NVIDIA Wants To Be A Better Linux Patron

[crazycheese]

There is a few "major" ways they can improve linux support without opening up the hardware.
...

1. At least provide some Partial opening up of the hardware on "critical" information for Open Source Developers. This information for example could be on the power management systems. After all, the Power management system would be critical in a sense that an improper setting could cause the video cards in mass to die due to overheating because of the driver having an incorrect combination of voltages, gpu frequency, and graphics card fan settings.
   ...

I have a feeling, the reason why overclocking or frequency changing is not available in proprietary under linux for fermi and up is not really lack of time.
But its due to them not wanting nouveau devs to reverse and unlock the card full performance mode (ie reclocking).
Because this would lead to much wider usage of the nouveau. And hence nvidia cards are much less dependent on driver optimizations, than amd - nouveau performance would be very competitive with that of blob.

This is strange, but I think its the true reason they don't implement it.

[crazycheese]

NVIDIA is exploiting Linux and doesn't play by the rules.
How is it difficult to understand that lots of FOSS people doesn't like that?

First come usable performance, then come opensource-ness.
This is the priority list from people who actually buy mid-high cards.

This results in major (and useless) conflict between FOSS- and this group of people.
But I guess - if you don't like nvidia - don't buy it. Buy something else that you think is better.

I don't see how nvidia is exploiting linux here. Got good opensource driver on performance hardware? - you'll be swarmed by buyers instantly. Its not the case now, though.

[Xake]

There is no Linux success outside of supercomputers, marginal server success and super-proprietary Android devices.

And virtualization-servers like XEN. and embedded. and SAN-servers. And data centers overall....
No, no success at all...

[allquixotic]

And virtualization-servers like XEN. and embedded. and SAN-servers. And data centers overall....
No, no success at all...

I agree with you, and I'll add that the server success is not "marginal" at all. Today, a majority of internet-facing web servers, email servers and even gaming servers are running Linux. It's extremely well-suited to these tasks, as well as less common ones such as streaming media (icecast), serving files over ftp/bittorrent/whatever, and "compute-heavy" tasks (with or without a GPU).

I work at a large enterprise. I asked around with our server guys, and about 25% of our server infrastructure runs Linux. I asked why more of it doesn't run Linux, and they said that they're planning to move most of their existing Windows infrastructure over to Linux within 5 years, but they haven't done so yet because it requires a lot of planning and red tape. So they definitely want to embrace it but they just need time and money to break away from Microsoft's vendor lock-in. It'll be better in the long run.

[johnc]

I agree with you, and I'll add that the server success is not "marginal" at all. Today, a majority of internet-facing web servers, email servers and even gaming servers are running Linux. It's extremely well-suited to these tasks, as well as less common ones such as streaming media (icecast), serving files over ftp/bittorrent/whatever, and "compute-heavy" tasks (with or without a GPU).

I work at a large enterprise. I asked around with our server guys, and about 25% of our server infrastructure runs Linux. I asked why more of it doesn't run Linux, and they said that they're planning to move most of their existing Windows infrastructure over to Linux within 5 years, but they haven't done so yet because it requires a lot of planning and red tape. So they definitely want to embrace it but they just need time and money to break away from Microsoft's vendor lock-in. It'll be better in the long run.

Like I said... I'm open to numbers. I couldn't find much except for the ones I quoted earlier.

I would love to know what the actual market share is.

[hal2k1]

Maybe I have my numbers wrong but my understanding is that Windows is top on servers. But all I have are IDC (non-revenue) numbers which showed Windows at 40%, Linux at 30%, and "other UNIX" at 30%. I'm looking for better numbers so if you have a (reliable) source, I'll definitely correct myself.

Okay, you got me on the embedded sector. It's in keyboards and routers too. Credit where credit is due.

If one were to look at the mobile market, you can see that a Linux kernel isn't necessary for success.

Android uses Linux.

Linux has top market share in: supercomputers, embedded (e.g. TVs, media players, cameras, microwaves, GPS units, automobiles et al), infrastructure (e.g. routers, web servers, email servers), massively parallel server farms (e.g. Google, facebook), point-of-sale and mobile, and it has a major share (if not the major share) of enterprise servers.

http://www.wired.com/wiredenterprise/2012/03/red-hat/

http://articles.businessinsider.com/...hat-jim-zemlin

Just about the only markets where Linux is not the dominant player are desktops (Windows) and tablets (iOS).

An average family might run one copy of Windows on their desktop computer, another on a laptop, one or two copies of iOS on their ipads, and perhaps a dozen copies of Linux on the remainder of their household electronics.

[hal2k1]

Like I said... I'm open to numbers. I couldn't find much except for the ones I quoted earlier.

I would love to know what the actual market share is.

Most of the consumer items which run Linux do not "advertise" the fact that they do run Linux. If a display with a graphics UI (e.g. a TV screen when you press any of the buttons on the remote) doesn't actually identify the OS, you can bet your bottom dollar that it isn't Windows, OSX or iOS, but rather it is Linux.

That makes it really hard to say what you mean by "market share", and who is doing the counting of "market share", and exactly what they are trying to achieve by their publishing of that figure.

If you are just curious, and you want to know the percentage of CPUs (which run an OS and have a graphics requirement) are those which run Linux, then my guess would be that well over 80% of the OSes running right now on the world's CPUs run a variant of Linux.

You won't find this out by reading published figures for "market share".

[V!NCENT]

Bypass the problem! Don't try to open up what can't be opened up! Instead, do something else with nVidia hardware, like supporting this:
http://www.phoronix.com/scan.php?pag...tem&px=MTA3OTI

What to take up to upper management:
Change CUDA from a 3dfx' Glide open source thing to a OpenGL-ish CUDA thing. What I mean by that is CONSORTIUM TO FORTIFY CUDA ACCEPTENCE BY MAKING IT AN OPEN STANDARD.

---

If you can get the above two things done, that would definately improve your image in my book. In fact, I might even get interested in nVidia cards if that means having the IGP be my 3D desktop engine and nVidia my parrallel compute engine! If I can't have the open 3D nVidia acceleration, at least make me able to completely bypass it (turn it off) and make me use it otherwhise.

---

It's nice to hear that there is some part inside of nVidia that wants change. You have my support if you can get the board to get behind the above.

PS: transfer the open kernel module + proprietary driver to a combination of Gdev kernel module and hook the proprietary driver into the Gdev infrastructure. Then it's solid gold. I can get the AMD Fusion to be the CPU+GPU and the nVidia card to be the 'IBM Cell' compute card'. And I could run the proprietary driver contained if need be, which must not fight against kernel components but work with them.

[V!NCENT]

Think about the market advantages! The board needs figures, so how about this:
-Jumpstart into the Cloud; imagine how much the likes of Facebook can get to buy this: massive parrallel tiny operations on CUDA because of kernel acceleration;
-Massively reduce server costs by needing less compute units (saving space, cut down on cooling, saving energy, reducing cost all over the board);
-Improving nVidia's image means free marketing;
-Offloading extra development costs for free to the likes of PathScale (let them sign an NDA for improving key parts, saving nVidia cost and improving PathScale income).

[korrode]

No.
A WONDOZE user who happens to be sitting in front of a Linux computer is satisfied with the useless blob.
A LINUX user is totally unsatisfied with dealing with bullshit 3rd party closed drivers that could very easily be SPYING on them. I dare you to PROVE that they aren't.

For over five years I have been using Linux as my primary OS (and by primary I don't mean 60% of the time, I mean 99.999999%. I actually think the HDD in my system holding a Windows installation failed in some way quite some months ago (was making some bad noises on power up), but I wouldn't know, because I haven't looked into it let alone tried to boot it).

I run a highly customised Arch Linux on my desktop, Debian on my server.

This makes me a "LINUX user" no?

"Satisfied" is an accurate enough description of my feelings about their binary blob. (Not like I'm against them open sourcing it, but I don't have any real qualms with the current situation.)

You see the cup as half-empty because nVidia, who provide the only actual good 3D solution on Linux, don't open source their driver.
I see the cup as half-full because nVidia, unlike ANYONE else, actually provide us with a good 3D solution for Linux.