NVIDIA Wants To Be A Better Linux Patron

[uid313]

Here is another plan ... ignore nvidia, and also ignore machines with UEFI secure boot designed for Windows 8.

Just run Ubuntu on new machines such as the Mele A1000 or the MK802 micro-PC.

On a 1 GHz single-core CPU?

I want at least quad-core and 8 gb RAM as bare minimum. Preferably better.

[hal2k1]

If they don't satisfy us, that would be an excellent plan B.

Who are they going to sue? Thousands op GPL contributors? Lols. And if nVidia can do what it wants, how can't we? It'll just be another middlefinger, but from the a much larger crowd.

Here is another plan ... ignore nvidia, and also ignore machines with UEFI secure boot designed for Windows 8.

Just run Ubuntu on new machines such as the Mele A1000 or the MK802 micro-PC.

http://liliputing.com/2012/03/mele-a1000-is-a-70-hackable-linux-friendly-arm-based-pc.html

Powered by a 1 GHz Allwinner A10 ARM Cortex-A8 processor, the Mele A1000 should be noticeably faster than the Raspberry Pi, which has a 700 MHz ARM11 chip.

While the Mele A1000 is described as a TV box, it’s basically a little ARM-based computer with HDMI, VGA, USB, and Ethernet ports, as well as support for an external SATA hard drive.

Now for the most interesting part (if you happen to be a free software enthusiast): The folks at Rhombus Tech have been working on a free and open source platform based around the Allwinner A10 chip.

The project is entirely GPL compliant, and the goal is to work with a Chinese company to produce a PCMCIA-sized computer which is capable of running Linux-based software without any proprietary, closed source drivers. Rhombus Tech hopes the platform can be used in all sorts of devices including tablets, notebooks, desktop PCs, and in-car systems. Theoretically the price could be as low as $15 per unit… assuming an order size of 100,000. Right now the cost is higher.

But thanks to the efforts of the folks at Rhombus Tech, developers have access to all the drivers needed to make use of any device with the Allwinner A10 processor, and that includes the Mele A1000.

A fully-fleged desktop PC, with full desktop software, (less monitor and keyboard) for $70 per seat.

Here is another:

http://www.pcworld.com/businesscenter/article/255837/this_tiny_pc_runs_linux_and_android_40and_costs_just_74.html

Featuring a single-core 1.5GHz AllWinner A10 Cortex A8 ARM processor, Android 4.0, 512MB of DDR3 high-capacity memory, and WiFi connectivity, the MK802 is now available on Aliexpress for $74 including free shipping to the United States via China Post.

With a MALI400 graphics processing unit, the device from Chinese brand rikomagic features 4GB Flash storage, a microSD slot, and two USB ports: one full-sized and one micro, according to CNXSoft. Video output is via 1080p HDMI--an HDMI cable needs to be added separately--and users can tap either an Android virtual keyboard or add a wireless mouse and keyboard.

Perhaps best of all is that users can run Ubuntu, Debian, or another Linux distribution of their choice via microSD card.

These $70 micro PCs both use the AllWinner A10 Cortex A8 ARM processor with the MALI400 GPU. They both use minimal power yet have enough grunt to run desktop software and decode and render 1080p video.

A company with generic office-type IT needs can potentially set up their entire IT infrastructure based on this type of hardware at perhaps only 5% of the cost-per-seat of a Windows 8 solution. Maintenance? Pffft ... if a box breaks, get a new one ... here, have a spare as well. OS corrupted? Here, plug in this new SD card.

http://www.calligra-suite.org/

http://www.netrunner-os.com/features/

Very nice desktop software. Excellent.

This solution would also be ideal in an educational setting.

[V!NCENT]

If they don't satisfy us, that would be an excellent plan B.

Who are they going to sue? Thousands op GPL contributors? Lols. And if nVidia can do what it wants, how can't we? It'll just be another middlefinger, but from the a much larger crowd.

[hal2k1]

You can't discriminately lock out only the nvidia blob just because you don't like nvidia. You'd have to lock all blobs out or risk a lawsuit.

Then lock out all blobs out of the kernel ... leave only firmware binaries (firmware runs on the card, it isn't part of the kernel). Easy peasy. UEFI secure boot will soon do this anyway.

BTW, you truncated the post you quoted. It actually finished like this:

I think it would be far better and cheaper for Nvidia and everyone if they just released specifications so that Linux developers could write their own drivers. Is this really too much to ask?

Why did you leave out the critical question at the end? Why is it too much to ask for programming specifications? Why on earth shouldn't nvidia release programming specifications if it wants to profit from selling a hardware subsystem that can be used as part of Linux machines?

"Linux machines" is now a market that is worth trillions. I'm not talking about only the desktop market, but rather the entire computing hardware market, into which nvidia could conceivably sell graphics hardware. Linux is a dominant player in this wider market.

If nvidia want to be a part of it, and profit from it, why shouldn't it be a requirement that nvidia releases programming specifications?

We can make it so. It can very easily be done. Lets do so, I say.

[V!NCENT]

Exactly. If nVidia wants to do everything in their power not to be able to benifit from the free software market, then let them. If they lose the 1% extra income that could make or break their quarterly finance reports, let them screw themselves.

In the event that nVidia realy does want to be a free software patron, lets help them. If it turns out to be a case of "actualy we don't care", then let them lose Android and Linux desktop customers permanently to AMD and Intel.

If nVidia realy can't go the Gdev route, then they can call their license contact and tell them "if we can't open up x part of y documentation, we loose money and are forced to not license from you again, so give us the ability to open up or you'll lose a big licensing customer. Don't think you are able to knock on door at AMD or Intel". Simple. Then change/add some part of a future hardware design if needed.

In the end it's all about money. Don't satisfy your customers? You'll lose. China is the biggest example of failure.

[XorEaxEax]

You can't discriminately lock out only the nvidia blob just because you don't like nvidia. You'd have to lock all blobs out or risk a lawsuit.

Yes, the whole banning 'proprietary blobs' is just nonsense, although it's not as if there's alot of them left. The way the devs are handling it now is just fine, they've announced their displeasure with binary blobs and the practical reasons behind it, they refuse to make it easy for binary blobs to work with the kernel and they go out of their way to make it as beneficial as possible to have your drivers as open source within the kernel tree.

If some companies still find that they want to retain the control over how and where people who bought hardware from them use that hardware and thus refuse to open either specs or (better yet) provide an open source driver then so be it, they will have to carry the burden of keeping a working driver out-of-tree.

As for those who seem so adamant to jump to NVidia's defence because -'they provide the best 3d drivers for the Linux desktop', again realize that the reason NVidia bothers to bring drivers to Linux has absolutely nothing to do with them giving a crap about the Linux desktop, they are catering for Linux high usage in 3D/SFX/scientific/industry etc where there are strong buyers interested in utilizing GPU's for high performance visuals and GPU based calculations.

However given that those Linux sectors are only getting stronger those of you currently using NVidia proprietary drivers on the Linux desktop have nothing to worry about, Torvalds and the rest of the Linux kernel devs could call NVidia 'terrorists' and they still wouldn't stop producing their proprietary Linux drivers.

On the other hand if you want to use NVidia tech which is particularly desktop/laptop oriented (optimus) then you are out of luck because that's not where NVidia's big Linux customers are and again they couldn't give a rats ass about the Linux desktop end users.

[V!NCENT]

If we all get behind 'my' idea, then that's what they have to do to please us. Then if they don't? It's no damage controll for them and they're done for. That's how it's done.

And if that would mean having to implement a new hardware design, it's in nVidia's interest to tell us that and keep us informed. We could then even (since the stack is open already (Gdev+PathScale+nVidia compiler(they open sourced it!))) implement a 'software Mesa' on top of it.

Calm down and get it together guys.

[V!NCENT]

Hey c'mon... WTF! nVidia doesn't HAVE to play nice with others, but then they are not getting 'our' support.

They asked what they COULD do to make us happy. I think my proposal @page 10 was fair (they don't have to open any magic, we get a consumer prices IBM Cell processing card for free, they can work on other kernel components and the blob works well with others, plus improved security).

If you don't like nVidia? By all means don't buy it, but if they ask us if they could do something else open sourcy for us, don't try to ban them and shoot yourself in the foot. C'mon... Are we 12 year old Stalins? At least give them a change to do some giving back to open source. It would only improve our situation and nobody has to by their cards.

[Xake]

You can't discriminately lock out only the nvidia blob just because you don't like nvidia. You'd have to lock all blobs out or risk a lawsuit.

Well, in practice with Secure Boot that would be the deal: no binary blobs will work unless they get the blobs signed by that specific distribution.
So nVidia will tell their costumers to just not enable Secure Boot. And nVidia and all other "blobbers" without a open implementation at hand will lose customers because some adapters really seem to like the idea of secure boot handled right...

[johnc]

So you think that Nvidia will have to maintain a custom kernel forever? That'll be lot of work, how much backporting they will have to do to catch up with the latest kernel development and drivers? You know, hardware changes all the time, and the mainline kernel will be years ahead of what they have.

This will cost them more money in the long run, I don't see every Ubuntu user having to downgrade their kernels for running the nvidia blob.

I think it would be better for Nvidia and everyone if they just released specifications so that Linux developers could write their own drivers.

You can't discriminately lock out only the nvidia blob just because you don't like nvidia. You'd have to lock all blobs out or risk a lawsuit.