Slashdot Mirror
Weather Channel Sponsors OSS ATI Radeon Drivers
jvmatthe writes "Jens Owen of Tungsten Graphics (mostly former VA Linux/Precision Insight employees) posted to the DRI developer's mailing list with some excellent news about the future of DRI drivers for the ATi Radeon 8500 video card: "The Weather Channel is funding TG to develop an open source 3D DRI driver for the ATI Radeon 8500 graphics card. The driver will be released to the XFree86 Project around Q4 of 2002, to be distributed to the public in future versions of the XFree86 X Server."
Presumably this means that this Weather Channel is the one footing the bill for the development. Given that the current Linux support for the 8500 is limited to a binary-only driver that is intended for a related professional-level card, the delivery of an open driver is excellent news. This is also listed at the bottom of the TG project page."
They're probably intrested in new and interesting graphics for their channel and most of their software probably runs on a system that uses Xwindows. By funding the development they not only get the drivers they need but also get to help out the community.
Hi The Weather channel bought about 5000 rackmount SGI O2's a few years ago and now they need replacing Jack.
This is great news and I really hope this will be the start for more open source drivers. The graphics acceleraters marked moves very fast, what was fast 1 year ago won't run the latest games today. We need drivers before the card is released or when it's released like they have in windows (the latest radeon drivers has support for rv300). Not something like 1 year after it's released.
Is it just me or is the world getting more surreal every day? I mean we now have everything from Russia joining NATO to The Weather Channel supporting drivers for my video card. Things are just getting to weird for me.
Keep Austin Weird!
The Nvidia still lacks 10-bit precision in the video overlay which is probably very important for broadcast video work. The ATI 8500 (and 7500) and the new Matrox all support at least that level of precision.
When information is power, privacy is freedom.
I know of no such card. That is, in fact, a chipset. Nvidia does not manufacture any cards themselves. And the problems with 2d quality have never had to do with the chipsets themselves, but with the post-DAC filtering. And this is highly dependent on the board maker. So, there have always been some people whose nvidia-based cards have looked just fine, and others who have gotten crap. It may be that the standards on the reference cards have gotten higher, so that there are more in the first category now than in the past, but it still sounds like it is highly variable depending on board manufacturer.
Meanwhile, get a Matrox card, and you are *guaranteed* top-class 2d output. This is why they are still business, considering the poor performance of their recent parts, and an area in which they still have yet to be bested.
"(Man) tries to live his own life as if he were telling a story. But you have to choose: live or tell." --Sartre
As a support to the Open Source community, I am gonna do my part, by watching the Weather Channel now.
Yeah, call me a brown-nose, but that's the least I can do, right ?
Muchas Gracias, Señor Edward Snowden !
ATI releases specifications of its hardware to developers.
An open DRI driver for Radeon 8500 (which seems to be completely different from Radeon 8800 wrt 3D) based on specs given by ATI has been in the work for some months, but development has been somewhat slow, and it's not ready yet (2D works fine, though).
In the meantime, ATI also released a (proprietary) driver for Radeon 8800 that seems also to provide 3D acceleration for Radeon 8500.
NVIDIA, OTOH, does not release specs, but only proprietary drivers, so there's little if no hope of seeing open drivers for that hardware no matter how much money is thrown at it.
Personally, when I'm asked what video cards are best supported by XFree86, I usually reply "buy an ATI or a Matrox: both have a long tradition of releasing specifications to developers".
The same way advertising does.
I think this may be a rather inventive way of advertising. Sponsorship. It has "worked" for decades in sports, why not in open source?
What would be cute is if they had a "Brought to you by..." message during the splash screen for the driver.
Myth? hardly...
You are more then welcome to check with a percise measurement equipment the output of Matrox VGA out connector and NVidia's one..
Matrox leads on that, even compared to GeForce4
Hetz (Heunique)
Like any other company, they probably weighed the cost of replacing their 5000 SGI boxes with something else. Found that this video card produced a very good benefit at low cost (especially compared to the SGI) but had poor support.
Think about them now buying 5000 cheap AMD rackmounts to put in cabletv headends around the country. Shareholders are saving a fortune - I bet they save a mint compared to the cost of taking the proprietary route.
The fact that they're willing to release the fruits of their labour is just an added plus. They 'get it'.
On a related microsoft-bashing note - imagine deploying 5000 lights-out, 'rock solid' NT boxes around the country in cable-tv headends all over the country. This is actually a great case study in the reliability and manageability requirements that these guys have and how linux beat out nt and other unix-like environments.
Oh, sorry, I mean GNU/Linux haha
Interesting you should mention Doom III... If you go read the actual announcement post (why wasn't this linked in the story itself, btw?), you'll soon see the follow-up from a certain Timothee Besset... With a very, shall we say, relevant company name in his e-mail address. Heh. There's hope, at least.
I wonder how big the chances are that a driver for the 8500 generation of hardware is any use on the upcoming "R300" chip, though. It would be so annoying if, once there is a driver, it's one generation behind. *Sigh*.
main(O){10<putchar(4^--O?77-(15&5128 >>4*O):10)&&main(2+O);}
A few points to add to this excellent post...
The Weather Channel's current headend rackmount machines are based on the Silicon Graphics O2. SGI has announced the End Of Production of the O2 and O2+, they will no longer be made as of November 2002. There is no similar replacement. The O2 was/is one hell of a video box... it's performance and capabilites were great for NTSC/PAL resolution video... that's what the box was designed for. But alas, there really is no replacement... The closest thing is the Octane2, but it's almost 3x the size... 2-4x the cost, uses about 3x the power.... and the only video option available for Octane is very expensive as it does mutliple channels of HD and SD video. There really is no O2 replacement.
I'm sure the Weather Channel has bought a few extras (they've already bought thousands) but is seeking some newer/better/cheaper alternative. Because the station, it's workstations, headends, and its network is mostly Unix based (mostly SGI IRIX, some HP-UX), Linux makes sense... especially with some of SGI's opensource offerings (XFS, Failsafe, and others).
What's interesting is that The Weather Channel continues to buy big SGI Origin servers and wiz-bang workstations (Octane2, Fuel) for internal use. It's also interesting to note that they are still in the process of upgrading many of the older headends to the latest (O2 based) hardware and newer software revisions.
I don't think they'll be switching really soon... but it is very cool to see that it's going to be a friendly community effort.
>Why not use existing drivers for other video cards instead? (like NVIDIA)
Because the broadcast studio will expect this hardware to be in use for at least half a decade. They aren't like PC users -- I wouldn't expect them to be throwing out their hardware every couple of years.
So, what happens when NVidia does a 3dfx? It took less than that amount of time for 3dfx to go from hero to zero, so it certainly can happen again.
You guessed it, they have to buy all new hardware from a vendor that is intelligent enough to provide open source drivers if they expect to ever update X, had they gone with NVidia.
For a broadcast studio, this could mean millions, not to mention that the downtime will cost even more dearly.
With open source, and open hardware specs The Weather Channel can expect a smooth ride should they find new software incompatible with their existing drivers. All they have to do is update them themselves. They don't have to wait days or weeks for another company to fix it for them (assming they are still in business).
If you could be told what you can see or read, then it follows that you could be told what to say or think - BoC
Given that the current Linux support for the 8500 is limited to a binary-only driver that is intended for a related professional-level card, the delivery of an open driver is excellent news.
I think this part is a bit misleading. XFree86 4.2 has support for several ATI 8500 chips, and there are and have been Radeon drivers in the Linux kernel that work with it. From what I remember reading when I got my card working (yes, I'm running an ATI 8500 under Linux with no problems) all of the 2D support is there, and some limited 3D support.
I agree more open source work for this card is good news, but that sentence almost makes it sound like you can't use the 8500 well with Linux at this point.
It has already been done by Hans Reiser, and the OSS community didn't quite enjoy it...
...
The reiserfs sponsor banner shouted its sponsors at kernel startup, but it has been replaced by a discreet "Check www.reiserfs.org for sponsor information".
Here's the log:
Checking ReiserFS transaction log (device 03:02)
Replayed 1 transactions in 4 seconds
ReiserFS core development sponsored by SuSE Labs (suse.com)
Journaling sponsored by MP3.com
Item handlers sponsored by Ecila.com
ReiserFS version 3.5.18
There are two very different sides to modern weather modeling and display...
1) The acutal hardcore computation (done on huge, non-graphical SGI, HP, and Cray boxes with more GB of ram than I have harddrive space).
2) The on-air graphics boxes that make pretty pictures for television and website weather.
The first requires insane amounts of CPU, large caches, lots of ram, and gobs of thruput between CPUs. The second requires graphics hardware capable of generating television-resolution static images or a few frame of animation from pre-processed data.
Both areas are constantly growing, especially the back-end number crunching. On air graphics continue to become more complex, especially with 3D cloud displays that some weather reports show. But even the SGI O2, introduced in 1996 and ending production in 2002, is more than powerful enough for this task. The Weather Channel is working on putting together a solution to eventually replace their 5000 rackmount O2s located across the country in cable headends generating on-air graphics ("and now your local forecast"). They're also getting ready for HD televison resolutions... something the O2 cannot handle (the Octane can, but that'll cost ya $40K). As a side note, there really is no O2 replacement.... the O2 was a really nifty box for rendering OpenGL direct to NTSC/PAL video out without having to go thru hoops or do any software/hardware hackery.
Check this out, it's very interesting.
c al/index.html
u ne/navy_weather.html
http://www.sgi.com/features/2001/dec/fleet_numeri
http://www.sgi.com/features/2001/jun/weather/
http://www.sgi.com/newsroom/press_releases/2001/j
There are drivers, and then there are drivers that you actually want to play games with.
On account of ATI's generally/traditionally superior image quality, I'd get a Radeon over a Geforce if it weren't for that nvidia linux drivers actually take advantage of all the card's hardware features. My experience with owning an older Radeon has been that even the most concerted DRI efforts haven't put half of the Radeon's features into effect. Thus, I do not want to play games in linux with my radeon.
Is it possible that The Weather Channel will get the Radeon's gaming features working in full? If not, then this doesn't really change anything for Radeons in linux.
If you don't care about gaming, then this news isn't very notable because the 2D is already good. Otherwise, any incompleteness of 3D features will just perpetuate the Radeon's status as pointless for linux when geforce available.
On the other hand, it's true that nothing comes close to radeons in the mobile department (considering things like their low energy consumption on dvd accel). So I guess there definately are people compelled to buy a radeon despite superiority of geforce drivers, who must we welcoming of any and everything they can get in radeon support.
So what GUI component is there to the filesystem, really? I can see having a config applet for the graphics, but really having a control panel for ReiserFS seems to me less than pointless. I know I don't perticularly want an applet whose only purpose is to be an "about" window.
± 29 dB
But what about things such as mpeg decoding? Is that not still in the world of "2d"? If your a film buff or a DVD freak you sure as heck do not want an nvidia based card.
"Things are more moderner than before- bigger, and yet smaller- it's computers-- San Dimas High School football RULES!"
What? ATI 8500 (and 7500) do not have 10-bit alpha/overlay. Where would the extra bit's come from? 32bits per pixel, 24 bits for RGB, 8 bits for Alpha/overlay.
Perhaps what you mean is they are using a 10bit DAC for output (a claim going back to the Rage128 cards from ATI).
How about providing some proof. A link or something, I find nothing on a search on Google that mentions 10bit overlay, only the 10bit DAC.
On Maxtrox they've actually gone to 2bit Alpha/Overlay, and 10-bits for each R G B for more color depth.
You are more then welcome to check with a percise measurement equipment the output of Matrox VGA out connector and NVidia's one..
... I thought so.
Matrox leads on that, even compared to GeForce4
If it's not a myth, please provide proof.
Something besides the typical subjective web review (and even those are admiting Nvidia has caught up).
How about a review that uses "precise measurement equipment" as you describe.
(and as someone else said, please test against a quality card, not the cheapest clone card you can find).
Oh wait, you mean there isn't any such review?
Great, so when ATI ships their new stuffs, the R250 and R300 chips, later this summer, Linux users will still be waiting for this corporate handout driver for the "old" 8500.
I think that I could get used to Nvidia's way of doing things...
I might be wrong, but I remember recompiling the NVidia drivers from source. I don't remember looking at the source though, so it might just be a wrapper.
:-)
Kernels and modules have version numbers, and a module with the different version no. than the kernel will not load. Thus, to ensure that their binary-only drivers worked in any kernel 2.4.x (or whatever) NVidia probably gave you the object files and one source file to compile and link against your particular kernel. That's my theory anyway.
Higher Logics: where programming meets science.
Sorry to get involved in a big video card flamefest (and have no fears of my continued involvement as I won't be viewing this thread again), but I happened to remember seeing something just like this. And sorry that this is coming out of the mouths of Matrox indirectly through Anandtech, but this is the one I remembered.
To me, the most exciting thing about this is that it may be the first example of what I believe is a coming shift in the economics of software. Right now, software is treated as a product like almost any other manufactured product. Programmers are laborers and they produce product for consumption.
Free Software turns this around. Programmers become professionals more like engineers, architects, and doctors. They are paid for the quality of their practice, not as units of production in a manufacturing enterprise. One of the most common complaints about Open Source and Free Software is that it is anti-capital and that it will put programmers out of work (or at least out of pay). I think that is just plain wrong. Most programmers working today do not work for software companies. Most work in MIS making systems of hardware and software fit the needs of specific businesses.
This is the first case I know of where a company that is not at all in the software industry is paying programmers to develop software that they need that will directly benefit them AND anyone else who wants to use it. (Several companies like RedHat and Mandrake have done the same over the last few years, but they are, at least in some hybrid sense, in the software industry).
I think this will happen more and more. This is happening right now only because Free Software and Open Source software must be being used at the Weather Channel widely enough that they need these drivers. Once Free Software reaches critical mass (I know: It's an abused term and I'm abusing it right now) this will happen more and more. Eventually it will make almost all kinds of software available for "free" (and Free) and programmers will be paid well for doing it. Instead of 21st century robber barons amassing gigantic fortunes for herding programmers together, thousands of programmers around the world will make more money than they do now developing software for "Free."
Why? Bcause computers and software have no inherent value. Their value comes only in how they improve the efficiency of other processes, or enable processes that could not be done without them. They are tools. They are used to make and do other things. It is this economic "amplification" effect that makes them valuable. Sure, there is some value in the software economy, but the efficiency boom that gave us the longest post-war economic expansion without substantial inflation wasn't entirely based on Microsoft's profits (or even slightly based on them. As rich as Microsoft is, they are just a drop in the bucket of the economy). No, it was the way the technology tools improved productitivty throughout the economy. They whole software industry (and by that I mean people who develop code industrially and keep it closed, raising the price by creating an artificial shortage) could vanish and be replaced by free software and programmers and the economy as a whole would get richer.
How will programmers get paid? Like this case.
So a company has to pay $150,000 to get something developed. If all the other software they use is both Free and free (libre and free? Free speech and Free beer?), they may well end up spending less on software while programmers get that reduced amount of money with little or no corporate overhead.
It becomes a profession, not an industrial enterprise.
My bit of pie-in-the-sky thinking for today...
> What if they feel like using one of the *BSD's?
:-)
The are using one of the BSD's.
Look at all the FreeBSD commit emails with "Sponsored by: The Weather Channel".
It was put in a rather long message that gets displayed when mkreiserfs is done creating a filesystem.
Sounds like exactly the right place to put it. First time I saw it I got a bit of a start, as I usually suspect an error message when I see such amounts of text scrolling by, but it makes sense to put it there.
Mart
"I know I will be modded down for this": where's the option '-1, Asking for it'?
ATI finally releases specs for their hardware DVD acceleration, if not out in the open, at least to this project. Their hardware DVD acceleration is GREAT in Windows. You can actually run a DVD-ROM on a Pentium 233MMX if you want to using one of their video cards if you don't mind not using the computer for anything else while the DVD plays.
ATI is to be commended for their relative openness compared to NVidia and Matrox. I think they have a ways to go on being cooperative, however since there are only a finite number of video card manufacturers they should be encouraged.
ATI should also be encouraged to find better coders for their Windows driver products, but that's another story for another place and time entirely.
Knowledge is power. Knowledge shared is power multiplied.
ATI is to be commended for their relative openness compared to NVidia and Matrox.
While I agree that ATi's willingness to release specs is heartening, I don't think you give Matrox enough credit. They have a pretty good history of cooperating with open source developers. I think they keep some portions of their linux drivers closed (DualHead maybe?), but by and large they have been quite supportive of OSS.
Plug "weather channel open source" into google, check a few articles, find out how heavily TWC has been relying on open source for at least the last 2 years (like replacing several NT boxes with one Linux box), realise what they're really doing is looking out for number 1.
Open source has saved them a bundle, so if doing this for their own benefit has the side effect of helping others, I'm sure they still feel that they're getting the best part of the deal.
Hmm, let's see. Weather Channel provides detailed advance information and education on severe weather, which contributes to increase in public safety. How about a new public discussion. "Which saves more lives, open source or the MS way of doing things?"
I see even classic Slashdot is now pretty much unusable on dial up anymore.
OSS zealots will most likely mod me down for saying this, but most of it is the truth mixed with a little bit of my own opinion. ;-)
:-p ) but you are being slightly less than honest when you give out a blanket statement to people that ATI or Matrox is best supported. That person is asking you for information and, instead, you are giving them your opinion dressed up to look like fact. A more truthful way to answer that question would be to ask them what kind or hardware/OS they plan on using and then to tell them that NVIDIA is best supported for x86 Linux and that ATI Matrox are best supported for all other platforms. If you must push your own OSS agenda, then follow up the statement with an explanation of why you think people should support ATI and Matrox because of the way in which they support the OSS community.
:-)
It always bothers me a little when I hear people say stuff like this. The truth is, as has been stated by NVIDIA many times, that there are patents shared by NVIDIA with other companies that have been used in the development of they video cards. One of these companies is, most likely, SGI. They can't open source their drivers without violating confidentiality agreements and opening themselves to, justified, lawsuits.
Yes they only release binary drivers, and it would be much nicer if they would release the source as well but when it comes right down to it their drivers are vastly better than any other 3D drivers out there. Unlike all other Linux drivers for 3D cards NVIDIA's drivers are about equivalent in performance with their windows counterparts (this is as of the last time I saw a benchmark, I've heard rumors that they may be faster now). Granted, this doesn't help people on non-x86 Linux or on a *BSD (and no, I don't believe *BSD is dying.
Well, that's just my $0.02. I personally believe in OSS, though I don't really believe in the extremism of people like RMS. I also believe that there is a place in the world for non-OSS in certain places as well as understanding for companies that simply are incapable of supporting OSS for external reasons beyond their control (such as legal reasons). I don't mean for this to come off as a flame so much as a friendly rebuttal.
-GameMaster
Rules of Conduct:
#1 - The DM is always right.
#2 - If the DM is wrong, see rule #1
>What type of research has Redhat funded for the
>sake of Linux?
Well, they employ:
Owen Taylor - one of the maintainers of GTK+
and the author of Pango
Havoc Pennington - author of GConf, Metacity
Jonathon Blandford - maintains the Gnome Control
Center, contributed to GTK+
Elliot Lee - wrote ORBit, works on embedded
GTK+
Dave Mason (until a couple days ago) - GNOME
documentation
Steve Tweedie - ext3
Alan Cox - kernel hacker extrodinaire
Not to mention the projects they contribute to:
GCC - remember Cygnus? remember who bought them?
CygWin - see above
eCos - see above
EL/IX - see above
newlib
tux
insight
mauve
source-navigator
pirahna
Aside from that they do a *lot* of testing and bug fixing on everything in the distribution.
Matt
MPEG2 acceleration really just means that the hardware can scale the picture and convert from yuv to rgb in realtime. Directdraw has interfaces for these functions and they're utilised accordingly..
> Plug that superb $400 Matrox video card into your cheap-as-shit monitor or even worse, a LCD (even worse color).
If you have a digital-input (DVI) LCD screen, it sees only digital signals. No analog components on the vid card come into play. The analog circuitry in the LCD is all that matters. I haven't used an LCD screen long enough to comment on quality.
#define X(x,y) x##y
Peter Cordes ; e-mail: X(peter@cordes ,
The NVidia driver source is a few C source files, an object file called something.o (compiled binary), and some shared libraries (libGL.so, or something). The source files have some wrapper functions. They provide a consistent interface for the already-compiled code; They are the interface between NVidia's driver and the Linux kernel.
#define X(x,y) x##y
Peter Cordes ; e-mail: X(peter@cordes ,
> I would think that needing to insert some binary-only code into a kernel could significantly decrease their vendor's ability to support their machines.
That's correct. If you report a kernel bug that happened while you had NVidia's drivers loaded, most kernel hackers will tell you to go away.
#define X(x,y) x##y
Peter Cordes ; e-mail: X(peter@cordes ,
Despite what people think about the other parts of an ATI graphics card chipset, you have to admit that ATI almost single-handedly killed off the need for a dedicated decoder card in terms of DVD playback for computers with AGP ports. Their support for both hardware motion compensation (HWMC) and Inverse Discrete Cosine Transform (iDCT) starting with the Rage 128 chipset meant you saved some 35-40% of CPU cycles decoding DVD's compared to an all-software solution using WinDVD or PowerDVD.
And everyone else is jumping on the bandwagon, too. Note that the nVidia's GeForce4 MX and GeForce4 Ti chipsets and the new Matrox Parhelia chipset have at least HWMC and iDCT support for DVD decoding.
Raymond in Mountain View, CA
I don't know the hardware well enough to say what's "comparable" to the O2, but there's a good reason it's the last of its type -- it's a money loser. Like all of SGIs desktop systems, the O2 just couldn't compete with comodity systems. It makes sense for SGI to concentrate on massively-parallel systems, where there's less competition.