[rescue] SMP on intel wasteful?

[Kurt Huhn]

> A took a <$2,000 PeeCee running Linux (a machine that would now be
> chipped according to the standards of Mr. McGuire's recycling friend)
> and a fully populated Sun E450.  I ran a number of EDA tools from
> Synopsys, most notably VCS and Design Compiler.  The PeeCee averaged
> 2.5x the overall performance of the Sun running the same jobs.  Sun was
> able to wring minor performance improvements out of the E450 but was
> unable to offer any greater explanation of why we should keep buying
> their hardware in the face of such a defeat.

I have no experience with either of the above, but it sounds a lot like
a specific application (or group of apps) designed for a specific
purpose, gave you better performance on an x86 box.  However, that does
not mean that x86 is better for all stuff, or that x86 is faster for all
things.  The world is not so cut-n-dry...

> 
> Since engineers were getting their results back much more quickly, they
> were able to increase productivity and reduce the time of the design
> cycle for their next DSP core.  EE's aren't cheap, so you can multiple
> the number of engineers times the number of days saved by switching to a
> PeeCee EDA farm and you can quickly come up with some big numbers beyond
> the obvious ones to show why the switch to PeeCee's and away from Suns
> for EDA was a wise move.
> 

Sounds as if, in this case, you made a good move.  However, keep in mind
that your circumstances are not transferable to all situations.  Plenty
of folks are making the move *away* from x86 for performance reasons.

Without understanding what the metrics for your testing were, how they
were performed, and what the desired end result was - it's difficult to
see the light you've lit...

> 
> Ah, I see.  I am the guy showing real world examples of why x86 is a
> viable business platform for UNIX workstations.  The other guy says
> "bite my arse hairs" and I am the one making an ass out of myself.
> 
> That's about as clear as mud.

Actually, that was "choke on my ass hair" :)

Anyway, I think that the reference was to the support of x86 as a viable
corporate environment where a Unix box *should* be, is the issue that
David was commenting on.

> 
> PC hardware was not a viable platform for UNIX workstations back then,
> based on its own merits.  Today, I contend that it is, based on its own
> merits.  The load balancing systems are irrelevant.
> 

My experience is far different.  The F5, Cisco, and Alteon units *are*
relevant - otherwise folks wouldn't be buying them.  *I* wouldn't have
bought them.  Your argument holds no water, as it were.  For instance,
check the archives more relation of a specific situation where I used
rather high-end x86 servers behind a load-balancer.

> But then the CPU is the bottleneck.
> 

At the risk of this getting out of context, this is not entirely true. 
I/O operations require very little CPU, and the crossbar technology
solves this anyway.  That's the beauty of crossbar systems, the CPU
doesn't have to process *everything*, and what it does have to process,
gets processed at amazing speeds.  For instance, my Octane can pass info
across the crossbar at 1.6GBps (that's a big 'B') *per XIO port* - of
which it has 4.  I have not seen a single x86 system display this
capability.  In terms of data movement, Unix workstations will beat up,
piss on, and not even kiss on the cheek, the x86 stuff.

However, you have to have the knowledge to design and implement it -
otherwise all that power will go to waste.

> 
> [demime 0.99c.1 removed an attachment of type application/pgp-signature which had a name of signature.asc]

Ugh - please stop this.  Paste it as text...
-- 
Kurt                   "What me look like, ricecake monster?   
kurt at k-huhn.com         Me Cookie Monster!  Me need COOKIE!" --Cookie
Monster