X.Org To Pay For Better Nouveau GPU Reclocking

[RussianNeuroMancer]

Or is it just a hack where everything is rendered by Radeon and output on Intel.

MUX-less "Intel HD + some other GPU" can't work in other way, on Windows too.

Thanks for your speculation.

What?! Robert Morell talk about Optimus? Where? Give me the link, give me the quote.

Just search on the web how many times the AMD driver caused BSODs on windows even with legit code, that ignoring needing tweaks to code that is JUST FINE.

Just like this code in Gnome, right? Oh, doesn't matter, smitty3268 already answer to this.

You never answered

I really need to? Anyway, if you intersted: I have four laptops with nVidia GPU (two 7300 Go (ASUS A8JN and RoverBook Voyager V550 WP), two 8600M GT (Acer Aspire 5920G-ZD1 and LG S900-U.CP38R)). On first two 3D broken one year ago (regression in proprietary driver; nouveau work fine), bug reported to nVidia, no fix for one year. On second two broken dynamic power management (screen flickering), bug reported too, no fix by nVidia for one year. Anyone have such regressions with Catalyst? I doubt.

[Gusar]

MUX-less "Intel HD + some other GPU" can't work in other way, on Windows too.

Did you read my post in it's entirety? No, this is not how muxless works. Bumblebee, while only being a very crude workaround, *does* work as Optimus is intended to work. Read my post again, and answer properly about what I'm asking about. Failing to do so will make me inclined believe Catalyst really does use just a hack instead of proper support.

Anyone have such regressions with Catalyst? I doubt.

I recall an issue where xv would crash the server, and we were telling people for months on forums to switch their players to a different output method as a workaround. So your implication that Catalyst is without bugs, bugs which sometimes linger around for months, is simply incorrect.
Also, it *does* take months for new kernels/X releases to be supported on Catalyst. And no, you don't need to post that link again, I'm familiar with it. All it shows is, their release workflow simply isn't a good fit for the Linux stack, to put it mildly.

On the other hand, RealNC and I reported xv vsync issues in the nvidia 302.07 beta driver (the first time I ever had xv issues on nvidia), and they promptly fixed it in, 302.11 works fine. And they're much quicker in supporting new kernels/X releases.

See, we all have our anecdotes. But the one big difference I noticed is, I've seen on more than one forum a huge thread regarding AMD graphics, so people constantly need to discuss how to get the thing to work properly. No such threads for nvidia. While that doesn't mean no one has problems with nvidia (I do frequent the nvnews forums), the amount of people with problems on the two sides is quite different.

[RussianNeuroMancer]

Did you read my post in it's entirety?

I expect you know - until DMA-BUF is implemented, proper switch without X-server restart impossible, but it's not AMD fault (and you know that too). Hack or not hack, key point is: with I+A or A+A MUX-less hardware AMD user can simply install the driver and switch GPU in CCC. nVidia doesn't provide even basic level of such functional.

I recall an issue where xv would crash the server, and we were telling people for months on forums to switch their players to a different output method as a workaround.

I remember many bugs with Catalyst, but looks like many people just forget nVidia bugs, just like if they had never existed.

So your implication that Catalyst is without bugs

Stop here, and give the quote. Where I talk that?

bugs which sometimes linger around for months, is simply incorrect.

For months, not for YEARS (like in nVidia case).

ll it shows is, their release workflow simply isn't a good fit for the Linux stack, to put it mildly.

Support of X-server and kernel of supported distributions (you can find list of supported here) always come in time, isnt' it?

I reported xv vsync issues in the nvidia 302.07 beta driver (the first time I ever had xv issues on nvidia), and they promptly fixed it in, 302.11 works fine

And what your recommendation to me? What I need to do with my nVidia hardware, that can not use VDPAU because the only one working option for this hardware now - is nouveau.

No such threads for nvidia.

So what? Is there enough data to make some sort conclusion from that?

the amount of people with problems on the two sides is quite different

I believe you have a representative statistics that confirm your words.

[Gusar]

I expect you know - until DMA-BUF is implemented, proper switch without X-server restart impossible, but it's not AMD fault (and you know that too).

Why couldn't you say that in the first place? You were berating Nvidia regarding their talking about their issues implementing Optimus, and saying that AMD is in a better position. Yes, they have the hack and Nvidia doesn't, but that's not proper muxless support.
And when it comes to hacks, for Nvidia there's Bumbleblee. Which works as muxless actually should, only activating the dedicated card for select apps and then deactivating it afterwards.

Stop here, and give the quote. Where I talk that?

How about the part I quoted? It implies that one won't encounter regressions with Catalyst.

Support of X-server and kernel of supported distributions (you can find list of supported here) always come in time, isnt' it?

Nice diversion. I mentioned latest X and kernel releases, and it's obvious I was talking about upstream releases, not select distributions. With Nvidia I'm not limited in my choice of distro to what's on a list. But like I said, nice diversion.

And what your recommendation to me? What I need to do with my nVidia hardware, that can not use VDPAU because the only one working option for this hardware now - is nouveau.

I don't have enough details about your issues to make suggestions.

So what? Is there enough data to make some sort conclusion from that?

Yes, that there's plenty of people having issues with AMD graphics, and that there's a lot more effort required to have Catalyst working properly, compared to Nvidia.

I believe you have a representative statistics that confirm your words.

You mean like the huge forums threads I mentioned? An example: https://bbs.archlinux.org/viewtopic.php?id=57084. There's actually a similar almost as large thread for the open radeon driver.

[RussianNeuroMancer]

Why couldn't you say that in the first place?

Because it's most hot X topic for last weeks. Why I need to "say that in the first place", instead of talking about key differences (basic level of support in proprietary driver out-of-the-box) that, as you see here, AMD-haters doesn't know about.

Yes, they have the hack and Nvidia doesn't, but that's not proper muxless support.
And when it comes to hacks, for Nvidia there's Bumbleblee.

I need to explain difference between fglrx implementation and Bumbleblee implementation from point of view of very common user, or you can understand it without my explanation?

How about the part I quoted? It implies that one won't encounter regressions with Catalyst.

You miss "such" word. Try again.

it's obvious I was talking about upstream releases, not select distributions. With Nvidia I'm not limited in my choice of distro to what's on a list.

xUbuntu and SUSE users doesn't care.

I don't have enough details about your issues to make suggestions.

Ask anything

Yes, that there's plenty of people having issues with AMD graphics, and that there's a lot more effort required to have Catalyst working properly, compared to Nvidia.

It depend on what forums you looking at. Try to сompare AMD bugtracker and nVidia bugtracker (let's say nvnews is nVidia public bugtracker; do you agree?) - you will get the opposite result.

You mean like the huge forums threads I mentioned? An example: https://bbs.archlinux.org/viewtopic.php?id=57084.

I just wonder, all this people ever contact with AMD tech. support or use feedback form? You know, they may talk about bugs forever, if they doesn't report it.

[vertexSymphony]

@Gusar → ignore RussianNeuromancer, he's just a blatant troll that will deny everything that you put in front of him, be it ignoring parts of your message, doing straw man arguments and such ... Just sit, relax and enjoy the show. things like the dma-buf being nvidia's fault, but not amd are just a fine piece of trolling art :3

[daniels]

And again someone that doesn't seem to be able to distinguish between his personal views and what's good for the larger community.

How about, in a few years, Open Source friendly companies having kick-ass drivers and hostile companies having bad/mediocare drivers.
How's that for an example?

What message are we now sending to our friends? 'We love you, but the money goes to our/your foes?'

Seriously.

you're right, i've always really wished that i could say to people, 'haha, joke's on you, you bought the wrong hardware but if you just bought a piece of shit longsoon laptop instead then everything would work perfectly'.

either that or linux works flawlessly out of the box, competitive with other operating systems, on hardware actual people really buy.

*shrug*

[RussianNeuroMancer]

Just sit, relax and enjoy the show

"Show" is right word for what nVidia do - ignore most requested feature, when AMD implement this feature at the same time. By the way you still doesn't show where Robert Morell talk about Optimus, but you still believe he talk about it. To be more clear: desktop-kind of Optimus "nVidia Maximus" is already supported in nVidia driver, and works. But this doesn't help to nVidia Optimus users at all. So if he doesn't talk exactly about Optimus, chances of official support are slim.

[asdx]

Indeed, it looks like I contradict myself, here is the reason why : we don't wait for nvidia to deprecate some cards before we start implementing support for it.

Fermi was released a year ago, 3D works great but performance is slow.
Let's say it takes a year before getting reclocking perfectly done and merged, Fermi will be 3 years old when nouveau will be ready for it.
3 years is old, isn't it?

That was my point.

As for your point about nvidia hw being not free at all, I agree. I used to think the same way up until I was given a perfect laptop ... with a nvidia card... What should I had done? Trash it? Resell it? Improve the reverse-engineered driver ?
What do you think about the Lima project? Is it worthless too?

My view: Minimal support for all devices/peripherals should be included in Linux, through reverse engineering or not.

Please ignore all the trolls and bitches from this forum and keep doing the best you can for nouveau. I'm very grateful for nouveau and it works great with my hardware (GeForce 9500 GT).

Thanks.

[crazycheese]

Right but remember that nouveau is only effective/useful on "old" cards. So, Nouveau is more about providing open source support for unsupported/poorly-supported cards.
Seeing things this way, we prevent people from trashing fairly good hardware just because it doesn't work anymore with the proprietary driver.
FYI, we still support tnt2 cards!

As for AMD and Intel hardware, open source support is already way better when it comes to power management than Nouveau. So, it is just a matter of providing at least a minimal service.

FYI GeForceMX card OpenGL support is completely broken - bad textures everywhere. Also, no one on IRC wanted these cards for free.

As of TNT, I have one, works in UMS as VGA. The card itself has virtually no advantages vs S3 PCI, due to OpenGL being almost completely accelerated in software.
For example, supertux2 is extremely slow.