[LUAU] Intel Doubles Down on Linux

[Jim Thompson]

On Jul 27, 2005, at 12:47 PM, Hawaii Linux Institute wrote:


> Jimen Ching wrote:
>
>
>
>> As for 'accelerated' drivers; I recommend taking those comments  
>> with a
>> large grain of salt.  At work, a vendor says the video card and  
>> the driver
>> they provided were 'accelerated'.  But we found otherwise during  
>> regular
>> use...
>>
>>
>>
>>
> I think we are getting into the core of this subject.  Writing a  
> device driver (& advertising it as such) is easy.  But writing an  
> optimized driver for a device that's worth hundreds of millions of  
> dollars (as in the case of nVidia's accelerated video cards), is  
> not.  It was not until very recently that I decided that there are  
> enough benefits to switch from "nv" to "nvidia" driver for my  
> nVidia FX 5200 cards.
>

The issue with nVidia isn't so much "complexity" as "obscurity".  Few  
outside nVidia grok the GPU pipeline of the nVidia cards, and these
are likely all under NDA, which obviates any "open source" (or  
*free*) drivers for these chipsets.

nVidia isn't alone in this regard, either.


> Everytime I heard complaints about how stupid/backward X is, I  
> always ask the instigator, whoever s/he is, to look at the Linux/ 
> UNIX version of Abode Reader  7.0 vis-a-vis the Windows version  
> (though I never did this in a public forum).  The point is not to  
> prod how great X is (am I going to kid myself?) but how far X has  
> progressed and how intimate the gap can be narrowed if enough  
> sources are devoted to improving an X app.
>

Cramming PS (or PDF) onto a page isn't that big a deal.   If you're  
talking level of polish/finish, then thats up to the programmer and  
designer.    I'd beat the point about imaging apps and the future of  
X .vs Windows (Avalon) and MacOS (Quartz Composer), but that horse is  
dead, or at least lying in the ditch.


> For a matured program running on a desktop machine (meaning that  
> the app does all you want to do and you are familiar with how the  
> app operates),
>
(and its stable)

> as far as user experience is concerned, driver is everything.
>
Uh....  the driver can't overcome a poorly-coded app, and its EZ (or  
at least straight-forward) to code an app that will perform quite  
poorly on any platform.


>   In the past, at least on the x86 side, device providers (most of  
> them are based in Taiwan),
>

nVidia's GPU software developers are mostly in Canada.


> either (1) don't know/care about the Linux kernel, (2) don't have  
> any control/influence over how Linux kernel is developed, (3) don't  
> give a damn about Linux driver or assign the job to entry-level  
> employees, or, most likely, (4) all of the above.
>
> Intel's move (to double down on Linux), if true, will eventually  
> elevate the status of certain (i.e., Intel-made) Linux device  
> drivers to that of Windows, thus opening up an opportunity for  
> Linux to be acceptably considered in the desktop arena.
>

Other vendors support their chipsets, and that hasn't made Desktop  
linux succeed.   There are a plethora of issues with "desktop linux"
for the mainstream.   i doubt that Intel fixes even half of them.


> (& "Intel Inside" will no longer mean "Idiot Inside".)  But how  
> should the Taiwanese periphery device makers respond to Intel's  
> move, is something their top execs should be deeply concerned  
> about.  (A case in point: Intel's Centrino chipset has pretty much  
> driven Taiwanese chipset makers out of the NB business.)
>

Hardly.  Nearly every major laptop vendor (Dell, IBM, HP, etc) offers  
a non-centrino notebook.   All but Dell offer AMD powered notebooks,  
and these can't be Centrino, either.

In any case, "Centrino" isn't a chipset, its a branding strategy.    
Unless you take all the specified Intel silicon content, you can't  
slap the "Centrino" brand on your notebook.   Specificly, you need to  
have a Pentium M processor, Intel's 855GME GPU (Centrino) or 915GME  
(Sonoma, which is required for "Centrino II") and Pro/Wireless Network).

Intel would *love* to have you believe that "Centrino" is a chipset,  
but its just not true.  Check these out:
http://blogs.zdnet.com/BTL/?p=1584
http://blogs.zdnet.com/BTL/?p=1557

In any case, actions speak louder than words, and Intel hasn't been  
that supportive of linux (wrt Centrino) in the past. Intel took well  
over a year after the initial Centrino launch to release even  
preliminary linux support for Centrino notebooks.  Intel also backed  
out of a commitment to have Michael Robertson (of lindows) on the  
original Centrino roadshow.

jim