Nvidia Reiterates Position on Closed Source Driver (OSnews)

Posted Jun 26, 2008 18:24 UTC (Thu) by johnh500 (guest, #49452)
Parent article: Nvidia Reiterates Position on Closed Source Driver (OSnews)

Most linux users, especially most developers, will not need the high-end 3D graphics and
computational capabilities of NVIDIA's latest offerings.  This allows users to pick and choose
between AMD cards (sometimes almost as good as NVIDIA's, though usually with a worse
price/performance ratio, and less reliable across the board); Intel integrated graphics (10x
to 1000x worse than a discrete graphics card), generally for simply 2D use; and open-source
drivers for NVIDIA (dicey, but ok for simple stuff).

The power users (3D designers; movie makers; geologist; and zillions more) who actually need
the very best have no choice but to use NVIDIA; there's no one else who comes close at the
high end.  However, open source is not an issue for these users; they only care about results:
does it work, can it be supported. And considering that NVIDIA's support is very strong, the
end users get precisely what they need.

It is long past time to look at the problem to be solved, rather than some random aspect ("is
it open source") of the problem. Linux is actually stronger as a result of real companies
putting billions of dollars into research and development; this cannot be duplicated (yet) by
open source efforts, and the resulting intellectual property has to be protected.  When was
the last time you were able to read the Verilog RTL that explains how Intel's CPU's work, for
example.  

Failing to recognize the basic economics of the situation will lead linux down the wrong path.
There is a place for closed source, and it is in the largest, most expensive research areas.
Again, how many chip companies release the source code to their drivers? Some, not all, and it
generally depends on being able to protect their silicon-based designs fairly secret.

Thanks for reading.
--JH

It's one thing to hide Verilog it's another to stuff huge binary in my kernel

Posted Jun 26, 2008 18:45 UTC (Thu) by khim (subscriber, #9252) [Link] (1 responses)

However, open source is not an issue for these users; they only care about results: does it work, can it be supported.

The question is support, of course. RedHat will not offer you support if you are using binary driver. Ditto for kernel developers. The same for many-many other people. It's quite clear and consistent stance: Linux is free, you can stuff it with binary wireless drivers or graphic drivers but don't come with questions and bugreports in this case.

When was the last time you were able to read the Verilog RTL that explains how Intel's CPU's work, for example.

Looks like you are diliberately confusing things. It's one thing to request verilog - may be it'll be useful, but it's not needed to program chip, only to build it. It's another thing to request documentation - this is a must and Intel's CPU documentation was always very good. nVidia noes not offer anything at all - not even chip interfaces! And with things like CUDA it starts to look more and more like another proprietary OS in the control of your computer. Thnx but no thnx. AMD and Intel now offer quite good documentation and looks like AMD offers good performance as well so there are noreason to stick with nVidia at all. You can expect even less tolerance to the users who are using nVidia drivers from now on.

It's one thing to hide Verilog it's another to stuff huge binary in my kernel

Posted Jun 26, 2008 18:57 UTC (Thu) by johnh500 (guest, #49452) [Link]

Sorry about the confusion; I'm lately finding that reading the RTL for our chip designs in my
own company is the only way to figure out what the crazy hardware guys are up to...

Anyway, you raise a good point: NVIDIA's documentation has always been their weakest point,
assuming you can even find the documentation. I should mention, however, that they do have a
public, documented SDK, and a huge variety of companies are using it to create their own
(proprietary, usually!) applications. The medical industry, for example, has some interesting
high-end, super high resolution flat panels, with drivers that they wrote, using NVIDIA SDK
APIs. Likewise for the movie industry, and so on.