[rc5] Re: Supercomputers & Bovine, Q&A

Carsten Jacobi carsten at jakes.kawo1.rwth-aachen.de
Tue Aug 5 00:37:54 EDT 1997


On 04-Aug-97 z wrote:
>>Do you have any ideea why's the PPC 604e client so fast ? (fast << && >>
>>?)

>Yes, because the ppc uses a risc chip which is inherently faster that cisc

Naanaanaano, you take the whole thing to easy. You can't just claim that risc
is faster then cisc, you can only say that it is different (if you consider
a problem where the cisc'er can play their card of complex instructions they
are of course in advantage). But speaking about the performance of the 604e:
The PowerPC-architecture includes a hardware rotate-instruction ("rlwimi" and
"rlwinm"; the mnemonics sound stupid but these instructions are very powerful!)
and they don't need to "shift left","shift right" and "or" as other risc-
machines.
I think it has been mentioned very often in this mailing list, that x86, Power-
PC and 68K are good architectures for RC5-encryption because they support hard-
ware rotates ...

>chips, and now that the ppc chips are just entering their maturity, their
>speeds are now exceeding intel's by a significant margin, especially the
>one's now being introduced this and next month, the ppc 750, when combined
>with a back side level 2 cach, running on it's own board at faster than the
>mother board shows that it is ~8X faster than a ppc 601.

I would wait until the chip is on the market and until they have developed the
board for that chip. And then, what do you run on that machine? MacOS again,
that doesn't support multitasking (sorry, but please no "keys per timeslice"-
settings) and memory-protection. I don't know ...

>It is expected
>that the intel technology (risc) will reach theoretical design limits
>sooner as well, so the ppc chips may well lengthen their leads even more as
>time goes on.

Ok, it is right that the instruction-set-architecture of the x86-chips is not
the newest. Correct me if I am wrong, but isn't it so, that the Pentium II from
'97 is still correctly executing 8085-programs from '77 (I mean despite that the
surrounding I/O is different)? It is impressive how fast they have made this
architecture; it has still its solid place at the top with highest performing
micro-processors. It has been claimed so often that the x86-technology is on the
end of its development and it has never become true so far ...

Carsten
----
To unsubscribe, send email to majordomo at llamas.net with 'unsubscribe rc5' in the body.



More information about the rc5 mailing list