I just bought an nvidia card

[grantek]

RealNC, can you tell me why the hell non-geeks should be upgrading their video driver at all? If they're doing it to maintain their system properly, then they should at least know what they're doing...

[jeffro-tull]

My brother and I just bought GeForce 9400GT's, and they work beautifully. The pragmatist in me forgives the closed-source blob on the grounds that I didn't have to fsck with it. The idealist in me is still, 3 years later, waiting for any kind of driver that will do everything for the Radeon X1300 one of those GeForces replaced.

[Milyardo]

In my opinion, Nvidia's "let's inlcude everything w/ the kitchen sink" philosophy doesn't make a good driver, nor does it provide any benefit to Linux development of OpenGL or its dependent applications.

On one hand, Nvidia's commitment to include many experimental and non ARB OpenGL extensions provides the community with early(or exclusive access in the case of some OpenGL extensions in Nvidia's driver that were meant to be Windows only) access to many different extensions to use that make OpenGL a powerful, and easy to use tool with many incentives over competing not so open platforms.

Sadly many of these experimental and/or proposed OpenGL extensions never make it past the ARB, or get accepted with altered specification. However, these extensions sometimes build up a legacy application base by time they are depreciated.

Still however, Nvidia does not remove these extensions after they've been officially depreciated, or rejected by the ARB. This IHMO, has the most detrimental effect on Linux and OpenGL application development. Its extensions like these that often lead novice OpenGL programmers to dismiss OpenGL as complicated, poorly supported and implemented, and buggy as I often see many OpenGL tutorials use these extensions in combination with newer extensions(This is much like using using DX7 with DX10 and expecting cohesive functionality between the two namespaces).

Conversely, I enjoy knowing that in ATi's driver, will follow the OpenGL ARB specification to a tee with most of their extensions(Even if they be a smaller subset of extensions). Unfortunately they have decided to deviate from specification before maintain compatibility with Nvidia's/Legacy(/radeon[-hd], though not so much anymore) implementations, such situations are inevitable. Still, fglrx's implementation of OpenGL the closest the to specifacation, which is why I, and I hope also any well learned Linux developer would, use ATi, both open and closed source drivers, as a reference implementation of OpenGL in their application development.

I know many would call ATi's conservative approach to OpenGL(and the driver development in general) what makes it "broken". Or even Nvidia's relesse philosophy fits in better with the development rhythm of the Linux kernel.

As far as my critisms for fglrx go, for almost a year now, fglrx was implemented without changing much of the existing Xorg stack, however since the end of support of r500 cards and below, this has started to change(fundamental functionalist would argue this is good thing, and the first step toward fglrx because like nvidia). From a usability perspective, memory management is obviously fglrx's greatest weakness. I would like to think this is because fglrx also is effected by changes in the kernel and DRM with the merging of GEM into the Xorg stack.

I lastly think arguments about the future of either driver is moot, hopefully soon to come changes in the XOrg and the Linux kernel will make vendor distributed XOrg stacks redundant and that both drivers are soon downsized to small DDX components with libraries only for patent encumbered use cases like BD+ and HDCP.

I believe ATi shares this vision for the future of fglrx, and proprietary drivers in general. Though I wouldn't hold my breath for Nvidia to help make this a reality.

[kraftman]

RealNC, can you tell me why the hell non-geeks should be upgrading their video driver at all? If they're doing it to maintain their system properly, then they should at least know what they're doing...

Yeah and it's as simple as adding new repo in Kubuntu. Simpler, faster and better then on windows.

[deanjo]

Yeah and it's as simple as adding new repo in Kubuntu. Simpler, faster and better then on windows.

In win land it's called windows update.

[nanonyme]

In win land it's called windows update.

<offtopic>There's imo exactly one thing the Windows alternative is better because of: you can fallback to situation before install on each install and iirc it even pretty much auto-fallbacks if it encounters an error during system component install (at least on new versions). We can hope Linux package managers get this eventually, it's actually pretty much critical if you want to avoid reinstalls at all costs.</offtopic>

[BlackStar]

In win land it's called windows update.

Video drivers on Windows update are horribly outdated and come without OpenGL support. I pity the poor devils that use windows update to get their video drivers.

[deanjo]

Video drivers on Windows update are horribly outdated and come without OpenGL support. I pity the poor devils that use windows update to get their video drivers.

For most people they work fine. Gamers will know to update, but for 95% of the users the "outdated" drivers work perfectly fine for what they want to do on their computers.

[curaga]

Video drivers on Windows update are horribly outdated and come without OpenGL support. I pity the poor devils that use windows update to get their video drivers.

Agreed. On more than one occasion, right after installing the latest Nvidia driver manually, Windows Update has proposed a driver from 2005 as an update to it. If an user doesn't notice that and presses go, it actually degrades the experience.

[BlackStar]

For most people they work fine. Gamers will know to update, but for 95% of the users the "outdated" drivers work perfectly fine for what they want to do on their computers.

A couple of months ago, I installed Vista for a friend. Windows update helpfully installed a driver for his 4850HD, which didn't even support the card (immediate blue screen). It then kept overwriting the correct driver on startup.

For the record, Jaunty installed and ran perfectly on the very same system, without configuring a single damn driver.

@Milyardo: it is true that Ati drivers tend to be more strict when enforcing the OpenGL specs. Unfortunately, their OpenGL implementation tends to be more buggy than Nvidia's (for example, FBO blits of multisampled depth renderbuffers have been broken for more than a year!)

You really need to test on both kinds of hardware while developing, if you wish to maintain any semblance of compatibility. If you actually need to support Intel hardware too... you are in for a world of pain.

Truth be told, there's a good reason why OpenGL is marginalized. OpenGL has two things going for it: cross-platform support and VR extensions. The API itself is ugly, inconsistent, inefficient (sampler state is still part of the texture object state) and downright awful at places (recompiling shaders every time you run your program!) Not to mention the god-awful driver support for anything more complex than spinning cubes.

I've lived and breathed OpenGL during the past three years; I've edited the specs directly; reported 15-year old bugs to Khronos; logged driver with issues Intel, Amd and Nvidia; I've implemented two different OpenGL bindings, plus a freaking 300Kloc platform abstraction to bring GL/ES/AL/CL to Mono/.Net.

I've invested so much time and energy that it physically hurts to admit that OpenGL is on the brink of death. Direct3D is a much better API, easier to use, more efficient and faster evolving. XNA is so simple that even the scientific community is jumping ship. Hell, even OpenGL ES is better - I'd use it in an instant if it existed on the desktop!

But we don't really have a choice: if you care for non-Windows support, OpenGL is the only way. Considering the odds, most developers simply don't bother anymore.