Re(2): [SunRescue] Ultrasparc CPU dismantling

[Tim Hauber]

rescue at sunhelp.org,Internet writes:
>Okay, here's the story on the UltraSPARC IIi processors.
>Apparently, the speed of the processor is controlled by
>the little module board, and the rated speed is ignored
>by an Ultra 5.  It is easy to swap CPUs between boards.
>The speed of the CPU moves with the board.  Which leads
>me to an interesting question -- the boards are relatively
>simple, and it is possible that mere resistors are indicating
>to the Darwin at what speed it is to run -- how would we
>modify them to overclock our CPUs?  The 333 mhz CPU that
>I took apart actually held a 360 mhz rated UltraSPARC...
>
>Just food for thought and you electrical engineers out
>there ;-)
>
>Chris

>From an electronic standpoint, CPU timing is a piece of cake, just a
simple little oscillator circuit.  unfortunately, that isn't all you need
to worry about.  In the PC world, overclocking is fairly easy if you buy
good parts, because there is so much junk out there the standards are well
below the timing capabilities of good pieces. On a real computer :-) the
timing tolerances of the memory bus, the various other interfaces to the
CPU  are not as sloppy.  If you speed up the processor you also have to
take into consideration all the other pieces on the processor card, which
probably includes cache, and maybe some kind of interrupt buffering, and
who knows what else.  The other thing which people are always talking
about on this list relating to sparcs is heat dissipation, the faster it
goes, the hotter it runs.  Overclocking won't kill a chip, it will just
get to a speed and quit working, but heat will kill it dead.