Forums | MacLife

Marketshare is all well and good but if the marketshare is maintainable it doesn't matter if it's low or high. Besides how many times has the popular choice been incongruous with choices that may have worked out better.

avkills wrote:

Mr. T wrote:

avkills wrote:

I don't care what you are converting, if there is converting taking place digitally it takes time and therefore effects performance.

That's certainly true for software, but it doesn't work like that at the hardware level.  As long as the conversion is done in one cycle (and it is - frequently less than one), there's no performance loss (btw, I realize this may seem a little counter-intuitive, but I have a pretty good background in this. ).

You just contradicted yourself.    Even if it takes 1/2 a cycle, that is 1/2 a cycle that normally would not have been used for straight processing.  So say you have a complex instruction that takes 1/2 cycle to break down in micro-ops (or reduced instructions); it would be faster if it was just going straight with the reduced instructions.

Anyway I know it can't be that much of a hit, otherwise modern x86 chips would really suck wind.

-mark

There's no contradiction.  The CPU doesn't burn a "half cycle" on the conversion before processing the micro-ops.  This is done concurrently with the execution of the micro-ops.  There's no wasted time.

Pariah wrote:

ScifiterX wrote:

Marketshare is all well and good but if the marketshare is maintainable it doesn't matter if it's low or high. Besides how many times has the popular choice been incongruous with choices that may have worked out better.

There is something pretty damned hilarious about Mac users all of a sudden deciding that popularity and PC rag reviews are meaningful.

I sure don't.  I've read PC Mag before and they all sound like a bunch Windows fanboys.  I sure as hell will not give them any credit as to knowing anything in depth about OS X or Apple software.  Hardware, yeah, well they do know that since Apple is now using the same commodity parts for the whole kit.

-mark

Mr. T wrote:

There's no contradiction.  The CPU doesn't burn a "half cycle" on the conversion before processing the micro-ops.  This is done concurrently with the execution of the micro-ops.  There's no wasted time.

So you are telling me the processor can decode the long word into the micro-ops and process the same micro-ops at the same time?  Hmmmm....

-mark

Czachorski wrote:

A reader survey is more interesting than a review to me, because it is an indicator of installed performance.

Not too long ago the general consensus around here was that 98% of computer users don't know smurf and thus Windows popularity.
I still think this is correct.

Czachorski wrote:

Pariah wrote:

Czachorski wrote:

A reader survey is more interesting than a review to me, because it is an indicator of installed performance.

Not too long ago the general consensus around here was that 98% of computer users don't know smurf and thus Windows popularity.
I still think this is correct.

I hear the point you are making - if you are judging popularity.  Reliability ratings from an actual installed user base are valuable information beyond a popularity indication.  All the people in the world can buy whatever piss-ass tool for the job that they think is best.  I like to know that the tools I buy are going to have some reliability, regardless of their popularity.

::agrees, thinking  of his Dozuki::