[Cooker] Doubt: x86_64 or i586?

[Adam Williamson-4]

On Mon, 2008-08-11 at 09:31 -0700, Adam Williamson wrote:

> > Yep
>
> The reason things are the way they are is a dbarth philosophy: he wanted
> One to be the version almost everyone used. Originally the page was not
> going to mention Free at all. Just for information.

To clarify (Thanks Fred) I'm not trying to bash this idea or say it was
a one-person show - I'm just explaining that the way the current page
works is the result of a conscious choice / philosophy, it's not just a
case of a badly-written description or something.
--
adamw

[Luiz Alberto Saba]

Adam Williamson escreveu:

> On Mon, 2008-08-11 at 16:54 +0200, Eric Piel wrote:
>  
>> Amrein-Marie Christophe schreef:
>>
>>    
>>> *Download the latest Mandriva Linux release*
>>>
>>> *** Mandriva Linux One 2008 Spring - 32bits life CD edition -*
>>>      
>>                                                ^^^^
>>                                                live
>>    
>>> */Recommended/: contains all drivers and common plugins
>>>
>>> *** Mandriva Linux Free 2008 Spring - 32bits & 64bits edition - The
>>> "purely" Free Software edition  (/See why you should care about Free
>>> Software <http://mandriva.com/en/community/free-software>/) *Warning:
>>> *does not contain any proprietary drivers or software
>>>
>>>
>>> Better ?
>>>      
>
>  
>> Yep
>>    
>
> The reason things are the way they are is a dbarth philosophy: he wanted
> One to be the version almost everyone used. Originally the page was not
> going to mention Free at all. Just for information.
>
> The problem with starting to introduce jargon like 32-bit, 64-bit and
> "live CD" is it confuses people who don't know what it means. But it
> still may be worthwhile. However, it's also worth bearing in mind we
> currently have another group of people saying the page should
> "obviously" make available both the KDE and GNOME versions of One, and
> another group saying it should "obviously" make the other language
> versions of One more prominent. If we start doing everything someone
> thinks is obvious, we wind up with a mess like the old download page
> (the reason we went with a simple download page for this iteration of
> mandriva.com is quite simple: we had a stat showing that onlt 33% of the
> people who visited the old download page actually got around to
> *downloading* anything, there was so much explanation to wade through.
>  

Fine. But... The reason for my original post is still here.
How can someone know which version is more adequated to his specific
hardware?
For an Intel quad-core should I use 32 or 64 bits? "Obviously" 64? Why?
This processor is not a true 64 bits.
What are the pros & cons of each version?
Free drivers?
Fine again...
But, if the user is a gamer or other kind of graphic intensive user, the
free drivers are not always the best choice...
It's not "marketing palatable" to put this kind of info in the download
page? At least a link to some kind of "How to choose the right version
for you"...

[Gustavo De Nardin (spuk)]

* Luiz Alberto Saba <las@...> [2008-08-11 14:21 -0300]:

> Adam Williamson escreveu:
> >The reason things are the way they are is a dbarth philosophy: he wanted
> >One to be the version almost everyone used. Originally the page was not
> >going to mention Free at all. Just for information.
> >
> >The problem with starting to introduce jargon like 32-bit, 64-bit and
> >"live CD" is it confuses people who don't know what it means. But it
> >still may be worthwhile. However, it's also worth bearing in mind we
> >currently have another group of people saying the page should
> >"obviously" make available both the KDE and GNOME versions of One, and
> >another group saying it should "obviously" make the other language
> >versions of One more prominent. If we start doing everything someone
> >thinks is obvious, we wind up with a mess like the old download page
> >(the reason we went with a simple download page for this iteration of
> >mandriva.com is quite simple: we had a stat showing that onlt 33% of the
> >people who visited the old download page actually got around to
> >*downloading* anything, there was so much explanation to wade through.
> >  
> Fine. But... The reason for my original post is still here.
> How can someone know which version is more adequated to his specific
> hardware?
> For an Intel quad-core should I use 32 or 64 bits? "Obviously" 64? Why?
> This processor is not a true 64 bits.

If it isn't 64bit, why are you choosing?


> What are the pros & cons of each version?
> Free drivers?
> Fine again...
> But, if the user is a gamer or other kind of graphic intensive user, the
> free drivers are not always the best choice...
> It's not "marketing palatable" to put this kind of info in the download
> page? At least a link to some kind of "How to choose the right version
> for you"...
>

[Luiz Alberto Saba]

Gustavo De Nardin (spuk) escreveu:

> * Luiz Alberto Saba <las@...> [2008-08-11 14:21 -0300]:
>  
>> Adam Williamson escreveu:
>>    
>>> The reason things are the way they are is a dbarth philosophy: he wanted
>>> One to be the version almost everyone used. Originally the page was not
>>> going to mention Free at all. Just for information.
>>>
>>> The problem with starting to introduce jargon like 32-bit, 64-bit and
>>> "live CD" is it confuses people who don't know what it means. But it
>>> still may be worthwhile. However, it's also worth bearing in mind we
>>> currently have another group of people saying the page should
>>> "obviously" make available both the KDE and GNOME versions of One, and
>>> another group saying it should "obviously" make the other language
>>> versions of One more prominent. If we start doing everything someone
>>> thinks is obvious, we wind up with a mess like the old download page
>>> (the reason we went with a simple download page for this iteration of
>>> mandriva.com is quite simple: we had a stat showing that onlt 33% of the
>>> people who visited the old download page actually got around to
>>> *downloading* anything, there was so much explanation to wade through.
>>>  
>>>      
>> Fine. But... The reason for my original post is still here.
>> How can someone know which version is more adequated to his specific
>> hardware?
>> For an Intel quad-core should I use 32 or 64 bits? "Obviously" 64? Why?
>> This processor is not a true 64 bits.
>>    
>
> If it isn't 64bit, why are you choosing?
>  

1. The Intel quad-core processors (exception made to xeon) are 32 bit
processors capablabe of dealing with memory "like" 64 bit processors
(EM64T). They're NOT 64 bit bus processors.
2. I'm NOT "choosing 64 bit" (in fact, I'm using i586). I'm complaining
about the lack of information about how to choose the right version.

>
>  
>> What are the pros & cons of each version?
>> Free drivers?
>> Fine again...
>> But, if the user is a gamer or other kind of graphic intensive user, the
>> free drivers are not always the best choice...
>> It's not "marketing palatable" to put this kind of info in the download
>> page? At least a link to some kind of "How to choose the right version
>> for you"...
>>
>>    
>
>  

[Thierry Vignaud]

Luiz Alberto Saba <las@...> writes:

> >> Fine. But... The reason for my original post is still here.
> >> How can someone know which version is more adequated to his
> >> specific hardware?
> >> For an Intel quad-core should I use 32 or 64 bits? "Obviously" 64?
> >> Why? This processor is not a true 64 bits.
> >>    
> >
> > If it isn't 64bit, why are you choosing?
> >  
> 1. The Intel quad-core processors (exception made to xeon) are 32 bit
> processors capablabe of dealing with memory "like" 64 bit processors
> (EM64T). They're NOT 64 bit bus processors.

This is meaningless. Which bus? there're tons of different ones on a
motherboard (CPU's address bus & data bus, different links (between
CPUs, between CPU & northbridge, ...).

Adress space width, data bus width and register size are all
orthogonal things...
Eg:
- 80286 is a 16bit CPU with a 24bit address bus
- ia32 processors had a 32bit address apce but already used a 64bit
  data bus since Pentium (== since _1993_).

What's more address space and actual physical memory space can be
different: 64 vs 40 bit (44 on newer models) on x86_64,

Pentium Pro & above are able to use 36bit physical addressing (in PAE
mode or in PSE-36 mode since PentiumIII).

It's mainly usefull for the system in order to manage more than 4Bg
ofphysical memory, not for applications directly which are still
restricted to 4Gb virtual space each.

Still their register size is 32 bit (partly accessible as 16 bit
registers for historical ones).


Since ia32 is just a 32-bit superset of the 8086/80286 instruction set
architecture, does this mean that ia32 is not a real 32 bit
architecture... :-) ? If you answer no, then since x86_64 is just a
64-bit superset of the x86 architecture, there's no confusing here.

Anyway, the x86_64 distribution targets x86_64 processors. Point.


> 2. I'm NOT "choosing 64 bit" (in fact, I'm using i586). I'm
> complaining about the lack of information about how to choose the
> right version.

The 64bit edition is the specific edition for x86_64 (aka for both
Intel EM64T processors & AMD's AMD64 processors).  Period.
Claiming x86_64 processors "re NOT 64 bit bus processors" doesn't
affect what target is the x86_64 distribution.

[Luiz Alberto Saba]

Thierry Vignaud escreveu:

Luiz Alberto Saba las@... writes:

  

Fine. But... The reason for my original post is still here.
How can someone know which version is more adequated to his
specific hardware?
For an Intel quad-core should I use 32 or 64 bits? "Obviously" 64?
Why? This processor is not a true 64 bits.
    
        

If it isn't 64bit, why are you choosing?
  
      

1. The Intel quad-core processors (exception made to xeon) are 32 bit
processors capablabe of dealing with memory "like" 64 bit processors
(EM64T). They're NOT 64 bit bus processors.
    

This is meaningless. Which bus? there're tons of different ones on a
motherboard (CPU's address bus & data bus, different links (between
CPUs, between CPU & northbridge, ...).

Adress space width, data bus width and register size are all
orthogonal things...
Eg:
- 80286 is a 16bit CPU with a 24bit address bus
- ia32 processors had a 32bit address apce but already used a 64bit
  data bus since Pentium (== since _1993_).

What's more address space and actual physical memory space can be
different: 64 vs 40 bit (44 on newer models) on x86_64,

Pentium Pro & above are able to use 36bit physical addressing (in PAE
mode or in PSE-36 mode since PentiumIII).

It's mainly usefull for the system in order to manage more than 4Bg
ofphysical memory, not for applications directly which are still
restricted to 4Gb virtual space each.

Still their register size is 32 bit (partly accessible as 16 bit
registers for historical ones).


Since ia32 is just a 32-bit superset of the 8086/80286 instruction set
architecture, does this mean that ia32 is not a real 32 bit
architecture... :-) ? If you answer no, then since x86_64 is just a
64-bit superset of the x86 architecture, there's no confusing here.

Anyway, the x86_64 distribution targets x86_64 processors. Point.


  

2. I'm NOT "choosing 64 bit" (in fact, I'm using i586). I'm
complaining about the lack of information about how to choose the
right version.
    

The 64bit edition is the specific edition for x86_64 (aka for both
Intel EM64T processors & AMD's AMD64 processors).  Period.
Claiming x86_64 processors "re NOT 64 bit bus processors" doesn't
affect what target is the x86_64 distribution.
  

Thanks, Thierry!

Why not make this kind of objective info available to the user at install time?
I think that it will be extremely usefull and will avoid confusions.