[Herts] Open graphics card + Computer dreams [rant]

Malcolm Smith spock at canopus22.demon.co.uk
Tue Jan 16 22:25:08 GMT 2007


On Mon, Jan 08, 2007 at 08:59:40AM +0000, Steve Clark wrote:
> I'm sure there was mention on this project at a recent meeting. I didn't
> realise it was so advanced
> 
> http://wiki.duskglow.com/tiki-index.php?page=Open-Graphics
> 
> It looks like they have prototypes of a PCI card, but I don't see when
> it is going to be generally available or at what price. They say it
> won't compete with the latest from the big boys, but it may be adequate
> for general usage by those who want an open option.
> 
I've been following the progress of this project since its inception
years ago. It is very cool that they have now made a real card that
does everything I would want from a VGA card: dual DVI, hi-res, silent
(no fan), and most importantly, *totally open* drivers. AFAIK there
is no other graphics card that fulfils all these criteria.

My Matrox G450 DualHead has open drivers but only 16MB RAM which I don't
think will power both of my 1280x1024 screens :-( The Matrox G550 is
only slightly better, but the last I heard (after extensive research) to
get Dual DVI, you needed to use Matrox's own proprietary driver, which
is unacceptable. NVidia are all non-free. IIRC, the last free drivers
for ATI cards were for Radeon 9000/9200. Since then, manufacturers have
closed their specs, possibly due to patent fears. All other modern
graphics card development has been led by gaming, i.e. hot fast noisy
monsters, all very closed source.

'Modern' [2004!] Video Cards with Open Specs?
http://slashdot.org/article.pl?sid=04/06/02/1814228

So the OGP card is the only option for me, at any price. It's good to see
hardware being developed like this, even if only a limited prototype at
high cost ($700+). Next comes a lower spec PCI single-head card, to be
mass-produced and marketed for low cost embedded/SFF use, then the 2nd
Generation should be a dual-head AGP/PCI-Express, which I will buy.

I was drawn to an insightful quote on this page:

http://wiki.duskglow.com/tiki-index.php?page=OGPN17

  List Discussions
  Dumb Terminals
 "[...] The 'CPU server' is just a 2.60 GHz Northwood - just one
 core, and probably can't cut (decent) real time 'raytracing'.
###############################################################
# Obviously it would be better to have 16 cores, but it seems #
# PC hardware is going nuts and replacing everything - I have #
# to upgrade from AGP and PCI to PCI Express, from DDR to DDR #
# Two, from PATA to SATA.                                     #
###############################################################
 Now, not only are commodity parts useless, but drivers have
 to be written"

He's right: I'm starting to think that the computer hardware industry
has lost its way. Things should be parallelising, becoming more modular
and scaleable. Instead we have to trash stuff every three years and
start all over again with a new cheap'n'cheerful box. If only we could
*add* to what we already have to get more processing+storage, we'd be
happier and there would be even more incentive to buy more tech without
the headache of incompatibility. The current PC market is dead because
everyone's computer is already way fast enough to cater for the average
person's needs, and it's just too much hassle to change. It's just the
freaks like me who want to process/generate 20+ audio/video streams
in realtime [and possibly M$ Vizta (l)users] that need the bleeding edge.

The industry would need to spend some time designing a modular standard
that could last for at least a decade, and even then still be useable
with the hypercomputers of the next gen. Start with some ridiculously
over-engineered bandwidth, and then just watch as it gets utilised :-)
I want to be able to buy a basic system, and add more processors as
and when I can afford.

The cost of implementing systems and programming their vast capabilities
is going to become the limiting factor, not the hardware cost. The world
needs systems that last longer than a few years - lifetimes need to
mature. Keep the old kit to do simple stuff like scanning for user
input, while the new gear can mesh with it and crunch the AI.

Building Chips Like LEGO
http://hardware.slashdot.org/article.pl?sid=07/01/15/1331234


Anyway, I'm hoping that there will one day be available some decently
affordable open LinuxBIOS 4-way/8-way computer that I can buy just in
time for one final upgrade before the DRM lockdown begins...

http://www.linuxbios.org/

Evilution: (all still in use!)                        Soon...
  Beeb -----> 486DX4 ---> PIII ---------------------> Octopteron
  1MHz *100 = 100MHz *7 = 733MHz * 10 * 8 DualCores = 8-way DualCore 7GHz
         ^            ^                               ^      Add one CPU+
Fun      |            |                               |     RAM each year
factor: 100 :-)       7 :-/                           160 ;-)

Memories...
  32KB *2000 = 64MB * 8 = 512MB * 32 * 8 = 16GB per DualCore proc

</Here endeth the sermon.>


> See you Wednesday.
> 
Sorry I missed it.
For those who missed December's violent demonstration:

http://www.thesmith.org.uk/words/journal/2006-12.html#scotalk


This looks fun too:

Alan Cox Files Patent For DRM
http://linux.slashdot.org/article.pl?sid=07/01/14/2252202

http://uncyclopedia.org/wiki/Time_Travel_Patents


Bye
-- 
Malc 
http://www.thesmith.org.uk/



More information about the Herts mailing list