NVIDIA's Linux Driver On Ubuntu Is Very Competitive With Windows 8

[schmidtbag]

That was a great play by Torvalds! Thanks to that nVidia started working seriously on Optimus for GNU/Linux. Linus is a very intelligent person, he knew that it would draw attention.
In the same way, it is good to criticize the flaws in AMD hoping that they read and take action on the matter. For me the open-closed driver discussion has no importance any more. It is important to provide a good driver, whether open-closed source. But you as a hardware manufacturer choose what you think is the best choice between the two, and use all your resources in a single work.

Assuming Nvidia tried harder because of Linus, they still didn't do what he wanted. Linus doesn't dislike Nvidia's software and drivers, what he didn't like is how they don't cooperate with the community and are a pain to work with because they always do things their own way. Nvidia's decision to put so much attention toward their closed-source linux drivers might have actually been their way of flipping off Linus, because it basically says "here's proof that we don't need to listen to you". While I wholeheartedly understand and agree with Linus' opinion of nvidia, as long as nvidia supports what linux users want (in terms of end results), it's hard to argue against nvidia's decisions.

[mendieta]

While Linux isn't going anywhere any time soon, Nvidia doesn't want to take the chance of putting the reliability of their products in the hands of people who can't be 100% trustworthy.

Right, the development community that built the linux kernel, the most widely deployed operating system kernel in the planet, the one that feeds 95% of the top supercomputing world, can't be 100% trustworthy. That's a pretty solid argument, right there. The open source model also clearly failed in Android, Samba, Apache, etc ... right?

I am grateful to NVIDIA for supporting Linux users the way they choose, but I'll buy hardware from manufacturers who make my life easier. Open source support makes my life easier (easier upgrades, more stability, easier to fix bugs), so there goes my money.

Cheers!

[YAFU]

Assuming Nvidia tried harder because of Linus, they still didn't do what he wanted. Linus doesn't dislike Nvidia's software and drivers, what he didn't like is how they don't cooperate with the community and are a pain to work with because they always do things their own way. Nvidia's decision to put so much attention toward their closed-source linux drivers might have actually been their way of flipping off Linus, because it basically says "here's proof that we don't need to listen to you". While I wholeheartedly understand and agree with Linus' opinion of nvidia, as long as nvidia supports what linux users want (in terms of end results), it's hard to argue against nvidia's decisions.

Well, if I remember correctly Torvalds reaction was for a question related to Optimus on Linux. Obviously Linus prefers OpenSource collaboration.
For me an ideal model of hardware manufacturer would be to put all their resources into OpenSource drivers (Intel is the closest to that model). Otherwise if the choice of the company is proprietary drivers, at least they collaborate releasing specifications for the community to develop basic OpenSource drivers. AMD fails to use all their resources to develop a better and unique driver, and nVidia fails to cooperate with the OpenSource community. But most importantly as I said before, is that the user has a good driver support in the development model that the company has chosen.

[wargames]

I am grateful to NVIDIA for supporting Linux users the way they choose, but I'll buy hardware from manufacturers who make my life easier. Open source support makes my life easier (easier upgrades, more stability, easier to fix bugs), so there goes my money.

- Easier upgrades: Nvidia binary driver (a.k.a the "blob") supports the latest hardware at launch time. Something that neither nouveau nor the open source AMD driver do. We are still told to use the 6000 series more than one year after the 7000 series launch (and the 9000 series are coming soon), because RadeonSI is not ready yet and you are stuck with OpenGL 2.0 (an API which was released on September 7, 2004. That is nearly 10 years ago).

- More stability: I wouldn't say the nouveau drivers are particularly stable. And yes, the open source radeon driver is stable but pretty slow and not feature complete. Why should I buy a $500 gfx card and not be able to run it at its full speed and be limited to use OpenGL 2.0 with no OpenCL support? Because of "freedom" ?

- Easier to fix bugs: Easier for whom? Programmers know how to fix bugs in both open source and closed source drivers. And don't tell me that having the source allows you to fix bugs 10 years after the company that manufactured your gfx card has ceased its support, because you are not going to be using the same hardware. We have also the issue of who (in the open source world) is supposed to fix the drivers for your 10 years old gfx card... YOU? Are you going to pay someone to do it? Or are you just going to demand someone to do it because it is open source and you think programmers work for free?

[Kano]

Basically i think the title is wrong as only OpenGL rendering was tested. The Unigine Heaven test has however Direct 3D as well. Especially when you begin testing with SLI/Multigpu or with Crossfire with AMD hardware then you get completely different results. D3D on Heaven is much better optimized, Nvidia OpenGL 4 is not that bad compared to D3D, AMD OpenGL 4 is worse. SLI works a bit with Nvidia OpenGL on Win with Heaven, AMD has got no CF OpenGL scaling at all with Heaven but interestingly when you play Rage which has no SLI profile. Without tesselation the difference between Linux+Win with AMD OpenGL with Heaven was not so extreme. I posted all results somewhere here when i did those benchmarks, feel free to search em. I also tested Killing Floor D3D (2 modes) vs. OpenGL on Win+Linux with several cards. KF D3D was always much smoother, Serious Sam 3 is much faster with D3D, OpenGL is similar with Linux+Win but slower. Source engine games suffer from the extra conversation layer especially on slower systems. When you really have got highend hardware then the conversation is not that critical, also Nvidia multithreaded the driver just for Source engine games. KF has got rendering bugs with Nvidia cards with OpenGL (on win as well btw.) but not with AMD cards. So to sum it up: when a game has an optimized D3D engine and the OpenGL engine is not optimized then you win nothing using it, the os does not matter much as the code is shared anyway.

[johnc]

I wish this was true for nVidia GT520 and games I play. On Windows, I get much smoother performance in Alien Arena (framerate is a bit worse, but the game looks smoother than on Linux - I have no idea why). The same for Source engine games: Team Fortress 2 on Windows: 20-40 FPS (lowest and highest), on Linux: 10-30 FPS (lowest and highest) and the game gets stuck everytime there are too many objects on screen moving (this does not happen on Windows)

Make sure multicore rendering is disabled in the options.

Also a +1 for the idea of benching D3D vs. OGL in Unigine.

[schmidtbag]

Right, the development community that built the linux kernel, the most widely deployed operating system kernel in the planet, the one that feeds 95% of the top supercomputing world, can't be 100% trustworthy. That's a pretty solid argument, right there. The open source model also clearly failed in Android, Samba, Apache, etc ... right?

Way to completely misinterpret what I said in the way you found convenient. I wasn't talking about the OS as a whole, I was talking about a select view developers who feel nvidia is making life difficult because of how they do things their own way. And all those products you mentioned (Android, Samba, Apache, etc) are developed by PAID DEVELOPERS. Not only that, but they have little to no relevancy with linux's GUI (android uses it's own). Those developers of those projects are trustworthy and reliable because they're contractually obligated to be. As far as I'm aware, the kernel is NOT completely composed of paid developers, and I get the impression wayland, mesa, X.org, and most desktop environments are mostly volunteer work. On top of that, there are relatively very few people who actually actively and continuously contribute to those projects. Sure you could argue Intel has a lot of dedicated developers that contribute, but that's irrelevant because they're focusing on promoting their own personal graphics platforms and could care less what nvidia does.

My point is Linux's graphical system is, as a whole, pretty disorganized and somewhat inefficient. While nvidia doing their own thing obviously isn't helping clean it up, they're picky about doing things right and they simply can't rely on volunteer developers to get things done the way they expect them to be done. You could argue that they could just pay these developers, but it'd be cheaper and easier for them to do things their own way. Again, I'm not saying I like this, but I can't fault nvidia for the way they do things.

[johnc]

The open source model also clearly failed in Android, Samba, Apache, etc ... right?

Closed-source drivers are somewhat common in Android devices.

I feel like Android gets too much of a pass from people who often deride NVIDIA for having a closed-source driver for x86 systems. It isn't obvious that Android would have been successful if every driver was required to be open source.

[Temar]

- Easier upgrades: Nvidia binary driver (a.k.a the "blob") supports the latest hardware at launch time. Something that neither nouveau nor the open source AMD driver do. We are still told to use the 6000 series more than one year after the 7000 series launch (and the 9000 series are coming soon), because RadeonSI is not ready yet and you are stuck with OpenGL 2.0 (an API which was released on September 7, 2004. That is nearly 10 years ago).

- More stability: I wouldn't say the nouveau drivers are particularly stable. And yes, the open source radeon driver is stable but pretty slow and not feature complete. Why should I buy a $500 gfx card and not be able to run it at its full speed and be limited to use OpenGL 2.0 with no OpenCL support? Because of "freedom" ?

- Easier to fix bugs: Easier for whom? Programmers know how to fix bugs in both open source and closed source drivers. And don't tell me that having the source allows you to fix bugs 10 years after the company that manufactured your gfx card has ceased its support, because you are not going to be using the same hardware. We have also the issue of who (in the open source world) is supposed to fix the drivers for your 10 years old gfx card... YOU? Are you going to pay someone to do it? Or are you just going to demand someone to do it because it is open source and you think programmers work for free?

^ THIS.

I fully agree. If you actually want to use the features of your GFX card, Nvidia is the only choice on Linux. Most applications just run better with a Nvidia card because even open source applications prefer the Nvidia APIs. Just think of XBMC/ffmpeg/vdr/mplayer which have been supporting VDPAU for ages while VAAPI was a pain in the ass for a very long time. The same is true for GPU computing where many apps properly support CUDA whereas OpenCL support is a mess (e.g. Blender) or not even existant.

[GreatEmerald]

How the heck do you search for these results in openbenchmarking.org? Michael mentioned last time that if you have SVG graphics you should be able to click in any plot and get redirected to openbenchmarking. This doesn't happen in any of my computers! It would be nice to have automatic linking of articles to the source test results ...

Click the minuscule boxy-box at the bottom left of the chart.