How To Request Better ATI Linux Support 192
An anonymous reader writes "Michael Larabel, the editor of Phoronix, has outlined some strategies for contacting ATI's customers (OEM/ODM/AIBs) to seek ATI Linux fglrx driver improvements. He opines that contacting ATI or AMD directly is the 'wrong approach.' He also states, 'I know for certain that at least one major OEM would like to see improved Linux support but is afraid that the Windows support would then be at risk.' Michael cites examples from the past where Lenovo had sought improved Linux display drivers, which resulted in several new features last year. He provides links to the feedback pages for a number of the vendors to whom ATI actually does listen."
Buy NVidia (Score:3, Insightful)
Not realy accurate statement. (Score:5, Informative)
They have completely and 100% open and free software drivers.
They are ahead of Nvidia when it comes to Linux desktop support. They will support sleep better, then will support hotplugging monitors better (when support for that sort of thing is added in X.org 7.3).
They supported technology like AIGLX before Nvidia.
They are quite fast enough for 3D desktop. The onboard GMA 950 can comfortably run either compiz or beryl 3d desktops with high efficiency.
As the display technology for Linux progresses the Intel onboard video cards will be the first.
Other advantages over Nvidia propriatory drivers include that they are much more inexpensive. The motherboards they come on have much better Linux support then the typical motherboards you find Nvidia onboard video drivers.
Laptops with Intel onboard video drivers will have advantages in price and battery life as well as stability when it comes to sleep and other advanced power management features.
The advantage that Nvidia video cards have over Intel is performance.
If you require performance for LInux desktop that goes beyond free software 3d games and good 3d desktop support and have requirements for newer video games or need 3d performance for your work then you have no choice but to buy nvidia.
There exists no open source 3d drivers that can support high end 3d performance nearly as well as what Nvidia provides.
But if your looking for cost effective and stable (much more stable then Nvidia) 2d/3d performance then Intel onboard video cards are the logical choice.
Plus they are open source.
Using Intel hardware I have absolutely no need for any propriatory software to drive my hardware. No SATA drivers, no video drivers, no wireless drivers, no nic card drivers, or no audio drivers need to be proprietory in any way.
(Intel is no freind of Free software, or realy even open source. They just see the financial advantage to supporting Linux properly.)
The current chipset for Linux to look for if you want as trouble free install as possible is the Intel 945g with the integrated GMA 950 video device. For non-bog-standard resolutions (ie widescreen) you will need to use the 915resolution hack for now, but this should go away in the future.
For special setups (for onboard devices) such as TV/componate/HD-out, DVI-out, and even dual DVI out you can purchase ADD2 cards for those features which plug into the PCI express port and interface the onboard intel cards. I don't know how well these work, but I am told by X.org folks that they _should_ work and will be _very_ interested if they don't.
Re:Not realy accurate statement. (Score:2, Informative)
Re:Not realy accurate statement. (Score:2)
Re:Not realy accurate statement. (Score:2)
However your points about opensource-ness are not refutable
I game with the Intel GMA 950.. (Score:4, Informative)
I am using Debian unstable, which currently uses the X.org 7.1 release.
The drivers that are supplied with that are not well optimized. They are good for compiz and can run Beryl very well, even with high amounts of eye candy except (almost everything turned on) for the one or two features that require special shading support. (basicly water effects).
It is capable of playing Quake2 and Quake3 well.
Return to Castle Wolfenstien is _very_ playable.
Enemy Territory is starting to push it and my currently favorite game, which is a full modification called True Combat: Elite is barely playable (it adds several more advanced features to the ET like HD lighting)
Nexuiz is not realy playable. Tremulous is fine, warsow is very playable. Cube/Cube2 is mostly playable except certain levels.
Benchmarks suggest that they offer decent enough performance for UT2k3 and UT2k4, but I don't know that for a fact.
If you use:
export INTEL_BATCH=1
and run 16bit RGB then everything is mostly playable. Also manually allocating memory to AGP and Texture stuff helps. And then allocating memory to a special buffer will enable HD-sized XV support. (see the 810 man file) A little tweaking is very helpfull. Expect a 75% boost in performance from that alone.
Now X.org, DRI, and Mesa folks (the DRI drivers are basicly made by taking Mesa and accelerating what they can), are working on efficient ways to manage video card memory, which is required for newer cards and usage patterns.
They are working on a special branch of DRI drivers called 915tex_dri.so (were the normal is just 915_dri.so). This adds lots of optimizations and efficient dynamic memory management and allocation. Using this you get quite a performance boost over the default drivers.
ET is very playable. As is True Combat: Elite and you rarely get framerates that drop below 20FPS. I keep my framerates limited at 70FPS and most of the time it's sitting at that limit.
If your using LCD display there is no point beyond having the limit set at 60 FPS. But you need to have good performance to keep it dropping down under 30FPS for good online play.
So the drivers are definately improving. I am expecting good things to happen with X.org 7.3. But due to the shared memory sceme and lack of accelerated texture and lighting effects these things will never be usefull for gaming, not like most 'gamers' expect.
Now with the GMA X3000 aviable with G965 video cards they will offer acceptable gaming performance once the drivers are optimized. They offer hardware acceleration for texture and lighting effects as well as shader support and other such things.
Technology-wise, at least on paper, they are on par with ATI's and Nvidia's entry-level video cards.
You can find benchmarks on Phoronix for the GMA 3000, which is from the Q965, which is a bit lower end then the X3000. It completely lacks T&L hardware acceleration and other such features. So it's sort of like a X3000 core, but with the GMA 950 features.
The X3000 should perform better then that, by quite a bit, but I don't think right now the drivers are realy all that optimized. Not until the memory management stuff gets worked out. Then it should meet the lower end requirements for Doom3 and Quake4 pretty well. At least enough to be playable.
But realy if your a 'gamer' that is more then casual then Nvidia is about it.
The nice thing about this is that you can get a 945G motherboard right now, get good 3d/2d support and if it doesn't work out for you then a Nvidia card is a easy add on.
Probably with Feisty the G965-based motherboards will probably be a good choice, but unfortunately I don't own one right now for personal testing. If Ubuntu was smart they'd be paying close attention to those chipsets, especially since they will be in the majority of next-generation laptops that people will be trying to use Ubuntu with.
Re:I game with the Intel GMA 950.. (Score:2)
Agreed! I'm not much of a gamer, but sometimes I fire up Tuxracer which runs very smoothly on my Intel 855/852 laptop. Beryl also runs fine.
It seems that many people have some kind of religion against integrated graphics, and they insist on buying external monster graphics cards no matter what kind of use. These are the same folks that insist on dedicated graphics memory without looking at any real world performance, even though things like AGP were specifically designed for fast access of motherboard RAM.
Then again, I also value open source and I like getting the necessary hardware drivers with the vanilla Linux kernel.
Re:Not realy accurate statement. (Score:2, Informative)
Intel graphics cards, to the best of my knowledge, only come on Intel motherboards, which are only compatible with Intel CPUs. As AMD and ATI have merged, this means that buying Intel graphics cards causes ATI/AMD to lose out on three sales, not just the graphics card.
Re:Not realy accurate statement. (Score:2, Funny)
The i915 is surprisingly good (Score:2)
In linux anyhow (in windows the freakin thing won't even show my accelerated RSS screensavers properly) it actually behaved better than a supposedly higher-end ATI card when using Beryl, and was in fact a fair bit more convenient to setup than even my NVidia card.
Re:Not realy accurate statement. (Score:4, Interesting)
My complaint is that that is done entirely in Mesa. The GMA 9x0 is rasterization hardware - that's all. I'm not surprised GMA 9x0 drivers work so well in Linux, and, in fact, I wouldn't be at all surprised if they work better than in Windows. I'm not surprised they're open source, either.
Consider, though, that they were able to open the source because all they had to do was rip out all of their 3D IP and let Mesa do all the legwork. Looking at it that way, it's not at all very interesting that they opened the source - after removing all the 3D code, all you've got is a really fast 2D driver - so what's there for them to hide?
NVIDIA has said on multiple occasions that the only thing keeping them from opening their source is all of the licensed 3D code they use in the driver. Until that can be stripped out, it stays closed.
ATi has only gone so far as to contract a developer and give him specs under NDA to develop an open source driver - for their outdated and obsolete hardware.
I won't be happy until an actual 3D chip has open source drivers - or perhaps until the GMA (X)3000 gets put in laptops. Until that happens, I couldn't give a rat's ass that Intel opened their drivers for their worthless hardware.
Re:Not realy accurate statement. (Score:2)
So it's barely better than the old 3DFX Voodoo I/II? IIRC, they were just texturing hardware. The CPU/driver had to rasterize the polygons, passing scan-lines off to the card to be drawn. That's why the original SLI setup was so easy - just pass odd-numbered lines to the first card, even-numbered to the second.
Not at all. (Score:2)
Completly false.
No data went between the Voodoo and the grafic card (proof is : if you didn't use special software that was aware of the Voodoo, when taking screenshots by dumping the VGA memory you got plain gray screens).
If it worked like you suggest (transfering the data and displaying them from the VGA board) it would be a waste of bandwitdh and you wouldn't have to connect the Voodoo in serie between the VGA and the monitor.
The rare circumstance where you can see data exchange between the VGA and the voodoo is with some software hacks that enables 3D-in-a-window (in this case the Voodoo renders the whole screen, but instead of sending the result to the monitor, the result is sent to a special buffer in the VGA board which in turn displays it inside an overlay window - like video : except that instead of decoding frames in software or with MPEG accelerator, you render 3D frames in hardware and give them to the overlay).
Voodoo I & II had complete polygon engine with 1 or 2 texturing chip(s) respectively (out of a supported maximum of 2, resp. 3).
It can completly function by itself, and on linux, you can find drivers to use them alone - X server running on Voodoo.
The reason you had to couple them with regular video cards was :
- They lacked basic VGA function (text modes, paletted modes, emulation of legac CGA/EGA hardware, etc.)
- Some 2D functions missing (no overlays for video, much less hardware scrolling and screen splitting than VGA)
- Other 2D functions not as good (2D blitting is very limited).
- no port-level compatibility with VGA (dos software can't run unmodified on Voodoo boards)
- No BIOS : PC can't boot on Voodoo boards (either the mother board has to be set to ignore missing video cards, or has to be set to use serial-port terminal as console. No way to use console on Voodoo)
For all these reasons you had to combine a VGA with your Voodoo board.
But the Voodoo board it-self is a complete (3D only) video card with all necessary hardware to render polygons. What it lacks compared to modern 3D card is geometric hardware (T&L), but has everything needed to render polygons and display them.
GMA950 has 3d hardware. (Score:2)
No, it isn't. If it was, it would mean that everything else would be done in software and all 3d graphics will run slow. At "slide-show" speeds. Which isn't the case.
3D acceleration is a complex task. On one hand you have hardware that is pretty low level (can draw polygons given that you give on-screen coordinate, pointer in memor for texture etc.). On other hand you have APIs that are more high level like OpenGL (here is a scene composed of polygons, render them).
You always need some kind of middle-ware between them that will get high level command from the API and translate them into a series of parameters to render each triangle, replacing "material id" with pointer indicating position in video memory where texture have been previously loaded, etc.
MESA is, in the beginning a complete OpenGL software implementation. Mesa Drivers are, special version of Mesa where everything possible has been replaced with hardware equivalent : ie. instead of drawing polygons in software, a 915 Mesa drivers will produce sequences of instruction to make the 915 render them. On some architecture that supports them, 3D geometry calculation will be offloaded to specialised hardware like T&L units or Vertex Shaders (some onboard GFX controllers and some older GFX cards lack them, in which case the Mesa driver is based on original Mesa code and geometry is offloaded to coprocessors like 3DMax or SSE, or done in software).
So of course, everything on linux uses Mesa. But that doesn't mean that everything on linux is done in software. That just means that Mesa does the front-end core work to give a standart OpenGL API, and the rendering of the triangles themselves is done by the hardware.
What Intel have done is the exact contrary to what you said : they have provided a 2D driver, and just bare bone 3D informations / code, that can draw triangles in hardware. Mesa is the glue in the middle that translate hi level OpenGL api to triangle rendering instructions for the 3D driver.
In case of Nvidia : 3D code *is striped*. all you have that remains is a pure 2D opensource "nv" driver. If you want to do free 3D with an nvidia board, you can only do it by letting Mesa do the whole legwork. Everything. Ie: even the polygons are done in software by the "full" mesa. No mesa drivers with specific code to do polygons/geometry in hardware.
As anybody else, they have to use a middleware between the API and the hardware. Their problem is, this is a huge monolithic pile of entangled code that includes lot of code that they didn't wrote themselves, but bought from other companies. It's either no 3D (current open-source variants) or a huge mass that contains all the hck they imported from 3rd parties to make 3D work faster. They can't open source that. That's understandable.
BUT :
- They refuse to go the Intel way and give just a small piece of code that does only the talking with the hardware, so that a Mesa driver can be written.
- They even refuse to release documentation about their hardware so a driver can be written from scratch.
The only way to get open-source drivers, is by reverse engineering, and writing a driver only using guess work and experimentation : the "nouveau" project. It's not necessarily impossible, in fact the "r300" project has done the same for Radeon 9500 to X850 cards. But ATI and nVidia aren't helping a lot. And that's inacceptable.
Re:GMA950 has 3d hardware. (Score:2)
Intel states in their documentation that the GMA9x0 is absent a 3D T&L engine. They added one in the GMA 3000, making it the first card that isn't simply a rasterizer. The GMA 9x0 have pixel shaders, which, I assume, is used to handle texture processing and other basic operations, but all of the other work - Matrix transformations, perspective, texture pre/post-processing/mip-maps, and so on, are all done in software, whereas the Voodoo II, as I understand, did the majority of that work in hardware.
That's what I'm complaining about, and that's why I hate the GMA 9x0.
Re:Not realy accurate statement. (Score:2)
My team's requirements from IT were that these machines must support two monitors and must run Linux. Obviously without X11 the first fails. The integrated graphics only support a single monitor, so the second requirement also fails.
Adding an nVidia card, which is what I would do for two monitors, turns out to be impossible since it only has a PCI-e 1x slot and the only cards out there are ATI or Matrox. I spent several days trying to get the ATI driver to install, and it's still quite buggy. The Radeon X1300 framebuffer driver was unusable since it was so slow. The dual monitor support from this card also sucks, with one of the outputs having poor impedance matching causing all sorts of ringing on the image.
Re:Not realy accurate statement. (Score:2)
00:02.0 Display controller: Intel Corporation 82915G/GV/910GL Integrated Graphics Controller (rev 04)
It *should* work, but it doesn't, at least with OpenSuSE 10.2. I suspect it might be the BIOS or something on this HP machine. In any event, the onboard graphics are useless to me since I need two monitors.
Re:Oh, and stay the fuck away from ATI. (Score:3, Informative)
nVidia used to be a tech support nightmare back in '97. Gateway 2000 used them on a lot of their system and the drivers were always screwing up. I think I did more nVidia driver reinstalls than any other single thing. They got serious about quality of their drivers and now they are some of the best. ATI could take this a step further by open sourcing their drivers (at least the Linux ones.)
I'd even bet that if ATI open sourced their Linux drivers, they'd be better than the Windows ones within 2 years, and within 4, they would be the basis of the Windows drivers, which would probably then go open source as well. (These are pessimistic numbers, because I'm not much of a betting man.)
So if I'm an nVidia fan, why am I saying that I hope ATI will do this? Simple: ATI has always had better hardware. All it would take for me to change loyalty would be to have ATI open source their drivers. I'd go buy one of their best cards tomorrow if they open sourced today. (I tend to be an early adopter. Go Feisty Fawn!)
Re:Oh, and stay the fuck away from ATI. (Score:2)
Re:Oh, and stay the fuck away from ATI. (Score:5, Informative)
You won't have absolute open source and you need to get used to that idea. You will always have a mishmash of open source and closed. What's important is that the OS and the underlying major technologies be open source. Drivers and applications do not need to be Open Source, and rightly shouldn't always be. You can see this by looking at gaming. No gaming developer is going to release their game into open source upon launch. It may be released 10 years later. There's absolutely no need to and should never be a pre-requisite for running on linux. The Kernel to the OS or the OS itself is licensed this way but that doesn't give zealots in the community the right to demand everything be open sourced or be no good.
Just accept that. It is important to bring in commercial ventures and you won't do it with the pure open source ideology. In fact, that would be killer to any attempt to bring in companies such as game developers. The OS yes, quality productivity apps yes, utiliites yes, drivers yes, games and other such products such as photoshop NO (and don't even consider it).
Opensource ATI exists (Score:2)
No, opensource drivers for ATI *already* exist.
To build drivers for Linux, you don't need the complete driver stack from ATI - You don't need their Direct3D driver nor their complete OpenGL implementation. All you need is just enough information to tell the hardware to draw a polygon on the screen. For the rest you can substitute parts of Mesa - so they don't have to reveal all secrets that their drivers do to increase speed.
In fact this has already happened.
Up to R200 (until Radeon 9250), documentation was available from ATI. They didn't reveal what their drivers did do under the hood. They only gave enough information about the hardware so you can instruct it to draw polygons. For the rest, code from Mesa was used.
Then the situation degraded : ATI stopped providing information.
For the R300 (Radeon 9500 to X850) reverse engineering has been done to understand how to make the hardware draw. Once again Mesa glued the rest together.
To have an open-source driver under Linux, ATI doesn't have to reveal the code they have licensed. They don't have to reveal any code. Everything the need to give is just enough information about how to use the hardware to draw stuff.
If they don't provide that, programmers have to reverse engineer : like "nouveau" project is doing for nVidia or "r300" did for previous generation ATI.
Gaming isn't the only application of 3D. There are people doing research in many fields such as medical imaging as an exemple. They need 3D too. And they need to do 3D on hardware that isn't necessarily supported by the official hardware makers. To run 3D on exotic hardware, they need open-source, so code can be ported or adapted to their needs. ATI is only interested in supporting their main market : Kids like you running games on Windows or Linux on a x86 platform. It's not economically feasible for them to support any possible hardware. So people must be able to adapt drivers them self, but without source this is impossible.
Sorry, but there are pretty good reasons to have open source drivers :
- Portability. As I said before, if you want to use hardware on non vendor-approved platform you need opensource. In fact, the r300 project started because Apple laptops where using Radeon 9600 board, but ATI supported only Intel x86 hardware. No way to get 3D acceleration on PowerPC (or back then, neither X86-64 for that matters). To be able to adapt drivers for PowerPC, you need access to the source, which wasn't available, so they built their drivers from scratch.
- Security/Stability. Part of the drivers runs in kernel mode. Which is dangerously high access. The rest of the opensource world (Linux kernel, OpenBSD, etc.) counts on source availability and public scrutinity to find and correct bugs (the "given enough eyes, all bugs get shallow" linus law). But with BLOBs, you can't control the quality of the code. You're left to hope that the card makers pay enough attention (which isn't necessarily the case, as their main and most lucrative market are the Windows users). Such critical bugs have already be found in some wireless cards' drivers.
- Support. As hardware gets older, company tends to abandon su
RTFP! (Score:2, Insightful)
There exists no open source 3d drivers that can support high end 3d performance nearly as well as what Nvidia provides.
But if your looking for cost effective and stable (much more stable then Nvidia) 2d/3d performance then Intel onboard video cards are the logical choice.
He concedes the performance point, but merely says Intel is fast enough for desktop use. I don't play doom, so I really could care less if my Intel video is fast enough to play it.
It is fast enough in the same way that my 92 Accord is fast enough to get me to work and back every day. It isn't as fast as a (insert high performance sports car here), but it gets me where I am going reliably and quickly enough (faster than one can legally go in my jurisdiction).
Re:Not realy accurate statement. (Score:2)
Re:Not realy accurate statement. (Score:2)
The GMA 9x0 is rasterization hardware. The pixel shaders are the only piece of actual computational hardware on the entire chip. The GMA (X)3000 is the first ever chip that does not do all of its 3D graphics work in software.
Re:Not realy accurate statement. (Score:2)
Having a 945GM (on a new laptop which my parents gave me last Christmas) doesn't suck compared to an ATI card from 2002-ish... why?
I don't care whether it's a driver issue or a hardware issue. Don't consider it if you care at all about graphics, esp. 3D.
Re:Not realy accurate statement. (Score:2)
Intel's own specification documents. The GMA 950 lacks 3D T&L. The GMA X3000 adds a hardware T&L engine. Look at the page you linked to.
I didn't intend to imply that it's not a full 3D rendering pipeline. I said it's a rasterizer. It handles texturing and that sort of thing, but all of the other essential computation is done outside the hardware - 3D transform and lighting is done primarily in software, and is supported by the extremely fast 2D rasterizer with its basic 3D extensions.
Re:Buy NVidia (Score:5, Insightful)
I basically have to agree. nVidia drivers now seem to work well, after much pain over the years. They even managed to open up their development process up a tiny little bit by leaking beta drivers. Kind of like a prude who secretly has a little on the side anyway.
Nonetheless, it behooves us to fight the good fight and annoy the shit out of graphics card manufacturers to reveal the eleven secret herbs and spices so we can fully enjoy the hardware we paid for. Discriminating against crappy support is a step.
Re:Buy NVidia (Score:5, Informative)
Re:Buy NVidia (Score:2)
They've gotten a _lot_ better (Score:2)
Re:Buy NVidia (Score:2)
I don't see what the problem is, why every comment for this article is negative. Maybe everyone's bad experience is due to the open-source drivers instead of the ones ATI releases...in that case it's not entirely ATI's fault the open-source drivers suck. I'm not going to carry a torch for OSS, I'll use what works.
Re:Buy NVidia (Score:2)
Re:Buy NVidia (Score:5, Insightful)
I really don't want DRIVERS from ATI or NVidia though, I want DOCS. We hear all sorts of whining about proprietary secret data, blah blah blah, but I DON'T CARE. If you want to sell me a fucking device, release the damn docs already. This goes for all the winmodems, winprinters, winscanners, wincameras, winwifi cards, winethernet, winsound cards, winkeyboards, winmice, winharddrive controllers too!!! I'm not buying WINDOWS hardware, I'm buying COMPUTER hardware. I want to be able to use it with ANY operating system. There is nothing so damn secret about how to program your device that would put you at a competitive disadvantage if everyone is releasing information. And if there really is? Tough shit. That is the price of being in the business.
Re:Buy NVidia (Score:2, Insightful)
Re:Buy NVidia (Score:5, Insightful)
Re:I would mod you up. (Score:4, Insightful)
Re:I would mod you up. (Score:2)
And what I was referring to was exactly this, it is like me going to the comp.os.linux.advocacy newsgroup and whine that Windows is shit and it does not do X and it is crap and whatever, knowing that people there will OF COURSE agree with me. That is just plain stupid. And yet, you are caught into his game when you "defend" him.
Re:I would mod you up. (Score:4, Interesting)
It's up to the device manufacturers to work with their vendors that they work with to ensure that documentation for subcomponents is available if needed in order to write a driver. They should make the subcomponent vendor aware that they can't use their parts unless this information is made available to third party developers. The VALUE of the part is less without documentation as it makes the hardware less compatible.
The fact is, in days gone past we USED to be able to get all the technical docs we wanted. IBM even used to release the source code to their BIOS. I've already debunked the "trade secret" red herring in another post.
Re:I would mod you up. (Score:2, Informative)
But... first of all, we're talking about documentation, not code. Second, even if we were talking about code, I'm not sure what you mean when you say that manufacturers "need to protect certain parts of their code" (trade secrets at least shouldn't really apply, since we'd only be talking about the code for a driver, not the VRML sources for the chips themselves). And third, while there might be NDAs involved even when we're only talking about documentation, I think that asking for documentation can still be a good thing in that case since it shows that there is demand for open documentation and that entering NDAs might therefore be detrimental.
Put another way, as a user, you've got something to gain and nothing to lose, but I can tell you right away that you won't win if you don't try. I don't have a crystal ball, and I can't tell you what the best course of action would be (FWIW, I don't actually believe that there is a single course of action that's *always* best), but that, too, shouldn't be a reason not to *try*.
As for "his game"... I'm not sure what you're referring to there, really, unless you mean that it's pointless to post on Slashdot saying that open documentation is needed/good since it amounts to preaching to the choire. That's certainly true, but I don't think you can deduce from that that discussion of these matters is pointless; outside of that fact that Slashdot is not as homogenous as one might believe, anyway (there is a certain majority opinion in many cases, but there's also a considerable "long tail"), the fact that you participate in this discussion also proves that you do not consider it a priori pointless to discuss these matters.
That being said, I apologise if I misunderstood what you were trying to say.
Re:I would mod you up. (Score:2)
Your post is kind of funny in that you use the exact thing you're complaining about to complain about the very same thing.
Re:I would mod you up. (Score:2)
Uh?
That was a difficult one to dissect.
But the difference is that I am not writing about it on the, say, "Closed source for profit greed group mailing list". Do you see the difference?
Re:Buy NVidia (Score:2, Insightful)
back to my point; yes, i'd love open drivers for nvidia. it'd be fair to say they'd most likely see a surge in sales if they went down this path. but that's not their business model so it's not going to happen anytime soon. on the other hand, they sell a solid PRODUCT. a product isn't just the hardware but the software also and, in my experience, they're worth the money i've forked out for them. i know other people's mileage varies a lot but their prodcuts 'just plain work' for me. that's what keeps me buying.
Re:Buy NVidia (Score:2)
Only for you, most of us dont give a crap as it works fine with the proprietary drivers they ship. In fact, most people are windows users so have never even been exposed to this debate.
Personally I run gentoo and the proprietary nvidia driver is fine, if I ever have a problem I might start complaining (to nvidia), but in the 3 years I have bought nvidia cards they have never given me any issues.
Re:Buy NVidia (Score:2)
So fglrx I had last tried caused occasional hangs on VT switches, hung on gdmflexiserver, and did nothing for sleep so the ATI chip was chugging down 5 watts of power in ACPI suspend. On resume the graphics would occasionally be corrupted. Just everything was wrong.
So I tried them again (had no choice), with 8.34.8. Now sleep the whole laptop uses 400 mW without me having to bend over bacwards, gdmflexiserver has yet to crash, I have not seen video corruption (this is Feisty after I set SAVE_VBE_STATE=false and POST_VIDEO=false, methods to work around some cards, but in this case essentially letting fglrx take care of it all, like I do for an nVidia system).
Maybe it's because of Lenovo demands, but my recent ATI driver has actually been on par with my nVidia driver experiences now. With the *very* notable exception of a complete lack of AIGLX, or even more basic Composite support. Feature wise they are still missing those, but it seems they have demonstrated the *capacity* for a decent driver at least...
Re:Buy NVidia (Score:2)
My issue is performance. Performance-wise, a top of the ATI card performs more like an entry level Nvidia card, both using proprietary drivers.
This sucks. I've waited through several generations of ATI cards (bought a 9800 Pro, replaced it with a Geforce FX 5900, now using an X1600 on my MacBook Pro), and Linux performance still is an order of magnitude worse than on Windows.
Nvidia cards are, by and large, performance competitive on Windows or Linux. I cannot really play World of Warcraft on Linux with an ATI, but everything works beautifully in Windows and/or OS X. This is unfortunate.
Ohh get off the horse! (Score:2)
Could we please do away with the ATI drivers doesn't work on linux mantra?
Yes you WILL have problems with the fglrx drivers and older boards, but so fucking what? Those boards are 4+ years old, buy a new one!
Re:Ohh get off the horse! (Score:2)
I'd actually be quite happy to buy a new graphic card to get better driver support if there was one that met my needs, but all the new cards seem to be intended for gaming and require active cooling which I can't tolerate in my quiet system. I also need an AGP based card and am not about to buy/build a new system just to get a new graphics card - my current 2.8GHz P4 (Northwood core) system is still plenty fast for my needs.
Re:Buy NVidia (Score:2)
Re:Buy NVidia (Score:2)
If you want to avoid this hassle entirely, just add "exclude=kernel" to fedora-core.repo. You'll then keep the same working kernel and driver. You might miss a security update, of course, but for single-user machines safely behind a firewall I suspect the danger is small.
I've actually had much bigger problems with Intel's supposedly open-sourced 945GM video under FC6. I can't seem to make the "intel" driver work on my daughter's Dell 640m laptop with a widescreen display no matter what I try. For the moment I'm stuck using vesa which is slower than molasses. I didn't have the same problems with the i810 series.
Mail in your proof of purchase (Score:5, Insightful)
Get yogether with your buddies and collect a pile of ATI and competitor proofs of purchase.
With the ATI ones say that you are a customer and would really like to see Linux support. With all the competitor ones, say that you would have bought ATI but for the driver issue. make sure you youtube it, blog it,...
I doubt that it would work (Score:5, Insightful)
I know, we all like to think that the customer is king, and that just because you have a proof of purchase for an old $30 graphics card, it means that a major corporation must bend over backwards for you and catter to your every whim. They should instantly hire a big team to code whatever you fancy today, open-source all their programs... Why, they should even come over and do your laundry. Dream on.
When cattering to mass-markets, you have to think in terms of ROI. If it costs X dollars to do something, will you even get those X dollars back? Is it likely that you'll even make a profit? If not, it's actually smarter to ignore that market segment.
Drivers nowadays are complex and expensive things, and frankly the Linux hobbyist market is tiny. And then they're likely to buy the lowest end card, or not even that as they're busy bitching about how binary drivers are evil.
So, basically, fully expect someone at ATI to at most have a chuckle as they dump your letter into the garbage bin.
OEM's are a whole other affair, because they move millions of boxes. If one of those says "we need linux drivers", then:
A. they probably know what their many corporate customers want. Dunno, maybe some major corporation or government department decided to standardize their desktop on Linux and actually needs 3D accelerated drivers. Basically if a big OEM bitches, they probably aren't doing it out of zealotry and fanboyism, but they know something about demand that you don't. You listen and take notes when those guys speak. And,
B. even if not, you want to listen to those anyway, because they're the guys who make your money. They're the "R" in "ROI". The last thing you want is Dell or IBM (Lenovo) standardizing exclusively on nVidia cards because you told them to fuck off when they complained that lack of Linux drivers hamstrings their server sales. If that were to happen, you'll see a big dip on your income chart, and the mere rumour would make your shares dive and the shareholders demand blood and rolling heads.
Basically you'll have a chance with your proofs of purchase when you fit at least one of the two criteria, preferrably both.
Re:I doubt that it would work (Score:2)
If that was true, NVIDIA & ATI wouldn't even have written their binary drivers. But they did, and they maintain them, and Intel even wrote an open-source one ! What you're missing is that Linux isn't easily segmented into markets. The "hobbyists" are the ones fixing the bugs for the pro market, and the pro market just wants something working and maintained, so keeping the hobbyists happy is important. The natural tendency towards free drivers is slow, but that will happen.
Re:I doubt that it would work (Score:2)
The way I heard that story, ATI made the binary drivers at all, again, because a large OEM told them something to the effect of "we need Linux drivers to sell Linux servers." I.e., again, the loud mouthed hobbysts had nothing whatsoever to do with it, and the whole action was driven by OEMs.
No offense, but dream on. The guys who take the decisions for the pro market usually only care about numbers, and, if you'll pardon the cuss, don't give a flying fuck about whether some of us whiny nerds are happy or not. I'll dare say that 99% of what's being bought and sold has nothing to do with whether some nerd was getting a boner about it, but with whether the nice salesman promised the moon and played a mean game of golf. If nerds had anything to do with it, we wouldn't have a thriving buzzword-driven snake-oil market in the first place.
I'm a nerd myself, so don't take it as an insult, but the world _doesn't_ revolve around us. You may have imagined that whole economies revolve around listening to sages like yourself, but they don't. They revolve around who did the CEO play golf with, and what ludicrious savings/ROI/TCO/whatever and buzzwords the nice salesman in a suit promised.
Even when a corporation adopted Linux, it had nothing to do with listening to the local nerds ranting and raving, but with some salesman from, say, IBM coming and saying something like, "see, if you bought our big Power 5 computers running Linux, you'd get sooo much more performance and stability out of the WebSphere we sold you last month." While we nerds were just ranting and raving about boycotting the evil MS, the nice salesman came and said magic words like "TCO", "ROI", "synergy", "support contract", and the like. _That_ is what changed management's mind.
_Some_ companies listen to their nerds, yes, but they tend to be small, and far and few in between. While they do somewhat enlarge the market share that depends on the opinion of those disgruntled hobbysts, it's still a tiny market share.
Re:I doubt that it would work (Score:2)
Except, it is those 1000 people (I imagine this figure is a very low estimation) who the others ask for advice on what to buy.
Re:I doubt that it would work (Score:2)
Case in point, my 3 most recent 3D card purchases have been: 7950GX2 (PCIe), 7600GT (PCIe), and 6600GT (AGP, some factory-overclocked job I got on ebay for $120 before my last motherboard upgrade.) I exclusively run Linux at home (with the exception of the FreeBSD server handling shell accounts and DNS, and the macbook that belongs to the guy who sleeps on my couch while he's looking for a house.)
A good friend of mine owns an 8800GTS (and is talking about buying a second one.) Before that, he had a 7800GT. Again, No Windows. He's also got a macbook, but doesn't do anything remotely 3D on it beyond expose and quartz extreme.
Of the 10-odd people I know who are either exclusively or predominantly Linux users on their personal machines, at least 8 of them own a mid-high or high-end graphics card. Most of us also subscribe to Cedega for some reason or another (at least five of that 8-person sample uses it predominantly to play warcrack, thankfully i'm not included in that five.) In my case, it's because I bought it so that I could finish Morrowind (back in 2002 when I owned an AGP GF4 4400) after deciding that Windows 2000 wasn't worth the pain and suffering that it was inflicting. Since then, I've played one or two Sid Meier games, Oblivion, the two C&C: Generals games, and a stack of others.
Re:Mail in your proof of purchase (Score:2)
I did this several times, never even got a response. I also had a ATI Video USB device, tried getting them to release the details needed to program it... no answer...
So I now buy other products and will not buy a ATI based product. If AMD management "fixes" ATI management, I might reconsider but until they are friendly to more than Microsoft they are off of my purchase list. ATI and Broadcom are on my no-buy lists.
So wich modern graphics card IS fully opensource (Score:5, Interesting)
Is there an actuall graphics card out there that IS capable of doing the eyecandy stuff, it don't have to do games, that is fully opensource with absolutely no binary bits.
I used to think matrox cards were the way to go but even they have a binary HAL bit that you need if you want the more advanced features needed for xgl and the likes.
Anyway the matrox cards are not supported anyway, as they are listed as missing certain features that are required.
The only lists I ever find mention ONLY nVidia and Ati cards. Yet I have seen some references that Intel was working on opensource drivers for its cards or at least hired some developers to do so.
So, is there a graphics card out there that I can use that is simply fully opensource, no hidden tricks, that is capable enough to give me the candy?
Because that would I think send the strongest message of all, if everyone who runs linux just buys a fully opensource card the others would be sure to take notice.
Intels onboard stuff is, I think. (Score:4, Informative)
Intels i810 and above are. Of course you can't get any graphics cards with them, since they're onboard solutions, so you're stuck with an Intel processor too. Which may or may not be a drawback.
Re:Intels onboard stuff is, I think. (Score:2)
Obviously I'm not saying the current chips are the quickest OpenGL graphics around, but really, for 3D graphics, how many choices do we have?
* NVidia - works well, but binary only.
* ATI - works mostly, but binary only.
* Intel - works for me, open source.
* Xorg - fairly limited support for 3D acceleration.
* Commercial X-servers - either don't support current cards, or too spendy for personal use, closed source.
I'm happy enough with the ATI on my laptop, but I'm equally happy with the open source Intel on my desktop.
Re:Intels onboard stuff is, I think. (Score:2)
Even though the newest Intel onboard graphic chips actually support dual screen, the motherboards often have only a single video output connector. And it quite often still is VGA.
For example, nearly all Dell systems come with an Intel onboard video chip and have only a single VGA output. When you opt for DVI or dual-monitor, you get an add-in videocard which usually is something like an ATI X300
(it used to be nVidia FX5200 in the past)
There are a few systems with DVI-I outputs and often they have wired the analog and digital outputs to two different screens, so a splitter cable enables you to connect two monitors (with different content).
However, that is still one digital and one analog.
Re:Intels onboard stuff is, I think. (Score:2)
Re:So wich modern graphics card IS fully opensourc (Score:4, Informative)
With older hardware there were more options. I have an old Ati Radeon 9200 in my closet, just in case I need AGP graphics card. It's the last Radeon that works completely with open source drivers. (also 7000 and 8500 and 9000 work).
So it doesn't look too good, does it..
Well there is hope. Intel is working on discrete graphics chips. read more here [beyond3d.com] and here. [theinquirer.net]
I believe Intel has no reason to change their Linux friendly policy. So I hope they come up with a decent discrete graphics card and release open source drivers with it.
Since Intel is such a big player it just might encourage others to do the same.
Re:So wich modern graphics card IS fully opensourc (Score:2)
On the one hand, it's probably clear that the CPU groups of both Intel and AMD understand Linux, Open Source, and have enough of the "got religion" involved to behave in the right way. Since the graphics group is growing up inside a CPU company, the don't have the ATI/nVidia "closed religion," or at least not in a big enough way to wag the CPU-oriented dog. In this light, it's pretty obvious that SOMEONE inside Intel knows that if/when they bring out a decent open-source capable graphics card, it's going to sweep its way through Linux purchases. I for one would be ready to put my money where my mouth is, though it all depends on my purchasing schedule.
On the other hand, even recognizing the value of an open source graphics solution and its appeal to the Linux crowd, there are no doubt forces inside Intel that would like to see it kept tied to Intel CPUs and chipsets.
The real question is whether Intel would rather see an Intel graphics solution working with an AMD CPU, or gamble that they can grab the whole kit'n'kaboodle by keeping things tied.
As an aside, it's also pretty clear that though AMD may understand Open Source on the CPU side, that understanding hasn't jumped the fence to the ATI side, yet.
*********************
The strongest message we can send is to buy discrete Intel graphics cards with Open Source drivers, if/when they become available. Buy them discretely or buy them in whatever systems we buy. Vote with our $$$.
Re:So wich modern graphics card IS fully opensourc (Score:2)
I believe Intel has no reason to change their Linux friendly policy. So I hope they come up with a decent discrete graphics card and release open source drivers with it.
Since Intel is such a big player it just might encourage others to do the same.
Perhaps, perhaps not. There are certainly those that argue that in order to produce a high-performance 3D card, Intel will have to license certain technologies that don't allow open-source implementations. If they can make it happen great, but I wouldn't count on it...
Matrox G550 PCIe -- fully open source, HAL not rqd (Score:2)
You're not right about this, as HAL is entirely optional. I run the Matrox G550 PCIe [matrox.com] card *without HAL* (pure source-based Gentoo distro with the standard G550/mga kernel/X11 driver) and have all the fancy OpenGL eye-candy goodness.
But it gets even better than mere 2D eye candy. You can even run full 3D OpenGL games on this card perfectly happily and at decent frame rates, as long as the game is coded efficiently for the standard OpenGL pipeline and doesn't require programmable shaders. As an example, I run the old FPS game Cube [sourceforge.net] on this card in a slowish P4, at a very acceptable 50 FPS, and it's extremely snappy like FPS games need to be.
So don't believe everything you hear. The pure open-source Matrox driver works just great, *without* HAL.
Desktop? (Score:2)
I'd guess that they still aren't great for the current generation of games etc, but for desktop acceleration, and some old games, Intel's commitment to Open-Source has been quite impressive. Of all the companies out there, I've found that they're the most FOSS-friendly, with OSS drivers for almost all their Soundcards, NIC's, Wireless NIC's, and Graphics cards.
Re:So wich modern graphics card IS fully opensourc (Score:2)
I'm using a Radeon 9250 with open source drivers as well which runs beryl just fine. The only problem is I have to unplug my second monitor to do it. Apparently there is a limitation with the driver that it can't handle textures bigger than 2048x2048, and Beryl needs to put one big texture in videoram that spans the entire desktop or something like that. For whatever reason though if I try to run Beryl with multiple monitors it breaks very badly [curtman.mine.nu].
Oh well. Maybe next year.
hmm (Score:4, Insightful)
Hmm, could it be Dell?
Re:hmm (Score:2)
Intel cards are already open source (Score:2)
In other words... (Score:5, Insightful)
ASUS [asus.com]
Lenovo [lenovo.com]
HP [hp.com]
Power Color [tul.com.tw]
HIS Tech [hisdigital.com]
Sapphire [sapphiretech.com]
The suggested letter is:
Subject: Product Feedback
To whom it may concern,
I recently purchased one of your [graphics cards || notebooks || desktops] that had contained an ATI GPU. While I realize your products are catered toward Microsoft Windows users as they are your largest consumer base, I wish to use this product with Linux. I had used the [your model number for their product] with the ATI Linux drivers, and while they have improved a great deal recently, I still feel there is much room for improvement. The drivers in their current form run much slower under Linux than Windows, lack support for AIGLX (a visual desktop feature), and other features found within the Windows Catalyst drivers but not Linux.
I do realize you may not officially support Linux and that you have limited control over the development of these drivers, but I would kindly ask that you forward this comment to AMD and that you ask them to channel additional resources to the development of these drivers. In good time you should make Linux support from AMD a requirement. Another step that I would hope to see is including the ATI Linux display drivers on your support/driver CD. As the adoption of Linux on desktops continues to increase, I hope you are able to jointly improve your Linux presence with ATI/AMD.
[your name]
Do not copy the "suggested letter" (Score:3, Insightful)
Re:Do not copy the "suggested letter" (Score:2, Funny)
Hay [Company Name]
You guys are such jerks. Get with the times man! Open Source is everything release GNU Drivers for your [video card model] now or you will be sorry in the future. Unless you change I am not going to get any of your crap, and I am going to warn everyone from buying your junk. Common you guys can be 1337 too!!!!!!
from
[your slashdot alias]
Re:Do not copy the "suggested letter" (Score:2)
Ideally you are correct and those who have the language skills should use their own words. Having a form letter to copy and paste is a good thing for people who have a hard time expressing themselves in writing though.
Re:Do not copy the "suggested letter" (Score:2)
What is "astroturfing", anyway? Anyone who bothers to send the letter obviously cares, and there's not really any unscrupulous companies pushing for free software drivers, so no one is getting paid to send the emails. Since when is thousands of legitimate customers or would-be customers requesting something from a company known as "astroturfing"?
GREAT! (Score:5, Funny)
Re:GREAT! (Score:2)
Free Drivers ? (Score:5, Insightful)
Support Open-Graphics (Score:5, Informative)
http://wiki.duskglow.com//tiki-index.php?page=Abo
The FAQ
http://wiki.duskglow.com//tiki-index.php?page=Fre
Join the mailing list.
http://lists.duskglow.com/mailman/listinfo/open-g
And
http://www.traversaltech.com/ [traversaltech.com]
Re:Support Open-Graphics (Score:2)
Re:Support Open-Graphics (Score:2)
Their Burden, Not Mine (Score:3, Interesting)
Why should I have to prove to them that I want better support? They should prove to me that they are providing better support. Until then, I will only purchase Intel video.
The Intel 950 GMA is sweet compared to any of ATI's cards with shoddy support.
Re:Their Burden, Not Mine (Score:2)
Then I got me a nvidia card, and it worked out of the box. I also have a laptop with a mobility 9600. fglrx is an awful driver, but at least it provides OpenGL with DRI.
Re:Their Burden, Not Mine (Score:2)
I admit that stinks, but Intel has open sourced most of their drivers (including the i965), which leaves the door open for someone to create decent drivers for it. The i965 might not work but the i950 works flawlessly.
I think they may have perceived their open-sourcing as relinquishing responsibility. Kind of a "Ball is in your court now" gesture. If only things were that simple. Still, I must commend them for providing open source drivers, while ATI and nVidia refuse. Getting a better driver than ATI or nVidia gives you isn't even an option, while it's possible someone could come out with a better driver for the i965 any day now (assuming they haven't already done so).
Re:Their Burden, Not Mine (Score:2)
It's hard to develop decent (3D) drivers when all you've got is a rasterizer (full 2D pipeline).
Or (Score:2)
If you don't get any harsher a punishment for selling heroin cut with brick dust than you do for selling pure heroin, then some drug dealers are going to cut it -- which means that if they want to stay in business, all dealers end up having to adulterate their product. The consumer doesn't know what they're buying; and the end result is often either poisoning from all the adulterants, or an overdose due to a batch of gear being stronger than they were expecting.
With graphics cards, the argument that manufacturers will use is that they might give away a competitive advantage (as though they weren't all reverse-engineering the living daylights out of one another's products anyway
An alternative approach. (Score:4, Informative)
Dirk Meyer [mailto]
His executive bio is here [amd.com]. Please ask him, nicely, to open the hardware documentation, and if he could provide some resources (people, money, hardware) to the X.org team so they can build drivers.
Why bother (Score:3, Insightful)
Composite support! (Score:2, Interesting)
Better Hardware isn't worth it.. (Score:2, Insightful)
be happy with your low end ati cards & systems (Score:2)
AMD in control of ATI now. (Score:2)
AMD will soon be bundling chipsets and CPU's. Linux users are not huge by volume, but they are a good part of the "technical group" that influence others. Forcing those people to become Intel adherents(the only open source alternative) is not a good business move for AMD.
Buy stocks in AMD/ATi! (Score:2)
*sigh* (Score:2)
Caring about Linux is important (Score:2)
Reading the news archive of slashdot might make you think that Germany is a pro-Linux country makes it hard to believe the facts that we've seen more and more: AMD and Fujitsu-Siemens doesn't really care about Linux and do not hire people for Linux-support etc.
I Agree on the Tactics (Score:2)
I've said all along that IBM, HP and other companies making money off Linux should be supporting - and even financing - driver development by peripheral companies.
And when corporations finally get fed up with Microsoft and start seriously moving to Linux, we will see the peripheral manufacturers tune change dramatically as Dell and the others decide to start selling fully certified Linux machines to those corporations. It will be Dell, HP, IBM and others who force the peripheral manufacturers to produce drivers, not end users or even Linux distros like Red Hat.
HP is already selling large scale desktop Linux deployments. This situation will increase and force the major retailers to demand Linux drivers of their peripheral suppliers.
The end users really don't need to do anything but wait until clueless corporate management finally learns that Microsoft is to damn expensive (when all costs are taken into account) to support any more.
ATi is missing the boat (Score:2)
This market is especially lucrative because oftentimes the company usually wants to spend at least $5k on each computer so it counts as a capital investment instead of an expense for the bean counters. So they do everything they can to load up a base PC with extra hard drives and premium GPUs (the NVidia Quadro line is milking this pretty well with CAD/CAM workstation GPUs that cost 10x as much as the equivalent consumer-grade Geforce chips they're based on) just to drag their basic $2k PC up into capital territory. And it's getting harder all of the time since the base PC components keep dropping in price so fast. But these people can afford it and even say they're saving so much money because it's still orders of magnitude cheaper than their old big iron platforms.
Sheesh, imagine having a $5k budget to build a PC.
Anyway, making a decent (and fun, for me at least) living compiling binary modules and hand-tuning nVidia xorg.conf files for various multi-screen sim visualization configurations, among other things.
Re:Here's an improvement -- open the source!! (Score:2)
Then we know where to go to and who to spam, or wether it is feasible to just drop those parts of the code, or wether to implement open alternatives. Maybe those parts are even some fraud patents, something like matrix multiplication or whatever.
Re:Driver support can mean increased sales (Score:3, Insightful)
Re:Best solution right here (Score:2)
And if it had instead been OpenGL from the beginning (even the partial-implementation MiniGL) then we probably never would have been saddled with Direct3D. But by promoting a proprietary standard they helped set us up for a fucking. Not that they could have reasonably foreseen it, but it was stupid to do anything other than OpenGL.