DVI output, ATI or nVidia

alan alan at clueserver.org
Tue Jun 26 23:51:53 UTC 2007

On Wed, 27 Jun 2007, Luciano Rocha wrote:

> On Wed, Jun 27, 2007 at 07:08:34AM +0800, Ed Greshko wrote:
>> Luciano Rocha wrote:
>>> You're comparing apples to oranges. On one hand, you have people
>>> complaining that a vendor refuses to give full documentation for a piece
>>> of hardware the user bought and wants to use.
>> So, it is a bad thing for a hardware maker not to give away its trade
>> secrets.  Right?
> No. But I hardly think that the documentation of how to make some piece
> of hardware do its work a mater of trace secrets. Do you also think that
> the x86_64 or altivec would bring more profits to their respective
> creators if they were kept as trade secrets?
>>> On the other hand, you have a piece of software that isn't needed for
>>> the system to work at it fullest potencial.
>> In my case I deem it so.  I need something like VMware (I actually use
>> Workstation) since I need to test products on multiple O/S without the need
>> to have tons of hardware or an N boot-able system.
>> Graphics performance is a very small part of my system's "potential".
> Well, in some cases you can't even *use* the graphics card. Not even for
> 2D.

There are some programs that do not run well without hardware 
acceleration.  Run LookingGlass without hardware acceleration for an 
example of this.

>>> I won't complain about vmware not being OSS. I'm not locked to it.
>>> I do complain that the Nvidia binary is a blob. It had a security
>>> problem some time ago and I had to wait for the vendor to issue a proper
>>> fix, and I can't use 3d acceleration in one of my systems (PPC).
>> Of course you are not locked into nVidia either.  Don't like their policy,
>> don't buy their hardware.  Sounds simple enough to me.
> So I don't. But I bought the hardware at a very affordable price (and
> at the only price I could afford the laptop) and at a time when there
> were no OSS 3d driver (for any driver). I would had gone for an ATI, but
> the new models were no longer supported by the OSS driver (ATI changed
> the internal workings and no longer bothered to release the specs).

ATI has a history of releasing specs that barely matched the hardware to 
begin with.  They just took it one step further.

>>> I'm not even complaining that nvidia doesn't release the drivers as OSS
>>> (though it would be nice), but I will keep complaining about releasing
>>> proper, unencumbered documentation about the hardware.
>> I guess I'm not clear on what kind of unencumbered hardware documentation
>> you want nVidia to supply.  Care to elaborate?  Or, maybe point to the
>> equivalent documentation that Intel provides on their video hardware.
> I don't know the documentation that Intel provides, it's late and I'm
> too tired to search for it. But they do provide full OSS 3d drivers.
> It's a (relatively) change for Intel, much appreciated, and when I can
> afford a new machine, I'll choose them.

Keith Packard works for Intel now.  He has been one of the main people 
writing the open source drivers.

>>> The Nouveau project is trying to document and implement a 3d driver, see
>>> http://nouveau.freedesktop.org/wiki/FAQ. But the last time I tried the
>>> graphics card entered an invalid state just after entering my username
>>> in gdm.
>> Yeah, I've heard about that...but I still can't figure out why I would use a
>>  free "clone" of something else that is already free and supported by the
>> OEM.  Can you help me out there?
> I thought I already did. The FAQ states:
> <quote>
> 1.9. Why are you doing this?
> We can't give you the answer, as each of the project members has his own
> motivation. Just a few answers from our staff, we got when this question
> was raised:
>    * Don't like binary blobs
>    * Want to give back to the OSS community
>    * Want to learn driver programming
>    * Yes, we can develop our own drivers regardless of what people at
>      NVidia may think
>    * Support for missing features
>    * Support for operating systems not supported by NVidia (any PowerPC
>      based OS for example)
>    * Just for the fun of it
>    * Binary driver keeps crashing even in 2D
>    * Slow Xorg "nv" driver (slow in performance and slow to get new
>      card support (Nvidia 8800 is currently not supported))
> <end quote>
> I haven't participated on the project, but my motivations for wanting an
> OSS driver are:
>    * Don't like binary blobs (can't patch vulnerabilities or try and
>      fix bus)
>    * Support under Linux/PPC.
> Please note that I have personal reasons for complaining, but I don't go
> out of my way to complain (I signed the petition, but never posted an
> "nvidia sucks because no docs or oss driver!", or something like that).
> I just wanted to explain that the people complaining about closed nvidia
> driver aren't the same people *not* complaining about closed vmware nor
> would they have the same reasons. I don't want to start a thread about
> my rights vs nvidia's rights.

There are also distributions that will not ship binary only drivers.  In 
order to get the card to "work out of the box", they want something they 
can ship.  2D only support is not enough these days.  Not with users 
wanting eye-candy like Compiz/Beryl/etc.  (Even if they don't want that, 
they really need the hardware acceleration.  That is not usually a part of 
the 2d-only card drivers.)

"ANSI C says access to the padding fields of a struct is undefined.
ANSI C also says that struct assignment is a memcpy. Therefore struct
assignment in ANSI C is a violation of ANSI C..."
                                   - Alan Cox

More information about the fedora-list mailing list