Potential Good News For NVIDIA Optimus On Linux

[Gusar]

Your typical diversion BS tactics I see...

Some people looking at this hardware from another side - ability to use more powerful GPU on laptop that they buy for desktop usage.

And how exactly does this address my argument that AMD's hack defeats the primary purpose of muxless tech? It doesn't, it's just a diversion.

And now you say something like "they delay xrandr support for 3.5 years for same reason".

Nvidia taking their time to support xrandr1.2 (which wasn't that big of a deal, as they had twinview) has what exactly to do with muxless tech? Nothing, again a BS diversion.

[RushPL]

With all the open source pledging efforts on different functionalities, is there a way to pledge money for NVidia Optimus support? (namely, this XRandR 1.5, possibly first using Nouveau + Intel, and finally, if NVidia has the will, Binary NVidia + Intel)

[RussianNeuroMancer]

And how exactly does this address my argument that AMD's hack defeats the primary purpose of muxless tech? It doesn't, it's just a diversion.

Easily - first comment in this thread about customer needs, not marketing purpose of some thing. Not all customers need switch between GPU one the fly, some need just powerful GPU, but there is no laptops without MUX-less graphics anymore. AMD provide service for such customers in the driver, officially. nVidia doesn't care about desktop usage at all.

Nvidia taking their time to support xrandr1.2 (which wasn't that big of a deal, as they had twinview)

Anyone who use both of TwinView and xrandr and deal with multi-display setup tell you xrandr is better (proper refresh rate detection by applications, ability to rotate not all displays but one, per-display applying of colour profiles, etc.)

has what exactly to do with muxless tech? Nothing, again a BS diversion.

Again. You said "Nvidia wants to do it properly." I said "they doesn't care about desktop usage at all, just face it". Proof: xrandr implementation 3.5 years story, Optimus support case (they already lag behind AMD for one year here). Now you get it?

[Gusar]

Nvidia is reviewing and through this improving xrandr1.5. They sent a patch a while back that would allow them to use dma-buf. And unrelated to that, they made VDPAU and they launched it *properly* - with full documentation on day one, plus patches for ffmpeg and mplayer also on day one. They do all this because they totally don't soooo care about the desktop. . It's not their priority, sure, that's clearly observable. But your "proof" that they don't care at all is weak at best, only works because of your other BS tactic - ignoring things that don't suit your argument.

And you still haven't addressed the argument that AMD's hack defeats the main purpose of muxless tech. "Not all customers need it" is not a counter-argument. It's a BS diversion, as I already said.
If I play the same BS tactic: not all people need to rotate just one display in a multi-display configuration; not all people need per-display color profiles; proper refresh rate reporting with twinview, while also something not all people need, can be restored by disabling dynamic twinview. So there - your arguments countered using your own tactic of diversion.

[RussianNeuroMancer]

some need just powerful GPU, but there is no laptops without MUX-less graphics anymore.

Correction: I mean MUX-ed hardware here.

They sent a patch a while back that would allow them to use dma-buf.

Do you even seen this "patch"? nVidia employee talk only about using it with Tegra, not on Optimus hardware. nVidia doesn't care about desktop usage at all, remember? (I hope after Linus speech they change mind about Linux desktop and implement Optimus support in Linux driver, but that not likely to happen.)

They do all this because they totally don't soooo care about the desktop.

Yes, and another proof of that: nVidia even doesn't make VDPAU fully compatible with compositing (if you disagree - argue with nVidia employee about that, not with me).

proper refresh rate reporting with twinview, while also something not all people need, can be restored by disabling dynamic twinview.

And get only one display.

not all people need to rotate just one display in a multi-display configuration; not all people need per-display color profiles

Sure, but this doesn't change the fact: nVidia lag behind Intel and AMD in xrandr support and still doesn't have official support of Optimus in proprietary driver. I not understand why you so afraid to face it and trying to argue.

It's a BS diversion, as I already said.

I want to remind you: talk about Bumblebee in conversation about official vendor support is bullshit diversion too. Do you realize that?

[Gusar]

Do you even seen this "patch"? nVidia employee talk only about using it with Tegra, not on Optimus hardware.

Quoting the very link you provided:

We'd only use the dma-buf interface in the
case of interoperating with the Intel device.

So you're flat out wrong here. Talk about shooting oneself in the foot...

nVidia doesn't care about desktop usage at all, remember?

Repeating something over and over does not make it true.

Yes, and another proof of that: nVidia even doesn't make VDPAU fully compatible with compositing (if you disagree - argue with nVidia employee about that, not with me).

Intel has tearing issues with Sandy Bridge. By your logic it means Intel doesn't care about the desktop at all then.

And get only one display.

And here you show that in addition to all your other BS, you're also clueless. I said "by disabling *dynamic* twinview", not disabling twinview as a whole.

nVidia lag behind Intel and AMD in xrandr support

Only if you ignore the 302.17 driver.

and still doesn't have official support of Optimus in proprietary driver.

Neither does AMD for their muxless tech. All they have is a hack.

I not understand why you so afraid to face it and trying to argue.

Dude, as long as you keep talking as if AMD has a solution (they don't), you're in no position to talk about "being afraid to face stuff".

I want to remind you: talk about Bumblebee in conversation about official vendor support is bullshit diversion too. Do you realize that?

Do you realize that getting so hung up on the word "official" is a total nonsense argument? So what if AMD's thing is "official". It's still a hack that defeats the main purpose of muxless tech. To use your own words: I not understand why you so afraid to face it and trying to argue.

But since you are so keen on the "official" word: Until xrandr1.5 is released X is incapable of properly supporting muxless tech. So all Nvidia could do is provide either an AMD-like hack, or a Bumblebee-like hack. And since there's Bumblebee, Nvidia decided not to provide their own hack. This is their *official* stance, as can be read in their PR response to Linus' middle finger.


Basically, you've lost this argument through and through. You haven't addressed the main issue that started it all. And your own link proved you wrong. Even if you now write another post with all sorts of stuff, it will still not address the main issue. So you lose. Instead of trying to argue, face it!

[RussianNeuroMancer]

So you're flat out wrong here.
So all Nvidia could do is provide either an AMD-like hack, or a Bumblebee-like hack.

You miss inconvenient (for your logic) option: at first provide official "hack" implementation that give hardware support to customers, and then provide proper implementation. This is best option for vendor who cares about customers, isn't it? But not for nVidia...

And since there's Bumblebee, Nvidia decided not to provide their own hack.

At first 2.5 years ago they said "fuck off!" to Optimus-hardware owners. That why Bumblebee was born. You, as expected, doesn't remember this.

nVidia doesn't care about desktop usage at all, remember?

Repeating something over and over does not make it true.

Ok, facts will talk instead of me:

Intel has tearing issues with Sandy Bridge. By your logic it means Intel doesn't care about the desktop at all then.

One issue doesn't make difference. Bunch of issues (like lack of proper xrandr support for years, lack of hybrid graphics support for years, lack of tear-free user experience under compositing environment for years, etc.) make the difference. Big difference in nVidia case.

And here you show that in addition to all your other BS, you're also clueless. I said "by disabling *dynamic* twinview", not disabling twinview as a whole.

What you propose? Re-configure xorg.conf by hands at every (for example) connetion projector to the laptop?

nVidia lag behind Intel and AMD in xrandr support

Only if you ignore the 302.17 driver.

nVidia already implement xrandr 1.3 support, but they do that 3.5 years later AMD Catalyst Team. This is just fact. This fact mean - they lag behind Intel and AMD in xrandr 1.3 support. This is already happened and will never change.

Neither does AMD for their muxless tech. All they have is a hack.

You may call it anything you want, but nVidia doesn't have even that in their driver. If user buy laptop with AMD hybrid graphics he may do 1-Click install in for example Jockey Ubuntu utility, but if user have laptop with nVidia Optimus, so... you know what he need to do

Dude, as long as you keep talking as if AMD has a solution (they don't)

I use Bumblebee? No. I use vgaswitcheroo? No. If they doesn't have solution, how I switch GPU? Talking "they don't has solution" is simply ignoring position, nothing more.

[Gusar]

All these words and yet you *still* have the ridiculous fixation of the word "official" and *still* haven't addressed the main argument, which is that AMD's hack defeats the main purpose of muxless tech.

I not understand why you so afraid to face it and trying to argue.

[johnc]

There is no Linux desktop.

Expecting companies to put in the time to implement features for anybody other than their customers (read: film / animation studios, HPC, etc.) is just kind of silly.

And who in their right mind would buy a laptop with a discrete NVIDIA GPU and put Linux on it?

The people at Valve must really have lost their minds trying to get into the Linux space.

And if it's bad now, just wait until the transition to Wayland hits full swing.

[RussianNeuroMancer]

All these words and yet you *still* have the ridiculous fixation of the word "official"

Compare Catalyst and Bumblebee installation/using switch. I hope someday you'll get the point of this word.

and *still* haven't addressed the main argument, which is that AMD's hack defeats the main purpose of muxless tech.

Some priorities of driver for hybrid graphics:
1. Ability to start X Server on MUX-less hardware.
2. Ability to disable dGPU to save the battery. This is what you call "main purpose of muxless tech". Catalyst can do this, and you know that.
3. If user need this - switch from iGPU to dGPU.
nVidia driver is not capable even with item 1, so for Catalyst there is nothing to compete with.