Slashdot Mirror
NVIDIA Is Joining the Linux Foundation
Norsefire writes "NVIDIA is joining the Linux Foundation, along with three other to-be-announced companies. From the article: 'As one of the three big makers of graphics chips for PCs--the other two are Intel and AMD, both of which are longtime Linux Foundation members--Nvidia's increased participation in Linux could be big news for users of the free and open source operating system.
Nvidia has long taken a closed approach to Linux drivers for its graphics cards, offering only a proprietary one and declining to participate in the open source Nouveau driver project, which has depended instead on reverse engineering.'"
Hedging my bets on Apple, Microsoft and McDonalds.
From the first article:
There's no -1 for "I don't get it."
The summary implies that the submitter thinks this is going to improve things with respect to their graphics drivers. Come on. Not likely. They're doing this as an ARM manufacturer, NOT as a GPU manufacturer.
Does that mean Nvidia gonna open source the driver for the graphic cards using Nvidia chips?
Does that mean that the Linux commodities finally got tweak the Nvidia drivers to the point that they can get to squeeze the last drop of performance out of Nvidia graphic chips?
If yes, welcome to the Linux Foundation
If no, then what's the meaning of joining?
Muchas Gracias, Señor Edward Snowden !
Posting anonymously, as I was deeply involved with the OSDL. I attended a number of their high-level meetings as I work in the IT industry and was asked to contribute. What I saw was a lot of Powerpoint among CxO types who talked about being the "center of gravity" in the Linux world. Most were guys who just wanted a cheaper operating system so they could get rid of proprietary Unix...not much love for the cost of Solaris + Veritas in those rooms. They were all eager to get a free-as-in-beer OS that would save them millions a year.
The OSDL had zero ability to get RedHat, etc. to modify their plans. RedHat is very enterprise-friendly, but that's from directly working with their customers, not through OSDL.
Anyway, funding dried up and that was that.
Wikipedia says the LF has "narrowing their respective focuses to that of promoting Linux in competition with Windows". Well, good luck with that. Jumping from an enterprise focus (OSDL) to a consumer focus (LF) would pretty tough for people who have their act together...and the OSDL crowd never did. After a couple years I found myself asking "what is the point of all this? we talk and look at powerpoint, but I don't see any actual change coming out of this organization," and so I left.
1. RTFA.
2. It does not mean any of that.
3. It means that they're pumping money into Linux. For what means... speculation includes the Tegra platform (which really is not a bad speculation at all), but who knows. You also have Oracle, Adobe, etc. in that list that have little to no support for Linux with their software (or other questionable attributes).
Your impression is five years out of date.
The radeon driver for the ATI/AMD cards has improved dramatically, the graphics cards have published programming specifications, and AMD actively support the driver. It supports 3D acceleration, and is a viable alternative to the proprietary catalyst drivers in many contexts.
In contrast, to have 3D acceleration on an Nvidia card, you are often forced to install a non-free driver, and Nvidia may or may not drop support for your card as you move to newer Linux kernels. The nouveau project, while making great advances under difficult circumstances, have to reverse engineer the programming interface to the card, and do not yet have sufficient 3D support for many applications. I would hope that one day Nvidia will give them more support.
Note that this is not a comment on the relative performance merits of graphics cards from the two different manufacturers. But if you want to run 3D graphics intensive applications, and also have the benefits of a libre software environment, then it is hard to justify using an Nvidia card at present.
It means that they meet the requirements for membership* and have paid their membership fees. Which basically means they're throwing a bunch of money into a pool intended to promote, support, improve, and defend Linux and other OSS projects and developers. And getting a tax credit.
Does that mean [other stuff]?
No.
If no, then what's the meaning of joining?
It means that they've thrown a bunch of money into a pool intended to...blah, blah. And gotten a tax credit. And the right to say "Member of the Linux Foundation" on their website and other promotional materials.
* Membership is open to "...individuals and entities that engage in or support the production, manufacture, use, sale, or standardization of Linux or other open source-based technologies." (Emphasis mine.) Note that you don't even have to engage in the use of Linux--you merely have to support it (whatever that means).
Not as long as there are ex-SGI people in the place that can tell their stories about the insane amount of time wasted in court over graphics software patents. So long as the patent trolls have to do some work to determine what Nvidia have they are a little bit safer from them.
It's just another insane direct consequence of software patents.
The Tegra platforms have not gotten a lot of traction in the Linux kernel development world, generally because they put all their GPIOs in the wrong place for udev and want to put all of their board-specific GPIO renaming in #ifdef's rather than putting them in separate platform description files.
This is particularly egregious for things like the auido jacks, which due to poor code arrangement, never end up sending udev events to subsystem audio, and instead send them to platform.
I would be very happy if the only thing that came out of it was that the names they assigned to pins in the source code matched the names that they have on their technical specifications instead of having weird-ass names for everything. Right now you have to translate through three layers of indirection to figure out what you have to poke to pull a pin high.
Really, what the ARM folks need to do is get together and decide on an ISA like the Intel/AMD/IBM/yada-yada folks did so that as engineers it posible to target a single real platform. Yes, I realize that would tend to commoditize them, but they are already budge also-ran chips.
-- Terry
You also have Oracle ... in that list that have little to no support for Linux with their software (or other questionable attributes).
Oracle? I mean sure Larry's O-monster is definitely one of the major Big Evil Corporations(tm), but you can't say they have no support for Linux. Hell, the flagship product Oracle Database has been available for Linux (and even certified on several distros) for at least 10 years now -- I was running 8i on a Slackware box back in 2003!
Many years ago, they came out with their own Linux distro (based on RHEL), and now you can even get a turn-key solution that includes an "appliance" server, which runs their software ... get this ... on Linux! They will fully support you with mission-critical issues, as long as you pay for the support contract ;)
Furthermore, most people don't even know that Oracle has dedicated team of paid staff that does nothing but work on FOSS. One of these projects is OCFS2, which I have personally been involved with (as a user & community member, not a developer) for 2-3 years now and has recently become part of the mainline Linux kernel.
They were the only ones who made a GPU driver that actually both worked and performed well. Whether or not it's open source is of secondary consideration - give me a fucking GPU driver that actually pumps pixels!
Claiming open source or not makes no difference just shows that you have no first hand experience. I am currently running the open source Radeon driver, and for the first time[1] ever in my 3D accelerator history I have a platform that never segfaults (any more) handles text mode properly (looking at you NVidia) and doesn't break on every kernel upgrade. This is a huge deal-maker for me because at this point I value stability over throughput, and over being able to run OGL 3+, which is the only reason I will boot the Catalyst driver on occasion. I do not disagree that the Catalyst driver pumps more pixels - and it also has other goodies like proper antialiased lines and FSAA - but that does not matter as much to me as being hassle-free. By the way, I can do 50K triangles/frame at 60 FPS with the Radeom driver on a fanless 4830 using 50% of one Phenom II core. Did I mention, I also value quiet? It's true.
[1] Except for Intel GMA, which is also open for too underpowered for serious development work.
Have you got your LWN subscription yet?
Right now I have 28 very nice Nvidia GPU's ticking away computing determinants of large matrices for me -- and this is just to estimate how much computer time I will need for the real calculation, which will use on the order of 100K GPU-hours. The high-performance computing crowd is switching from conventional supercomputers to Nvidia GPU's as fast as the code can be written.
These things ain't cheap: the new ones that they're putting into clusters cost $1.5k each, and I bet the profit margin on them is a lot bigger than on Geforce 555M's. More importantly this is an avenue for Nvidia to dominate the high-performance computing market, especially if they do things like implement a way for a GPU on one node to talk to a GPU on another node (by a direct-to-Infiniband link or something), bypassing the PCI Express busses. (Right now it's GPU -> PCI Express -> RAM -> Infiniband -> RAM -> PCI Express -> GPU.)
Needless to say the overwhelming majority of these machines run Linux. (Your average physicist can't even imagine what a Windows supercomputer would look like. I sure can't.)