Hi,

> heh, come to think of it, i don't think SUN has ever put the cache on
> the same piece of ceramic (ah ha!, i can spell it now!!) let alone on
> the same die.

On die is good as you can run it a hell-of-a-lot(tm) faster than
external cache as the signal doesnt have to travel too far, hence
higher clock rates (according to an ex VLSI designer here).  This
really only makes much of a difference when you start running
CPUs really fast (SM50s dont classify as fast etc etc).

> it's hard to say since they like to hide everything under those
> enormous heatsinks.

Heatsinks are good :)  You want to see the ones on HP N-class 9000
servers (the machines that look like giant air-conditioners).  6
inch tall ribbed heatsinks (no dick jokes please).

> > seem to avoid the cache speed discussion on their web site I notice.
> > Anyone know what the clock ratio is?  1:1, 2:1 etc?
>
> that's a good question.  if anyone knows, i'd be interested as well.

Yep - that really would affect the performance if it was 2:1.

> and even then, they didn't make a CPU that was used in what could
> be classifed as a computer until 1974 with the 8080,

...which a load of pissed of Intel engineers left and started
Zilog who produced the Z80 and Sun Serial ports :)

> and didn't make a CPU that was truely mass-produced until 1978
> with the 8086/8088, AND didn't figure out 32-bit until 1985.
> so they got an early start, but they were sufficiently slow
> with their design process that it took them 20 years to catch
> up.

Dare I say AMD have advanced in large bounds comparatively.
Problem is that they're just selling 1GHz 386's with some
extra instructions strapped on the side...

> i've got to start mucking with the Netra's at work, i didn't
> even know they could do this (i don't really use them very
> often).  this is far too cool.  i though the ROM monitor was
> neat enough, but boy was i wrong.

Can you put values in memory and run them???  It'd be a very
`hackish' style of coding :)  Back to the old days at school
of trying to poke values into memory on Commodore Pets to
overfrequency the monitor and consequntially blow the flyback
transformer.  Serves the teacher right for saying `you cant
damage the hardware with software'.

- Chris.