[rescue] RE: divide and conquer? (CPU design)

[Joshua D. Boyd]

On Thu, 21 Jun 2001, Ken Hansen wrote:

> I am *far* from being an EE, but I do know that once you move componentls 
> off the same die, peformance takes a *huge* hit - witness the PPro with
> it's on-die cache compared with similar chips with cache on a separate
> die (PII?).
> 
> There are somethings in life where I defer to the engineering department, 
> like high-speed computing and automatic transmissions*. I look to Cray,
> and see that their latest designs started to use (IIRC, Dave McGuire
> feel free to correct me) collections of "off-the-shelf" CPUs as opposed
> to dividing the CPU into numerous components, as you describe.

It depends on the Cray.  For the T3E line, you are certainly correct.  If 
you are talking about the SuperCluster line, you are even more correct.  I
think everything else currently sold is completely custom.
 
> Real gains in speed are achieved by proximity, and dividing a CPU into 
> multiple pieces of silicon (or whatever) will have a negative effect.

I beleive that he was trying to say that it should be 8 CPUs on the same
die, as opposed to one CPU with 8 execution units.

I think that is a nifty idea, but I wonder how that performance would
compare to a VLIW design.  The IA-64 sorta does this, but only for a
maximum of 4 instructions per bundle, and 4 is what you get when you are
lucky.  I don't trust anything Intel does as being a good example of a
design strategy generally.  

--
Joshua Boyd