
an egg into its radiator.

I'm willing to bet that it is from the eBay vendor who takes the
heat sinks off to photograph the CPU chip's identifier -- and in the
process has reduced the efficiency of the heat conductivity from the
chip to the heat sink.

Now I am left with the choice of using one working 900MHz module or a
pair of 750MHz modules.  Given the electricity that these things pull
and the trouble I've been having with hardware, though, maybe it's
time I just sold these two Blade 1000's off and save the money I've
been spending on their maintenance on a new x86.

[ ... ]

As an afterthought, has anyone tried swapping the physical chip from a
US III Cu module onto a non-Cu module? I have quite a few extra
(working) 750MHz modules, and if my 900MHz chip itself is still
working, it would be worth my while to transplant it onto one of these
other working modules and just toss the non-Cu chip. There are no
apparent physical differences between the Cu and non-Cu modules, but
I'd hate to damage something further by trying something which should
not be done.

Hmm ... the 900 MHz Cu and the 750 MHz non-Cu modules have the
same cache size. But I'm not sure whether the Cu chips might work at a
lower voltage, and the module might include voltage regulators to feed
the chips.

I've never had the heat sink off, so I don't know whether the
CPU is in a socket or is soldered to the board. And I'm not sure what
torque limits to set the heat sink screws to. It looks as though there
are springs under the heads of the screws and if the screws have limited
travel the springs may simply provide the proper compression -- and I
suspect the tightening sequence for these is also critical as is the
sequence for the jackscrews installing the CPU modules. Anyway --
assuming that the CPU chip is socketed, and that there are no problems
with the mounting screws and springs, make sure that whatever is between
the CPU and the heatsink is properly replaced. If it is one of those
silicone rubber heat conductor pads, you should be able to reuse it. If
it uses a heat sink compound, you should clean the surface and renew the
compound when you swap CPU chips.

Let me know how it goes. I've got a damaged module with a good
900 MHz Cu CPU and several 750 MHz CPUs.

Good Luck,

Email: <dnichols@xxxxxxxxxxx> | Voice (all times): (703) 938-4564
(too) near Washington D.C. |
--- Black Holes are where God is dividing by zero ---