1) Nouveau is buggy and problematic (related more to performance and a lack of support for many features)
2) The Nvidia drivers are buggy and problematic (related more to setup and compatibility with the rest of the system)
Both of these problems can be traced back to the same root cause, which is simply that Nvidia refuses to do a good job of supporting Linux in any way. They could give Nouveau some funding and information about their hardware, but they don't do that. They could develop proper Linux drivers, but they don't do that either. Every way you look at it Nvidia is at the root of the problem.
The solution on the consumer side is to not buy Nvidia products, and to encourage other people to not buy Nvidia products. The solution on the producer side is to contribute to Nouveau (easier said than done unfortunately, your average web dev can't just jump in and reverse engineer graphics drivers on a weekend).
I really dislike the way Nvidia makes it difficult for projects like Nouveau to reverse engineer an open driver, however I don't understand what you mean by poor support for their Linux driver, it's probably the one thing I can't fault them for as it is performing very well and stable.
I'm with you on not buying NVidia products though, that is the best way to send a message.
All my experience has been on laptops and there are lots of reports of Optimus related problems on Linux so there's that. I gave up on the hope of having Optimus actually work years ago. I would like to just be able to enable the latest version of Nvidia's driver and see my laptop start using the discrete gpu. That would be a big step forward from the past 7 years.
FWIW there's Linus' famous Nvidia F-bomb :) "Nvidia has been the single worst company we ever dealt with" https://www.youtube.com/watch?v=IVpOyKCNZYw
There's a reason Microsoft shimmed the ability for the display driver to crash and restart without taking out the rest of Windows...
Anyways, it would be good if NVidia and Ubuntu got together and at least made it easier to make use of the official driver.
Aren't scientific farm servers running on Linux kernel with Nvidia cards ? It would not be used if it was that buggy, so I think that claim needs precision
Reproducibility in computational science is recognized as a huge problem: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3383002/
A lab or institution claiming a publishable result based on discounted hardware from the manufacturer running code that does who knows what for their TensorFlow/whatever calculations is engaged in something other than science. I wouldn't go quite so far as to say that they're just engaged in advertising demonstrations, but it's coming close.
Reproducibility doesn't necessarily mean every component of the system is completely open and inspectable.
I wonder what percentage of "Machine Learning" results are based on such shaky foundations and when they'll receive the same amount of skepticism and scorn we see heaped(deservedly) on the field of psychology.
That doesn't follow. They can (attempt to) reproduce with the same hardware, firmware, and drivers. They can reproduce with similar hardware from another vendor (AMD). If you're using OpenCL, you can reproduce on CPUs, albeit much slower and probably only for partial results. Maybe you want to break out some pen and paper to verify a few calculations of the CPU for even more partial results because you don't trust Intel's firmware blobs, even though you've presumably cross checked against AMD hardware, but you've gotten quite silly at that point.
Maybe some of these changes won't give you bitwise identical results - but then again, neither does most science. You check if the broad results and patterns are statistically similar. Maybe they are, and the results survive a large variety of similar but different test setups. Maybe they don't, and you discover an uncontrolled variable, like the exact chemical composition of the surface of your glass beakers being contaminated by impurities, or your AI being sensitive to the exact floating point rounding behavior of your GPU.
If anything, you want to see if your results survive variations in the uncontrolled variables, to make sure you've properly determined what factors are controlling the results.
If you cannot reproduce then you do not know whether it is due to the equipment/materiel varying or not and you have severely reduced options for investigating that.
If you look at the problems with even reproducing and testing the bugs that the specific situation under discussion exposes it is illustrative of the difficulties involved with working with a non-modifiable, unexaminable platform.
In your example, the researchers would be forbidden by IP laws from examining and testing the surface of the beakers for contaminants and publishing a description which allowed other researchers to make similar modifications.
It's the most sub-optimal situation in which to obtain something which starts to approximate reality. Not impossible, just difficult, and likely to lead to shenanigans.
So Google doesn't have the resources to test the default driver on the most widespread linux distribution for one of the two most widespread GPU producers in the market.
> The default driver on the distribution we support is broken.
They do test the default driver on the distributions they support. The tests show that the driver is unacceptably unstable. It may work better on other distros/versions, but they don't have time to test those. They're happy for others to help out in testing these unsupported combinations, however:
> Again, if someone wants to spend the time to test thoroughly to narrow down the blacklist, we will accept patches.
And in any case the reason nouveau sucks (and yes, folks, nouveau sucks) isn't Google's fault to begin with.
You seriously want to complain about Google's lack of support for an open source driver that barely exists because of NVIDIA's lack of support?
It's crazy to suggest they should give web access (not just webgl, but CSS styling) to a buggy GPU driver by default without an opt in by the user.
(Not bring cynical, I genuinely don't know.)
From reading more about it, it seems these are all benign (in the sense of Chromium couldn't ship h264 because it's licensed but Chrome can).
This just about sums up the problem with open source. The response is always "if you don't like it, fix it yourself". Because everybody is a developer with copious free time to learn how to fix the problem in their favorite product. It's not like cloning and building Chromium takes hours. Or like you would lose anything by using Chromium instead of Chrome.
But that's the point. The response always works because it's proportional to the problem. For problems that are legitimately small and you could fix yourself, or pay someone to do it for an amount of money that a normal person could feasibly pay, it's a valid way to solve the problem.
And for problems bigger than that, it invites the user to consider what they're really asking for and who they're asking for it. Actually fixing nVidia's dumpster fire would be a huge ordeal for anyone other than nVidia. Asking a third party to do it without documentation... let's just say there is a reason nouveau is in the state that it's in, and it's not a general lack of interest in fixing it.
So in that case "go fix it yourself" means "if you think it's so easy then let's see you do it."
It takes copious free time to learn how to run chromium with an --ignore-gpu-blacklist argument? What kind of strawman are you trying here?
Look. The open source NVIDIA driver is terrible, and Google doesn't want it messing up the experience of the people using its product. Who are you to demand that Google fix someone else's terrible driver (which is terrible through no fault of Google's -- NVIDIA would prefer it not exist at all)?
If you really cared about this at all, you'd be upset with NVIDIA. But you're not. You apparently don't even like open source software, which presumably means you're using the binary drivers and unaffected, right?
The driver is hopeless? Fine.
You don't want to spend money on people that will most probably install an ad blocker? Go ahead.
You don't have the resources? Come on.
It's not as simple as "one of the two most widespread GPU producers in the market".
Chrome team, on the other hand, does not have the resources. Which is to say that this is not the thing they believe is most important to work on at the margin, and given the amount of engineers Google will fund for the project.
If Nvidia wants their cards to work well on an all-free-software stack, they can provide a free software driver. If users care about making an unofficial driver work, they can do the work or fund the work. Chrome is just trying to get web pages on the screen and it doesn't seem like it's worth the boil-the-ocean level of effort of taking responsibility for implementing a reverse-engineered graphics driver when they can just get web pages on the screen using software rendering.
Something doesn't add up to me there, and no one that I saw ever addressed that in more detail.
they are also saying their that driver may work fine on some less common linux configurations, and the chrome team does not have the resources to test those.
you are literally saying they are saying the opposite of what they said, then blaming them for it.
The common distribution was blacklisted not because it was not tested, but because it clearly has issues observed and reported by the users.
If a driver does not work in popular configs, and needs to be blacklisted in those configs, your solution is to still whitelist it in other configs and have the browser likely not work.
yes, "every company" does exactly what google did. you are being purposely dense, trying to catch a "technically correct" chance of not admitting the stupidity of your question. The problem is, you are not even technically correct - just purposely dense.
And the test suite failed on a minor issue where the cause was uncertain, which gets in the way of checking for the bigger definitely-the-driver problem.
I solved the Chrome/Chromium on Linux problem long ago by switching to Firefox. I'll still fire up Chromium for Google sites and a handful of other sites, but for the most part Firefox has been getting it done for me.
It's an Ubuntu/Nvidia problem, not a Google problem. AMD and Intel open source drivers work just fine.
That's still less than a percent of the market I think. Maybe shave off another order of magnitude. So the economic argument is not the one to make. However, it sucks nonetheless.
Honestly, you could fill up the number of people using nouveau with chromium in a stadium more likely. From the small pie of linux users, you drill down to nvidia, and then further down to nouveau. All laptop users, even with nvidia cards, will only use the intel gpu unless they explicitly enable the discrete mode and not install the nvidia drivers. Desktop users obviously only buy nvidia if they want to use the binary driver. What do you think remains of the pie ? I haven't met a single person in real life who uses this stack. I love nouveau. That doesn't change the numbers though.
To save those users from themselves, Google now disabled GPU accelleration for all Ubuntu users.
Seems like a bit silly to me.
I don't get this. All linux distros under the sun ship with nouveau, and always have. It is a kernel driver afterall. What else are they going to ship ?
The NVIDIA driver is the driver that yields the best performance and experience; AFAICT the only reason it's not shipped by default is because the kernel is then considered "impure" or something (because source isn't available to debug issues in the NVIDIA driver), hence bugs can't be filed upstream for any issues in the kernel if the NVIDIA driver is installed.
However, none of this is really the concern of the Chrome team. Their responsibility is to give the user the best experience, which they have defined as security, stability, and performance in that order. Since the nouveau driver is causing stability bugs, it's perfectly reasonable for them to fallback to software rendering and take the performance hit in order to preserve stability.
The term is "tainted".
Including Nvidia's binary driver by default. Obvs Debian, gNewSense, and GUIX can't do that, but since when have corps cared about them?
Like, don't get me wrong, I'm a FOSS or GTFO kind of guy, running Debian and Replicant, but shipping proprietary drivers by default is what chrome is hinting at obviously.
The combining of proprietary and GPL code has to be made by the user and the result is un-distributable. That's why distributions cannot ship it.
What they can do is to make the above step easy (which Ubuntu does).
AFAIU, Nvidia's legal argument is that their driver can't be 'derived from' the Linux kernel even though it's linked into the kernel since it's older than their Linux port and only is linked against their GPLed (and if push comes to shove, dual licensed GPL and proprietary) kernel abstraction layer. Judges don't care about linkers and the GPL doesn't actually say anything about linking.
This is about GPL infrigment; you can combine GPL licensed code with differently licensed code, but only in your privacy, without distributing the result to others. Once you would distribute the result, you would be breaking GPL, thus having no right to distribute.
> The "Program", below, refers to any such program or work, and a "work based on the Program" means either the Program or any derivative work under copyright law: that is to say, a work containing the Program or a portion of it, either verbatim or with modifications and/or translated into another language.
The argument that I've is that even when linked together, the Nvidia binary driver isn't a 'derivative under copyright law' because it doesn't have Linux specific entry points, and is older than their Linux port, and therefore isn't part of "The 'Program'" as listed in the GPL.
Your argument is, why some piece of software should (or should not) be relicensed to GPL.
My argument is, that since that piece of software is currently not GPL licensed; avoids the relicensing topics and explores how to solve the current situation wrt distribution of the combined work. Use by the end-user is non-issue, due to freedom 0.
And second, to answer your question, partly because Linux is sufficiently good and has wide enough hardware support that it's still easier to use Linux legally under the GPL than to use another operating system, and partly because if you're going to use another operating system with narrower hardware support you might as well use a tinier embedded one and ship cheaper hardware (which is what Linksys/Cisco did with the WRT54G, shipping VxWorks and less RAM, and then selling the more-Linux-capable WRT54GL as a higher-end product).
AFAIK it works with Gnome's Wayland Implementation, but not with the one from KDE.
Chances are, that you are using modesetting X driver anyway (i.e. the KMS driver, the same driver/backed, that is used by Wayland).
Otherwise, when using it in KVM VM, it can run the desktop compositor quite nicely.
Then are Nouveau users only losing performance and not functionality as a result of this Google decision?
That's not the point. The point is that it results in brokenness, and the easiest way to mitigate that for the bulk of non-technical users is to use software rendering instead (i.e. blacklist it). They're open to more refined solutions, but these will take some effort.
The composition is GPU accelerated too. Chrome has it enabled by default, Firefox doesn't.
In the end, both browsers treat Linux as a second-class citizen. To make things even more weird, ChromeOS uses the same API (VA-API) and the same driver (for Intel GPUs) for video decoding acceleration that the stock Linux does, and the video decoding is supported on ChromeOS but not on Linux... And Firefox is just plain mess.
I'm facing this problem and it's very annoying. I have to run Chrome with --disable-gpu-driver-bug-workarounds, which ironically makes it work perfectly---with the "driver bug workarounds", it's terribly broken. But now, any Electron app I run (e.g. VSCode, Postman, Discord, etc.) is also broken by default and needs the flag as well. Frustrating.
> Please remove the blacklist as all Nouveau users expect this corrupted behaviour. If at all possible it may be worthwhile to let users know it doesn’t have to be this bad if they flip to the binary driver. Perhaps a pop up letting them know?
Let's say 50 options. If it costs $1 to test each combo it would cost 2^50, around $10^15 dollars. Google only made $10^9 dollars profit last year.
Now of course I'm being facetious but do not underestimate the cost of testing.
And still, Mozilla manages to do a better job in this regard.
The current link says it as if Chrome was not being fair, but reading their reasoning it makes total sense. Key quote from said bug report:
> We thought about blacklisting nouveau driver long time ago, and decided to let more adventurous users to play around. Now that Ubuntu ships with nouveau on default, maybe it's time to blacklist it in Chrome.
> We received quite some bug reports on other rendering issues with Nouveau, and this bug is just one of them. So we will disable all GPU acceleration by default. If someone wants to bypass that, there are two options: 1) install proprietary NVidia drivers (see #24) 2) run Chrome with --ignore-gpu-blacklist, but be aware you are taking a risk here
> Unfortunately we don't have the resources to test every variation of every GPU/driver combination on linux, let alone investigate & fix bugs in drivers. We want a stable & secure browser first, a GPU-accelerated one second, only if possible.
> The default driver on Ubuntu LTS has severe issues, asking non-technical users to update their driver is just not acceptable as a prerequisite to use Chrome.
> If someone is interested in well-scoping the brokenness (version range and/or devices affected), we're happy to take a patch to the blacklist.
So the driver fails in some (common enough) cases, they let it run anyway but now it's default on ubuntu so a lot of non tech users are/will be affected, so they block it but you can still bypass the block if you want. And if newer version fix it and someone can help them figure out which versions work for what, they are willing to limit the blocking.