Re: [MLUG - DISCUSSION] VR3

[Kmicic <EMAIL:PROTECTED>]

"Ross, Matt" wrote:
 
> Its only a misconception if its wrong.  They wouldn't still be wasting two
> processors on a motherboard if it had proven to be wrong.

for a long time manufacturers have been trying to push DDR ram when
there was virtually no advantage to it (may have changed by now). I
don't buy this argument, although I have to admit I have not provided
any evidence to the contrary either. 

> > > http://physics.indiana.edu/~sg/CSE2001/singlenode.html -
> > some benchmarking
> > > results involving both single and dual
> >
> > Correct me if I'm wrong, but in this experiment the higher the number,
> > the better the performance. Shows that dual Athlons are consistently
> > behind the same single processor machines on the tested tasks.
> 
> Read the page, some of them say "higher number = faster" some say "lower
> number = faster".  Not one of them showed that a dual athlon was behind a
> single.

I have read it. Towards the bottom it has a comparison of single and
dual Athlons. On this page "higher number" consistently means better,
although I have not been able to track down explanation of the numbers
(must be buried on some other part of the site). I think it has
something to do with MIPS, but not sure.


 
> Thats a rather interesting, and incomplete analysis.  I think any server
> administrator will tell you he's leaving out a whole spectrum of
> applications, including database management systems, which we were told the
> box in question pertains to.

When you look at Jason's post he says he's doing a single user
development work, and his server is not public. Reason multiple
processors are good for servers is because of the underlying assumption
that servers will service many independent requests which rarely come
into conflict. 

Going back to java - which prompted my outburst (Sorry, looks like I
fired off two offensive posts there). In java (at least on linux) you
have green threads of native threads, and you can run either or, but not
both. Native threads run in kernel space as processes. Green threads run
in user space under a single process. Everybody uses green threads -
because it is a known fact that they're generally much faster that way.
I would like and welcome evidence showing native threads on dual proc
beating green threads on single proc. 

> > I'll leave it at that - don't really have the time for a quote war. As
> > far as evidence to support my claims - it can be found in any
> > distributed computing textbook.
> 
> I think we've all read our fair share of them, but get real, if there
> weren't a distinct advantage to having two processors, why would there even
> be such a beast after so many years?

For example: so that geeks would have to explain to everyone else that
"no, 2 processors x 1Ghz is NOT 2Ghz" while the salesmen give them the
funny look and give you two processors for (almost) the price of one. 

I'll probably try a dual proc sooner or later, but I don't expect much
from it in terms of reliability (and performance). Both my AMD's (athlon
and duron) freeze regularly once or twice per month under win2k (asus
and abit mobo's). I don't think putting extra processor is necessarily a
way to make AMD+win2k run more reliably.

Paul
--
To unsubscribe, go to http://mlug.missouri.edu/members/edit.php

Archives are available at http://mlug.missouri.edu/list-archives/