Thursday, November 25 2004 @ 07:18 PM CET Contributed by: theosib Views: 6216
In my experience, and apparently the experience of others, finding a
graphics card which is supported by open source drivers is a bit of a
challenge. Either the open source driver support is poor or limited,
or there are simply no open source drivers at all. Many people resort
to buying used cards from eBay.
Frustrated, I have decided to start a project to solve the problem.
Since I am a chip designer, and I work for a company that sells
graphics cards, I decided to approach management with the idea, and
they agreed that if I could justify it economically, we would do it.
The primary objective is to produce a graphics product which has fully
open documentation and fully open source driver.
When I first brought up the idea on the Linux Kernel Mailing List, it
received quite a bit of attention. Lots of discussion on LKML
resulted in an article on kerneltrap.org, followed by a link from
Slashdot.
In addition, one of the list members put online a petition pertaining
to pricing of the proposed graphics card: http://www.petitiononline.com/3dc4rdlb/petition.html
(I think the petition will be valuable feedback for the company to
determine viability.)
As someone who uses Linux as his primary platform, both at home and at
work, I would really love to see something like this come to fruition.
Unfortunately, the attention people have paid to this project has
started to wane. I figure either people really aren't that
interested, or they just don't know about it. I decided to test the
waters by posting to a few different forums around on the web. It
turns out that there are plenty of interested people, but they just
haven't heard about it because word hasn't spread.
Although my employer is capable of and willing to help me to market
this, they have not taken an intrusive role in the project, in part
because they understand that this is not a normal project. This is
more of a community project than it is a "Tech Source" project, and
they have respected that. Unfortunately, being an engineer rather
than a sales person, I haven't had huge success on my own in
generating the necessary awareness among those who would be interested
in this sort of thing.
To that end, I would like to humbly ask interested parties who read
this to please pass the word around. (And also participate in the
discussions and development process.)
I hope you don't see this as a plea for free advertizing but rather as
an opportunity to take part in the development of graphics hardware
which is compatible with Free Software principles. I honestly believe
that there is a major need, and I would like to be part of meeting that need.
The following comments are owned by whomever posted them. This site is not responsible for what they say.
Graphics card with open source drivers!
Authored by:
bart on
Thursday, November 25 2004 @ 11:33 PM CET
I definitely see a need for this. As you can read in my article about comparison of raphics cards on FreeBSD, the only thing that really works and performs well is the hardware that NVIDIA offers, and that doesn't come with open source drivers
In the past I went with Matrox because of their good support for OSS, but with the P series that seems to have gone out of the window, and the drivers they have are not usable for most people from what I can tell, which basicly means you can go with outdated hardware and OSS drivers, or modern hardware and no accelerated 3d support if you get drivers at all (unless we are talking nvidia or in some cases ati)
So.. keep up the good work, and keep us posted please.
Authored by:
theosib on
Thursday, November 25 2004 @ 11:47 PM CET
Thank you for the encouragement. I'm hoping that people
will talk about this amongst each other and that soon the
awareness will reach critical mass. I'm very anxious to
get started on the implementation, but that can't happen
until enough people (in the thousands) have demonstrated
an interest.
Authored by:
bart on
Friday, November 26 2004 @ 12:06 AM CET
Reaching critical mass is often the problem with such projects it seems. Judging from the references I get from Google, quite a few people are in fact looking for information on graphics cards on Linux (and to a lesser extent, FreeBSD), so you'd think the potential is there.
Looking around myself I find lots of different posts to mailinglists but no prominent repository of information and a (web based) discussion board related to graphics hardware on open source platforms.
Authored by:
theosib on
Friday, November 26 2004 @ 12:43 AM CET
Well, if my mailing list (mentioned in the article) isn't
enough, then, yes, someone should set up a web page all
about this. Any suggestions? Perhaps someone on my
mailing list could do it.
Authored by:
bart on
Friday, November 26 2004 @ 12:50 AM CET
Hmm, a mailinglist works fine for discussing I think, even more so when it has a good archive. I do think that web based discussions are easier to find for 'outsiders', and also easier to read and join. Mailinglists are somewhat easier to follow for 'insiders' tho I think.
I can easily setup a site for it, but I'd only do so if people are interested in participating (heh... the critical mass problem again) so that it gets content from multiple sources.
Right now I feel more like adding a seperate section for it to my site tho.
Authored by:
theosib on
Friday, November 26 2004 @ 01:22 AM CET
I would be very happy about it if you had a special
section for this on your site. We should cross link
between the mailing list and your discussion forum. Just
give me an appropriate URL, and I'll add the link to the
list main page.
Authored by:
theosib on
Friday, November 26 2004 @ 01:46 AM CET
Excellent! Thank you! Now, my only concern is that I may
not be able to do a good job of following both forums,
particularly when I get onto doing the development work.
Hopefully, some members of both forums will cross-post
summaries of interesting discussions.
Authored by:
Anonymous on
Saturday, November 27 2004 @ 05:57 PM CET
Just an artist passing through. I wanted to let you know that this is very much a valuable venture for all concerned. I am keenly interested in a cutting edge FOSS compatible graphics card with good nonproprietary drivers. You should really also seriously consider informing GIMP users also. I heard of your efforts on a Blender3d site and was greatly encouraged by the idea.
Authored by:
theosib on
Saturday, November 27 2004 @ 10:02 PM CET
Thanks for the tip. I checked out the GIMP site, and I
decided that it was probably off-topic to post to any of
their mailing lists (I can't affort to get on people's bad
side). I decided to email their webmaster, but I don't
know if it'll get anywhere.
While I have been received well in the places I have
posted, I'm concerned that my obvious bias towards my own
project will count against me. It's better when people
who are not me spread the word around, especially when
there is more than one of them.
Authored by:
Anonymous on
Sunday, November 28 2004 @ 07:13 AM CET
After reading the spec and thinking about how the bazaar works, I recommend you add the "VGA feature connector" to the spec. Make sure that the pin assignments onto the logic chip are compatible with the signal directions called out in the specification ... then ignore the connector.
In the short term, it will be handy as a high speed debug port. It would be easy to put two cards in adjacent slots and program one to be a high performance data capture service that monitors the feature connector of the other. That will let someone with two cards contribute to cutting edge development (single headed) and then reflash the boot chips with the stable binary for a normal day-to-day (dual headed) display.
In the medium term, someone (not me, thankfully) who has a legacy need for the feature interface might actually generate a patch to the main logic files to implement the proper capability.
In the longer term, it seems feasible to have several boards in a multiheaded configuration that are nominally rendering in parallel. However, for the display seams and the like, it is handy to have a direct connection between the boards. Even if there is not enough space to do T&L in one chip, it might be effective to stream the T&L data to one board over DMA, have it deliver the preprocessed and optimized data over the one inch ribbon cable to the other board, where its DMA channel is delivering the non-coordinate data and these two inputs are merged into the rendering engine.
Authored by:
bart on
Sunday, November 28 2004 @ 06:48 PM CET
Out of curiosity, is there hardware around that actually made use of this connector? and if so what kidn of hardware? I have seen the connector quite often (and still have 2 cards around that have one), but I never saw it actually being used for anything other then sticking a mpeg decoder card onto your video card.
Another question is that for all I know the feature conenctor is 8bit. Woudl that be workable with a modern graphics card?
In the past I went with Matrox because of their good support for OSS, but with the P series that seems to have gone out of the window, and the drivers they have are not usable for most people from what I can tell, which basicly means you can go with outdated hardware and OSS drivers, or modern hardware and no accelerated 3d support if you get drivers at all (unless we are talking nvidia or in some cases ati)
So.. keep up the good work, and keep us posted please.