Why "Limiting Factors Theory" is so important for OLPC devel. (Re: [Trac #125] Geode GX is a poor CPU choice (fwd)

[supat at supat.eu.org]

On Tue, 26 Sep 2006, Mitch Bradley wrote:

> supat at supat.eu.org wrote:
>> 
>> Did you understand the theory?
> I can assure that Ivan understands the theory.
> 
> By the way, the English idiom for "first limiting factor" is the word 
> "bottleneck".

No. "first limiting factor" and "bottleneck" is not the same thing.
Seem you did not understand the theory correctly too.

>> 
>> And from experiments it was proven that the theory is correct.
> However, since the final OLPC machine will use on-board NAND FLASH instead of 
> USB storage, the performance of USB storage is largely irrelevant.
>> 
>> Sure, processor has effect to pc performance but not the first limiting 
>> factor.
>> 
> It depends entirely on the application.  For the case that you tested, USB 
> access time was indeed the bottleneck.  It's probably correct to assume that, 
> in your application, mass storage access time is the critical factor.  It is 
> not correct that mass storage access time is the critical factor in *all* 
> applications. It is very easy to find applications that are CPU-bound.  It is 
> also easy to find applications that are memory-bound.
>

It is not exactly fall into your explanation. However, you have better 
understanding about problems more than him.

I know very well about I/O bounded and CPU bounded work and I assure you that 
"the first limiting factor theory" did applied to all works including most 
problems in this world.

Before it has USB problem, it has AMD drivers problems and usb8xxx problems but 
after those problems were solved it appear that USB storage become problem.

I know that we plan to use internal storage and get rid off USB problem but 
those event did not happen yet. If it happen then processor may become the 1st 
limiting factor. I am scientist and will not say the thing that did not happen 
yet.

At this moment total performance of OLPC with AMD processor is acceptible based 
on my experiment.

Sure, changing processor may increase OLPC performance but so far I cannot say 
increasing in performance are significantly different.

good luck,
supat
ps. conclusion: limiting factor is wider meaning than simply bottle neck.