Hacker News Reader

When nVidia got what it wanted from Linux by dominating compute space using CUDA, they've just left visual part of their drivers to rot.

Moreover, they stay silent in the forums now. They don't listen to customer feedback or at least give any signs of hope or assurance. Their company image became too dominant and arrogant (I think it runs more than skin-deep).

Some of the nVidia's huge market share is coming from the momentum they have. Linux users do not refresh their hardware too frequently generally.

I'm running a GTX680 on my primary Linux box, but it won't be nVidia next time. I'll be returning back to red team and I'd be a happy camper.

> When nVidia got what it wanted from Linux by dominating compute space using CUDA, they've just left visual part of their drivers to rot.

The flip side is that if you want to do GPGPU work (even if it's just local testing) you more-or-less need an nVidia card.

You're right but, I'm deeply interested what AMD can do now or what's in the works.

ATI was left in the dust while the foundation has been forming, unfortunately, but the history is not that straightforward.

TBH, while I appreciate what nVidia is doing, I don't appreciate how did they came to this point.

I have some close friends who're doing game engine and very low level 3D development. His thoughts about nVidia's driver behavior, even under Windows, is worth another rant.

> When nVidia got what it wanted from Linux by dominating compute space using CUDA, they've just left visual part of their drivers to rot.

Have they ever taken proper care of the visual part? I've thought that it exists in the first place because of the SFX industry's heavy use of GNU/Linux and Nvidia. That's a fraction of their market share so only got a fraction of their attention, likely mainly to avoid vacating the area for competitors than to actually make money (bad ROI compared to Windows).

>> bad ROI compared to Windows

Windows is essentially just the gaming market. Gamers are fickle, only buying a card every couple years and are very price conscious. But AI, and any other datacenter uses of GPUs (mining) buy cards in bulk, are not fickle, and are not obsessed with daily price fluctuations. Those are the markets Nvidea wants to be in and they are dominated by linux. Not playing nice with linux gamers will come back to haunt Nvidea.

Also, one should not underestimate the bias that gaming generates. All those linux gamers who suffer poor drivers while gaming saturday night walk into jobs at datacenters and AI research groups each monday. They remember. I certainly do. All things being equal, I lean towards AMD whenever the boss asks my opinion on our next purchase.

It's much more complicated. The application diversity is much higher than it looks.

There are also Quadro cards aimed at professional OpenGL acceleration for GIS, 3D modeling, CAD, etc. These applications run on both Windows and Linux (soem are cross platform, some are platform locked) and, the ROI pie is a complicated picture.

Windows is dominated by gamers and casual users, that's right, however it's not the sole market for gaming. Steam is pushing for Linux in a relentless pace. Steam in Linux is getting rooted like a Sequoia tree. When it becomes visible, its roots will be much deeper and stronger than the trunk (which is a good thing).

nVidia used to care for Linux users. Extensions required by composition, off-screen rendering and other stuff came to nVidia first (when compared to AMD). Video acceleration is also the same (AMD's video acceleration technologies can were much better than nVidia due to their legacy but, they acted slow. They're even now. Maybe AMD is still better. I need to test).

This trend reared its ugly head ~3 years ago for me. I've upgraded my monitor to a DP one and DPMS went out of window. Screen suspension and standby has gone. They didn't fix it for older cards for ~1 year IIRC. Currently I'm experiencing a strange frame-rate drop. Will investigate when I have time. If an onboard Intel card can drive same DE (KDE) on a same resolution screen better with a much inferior CPU when compared to a triple-slot, out of the box overclocked GTX680, then there's something very, very wrong.

nVidia was top of the line, just works out of the box GPU vendor. Now due to its arrogance and dominance, cracks start to appear. Even Intel's driver blunders are solved faster (ah, they're open source).

> Linux users do not refresh their hardware too frequently generally.

> I'm running a GTX680 on my primary Linux box, but it won't be nVidia next time.

Well, you're kind of making their point. Your business is probably insignificant for them, you have an 8 year old GPU :-)

Just because I'm changing my GPU once a decade, it doesn't mean that I don't access to much bigger hardware pools and use their newer hardware for other purposes.

> When nVidia got what it wanted from Linux by dominating compute space using CUDA, they've just left visual part of their drivers to rot.

In their defense, from their point of view, gaming on Linux is a tiny fraction of their overall business and, presumably, even their Linux one. For them, Linux is the same as HPC and ML model training and, maybe, cryptocurrency mining.

Just don't forget scientific and engineering applications which make use of their Quadro cards, under Linux.

Can we call it "small-scale HPC" or "single-node HPC"? ;-)

Nope. HPC is not interactive in most cases. The calculations are done unattended.

Quadro cards are used for real-time rendering of ultra-high resolution models for post-compute analysis or pre-prod/pre-analysis design work.

Quadro cards are not more useful in CUDA computations when compared to the Tesla cards.

They're different beasts for different scenarios.

Disclaimer: I'm an HPC administrator.

That's true. I remember the SGIs and Suns tied to their larger siblings where the rack-sized behemoth did the number crunching and the desktop workstation did the visualization of the data. I didn't suspect Linux would still be relevant in this role - I see lots of Windows boxes doing that, even when the app is presenting itself through X (something I find quite amusing).

Nvidia spends several times as much as AMD on software. If they have that much more money, why is AMD able to ship more stable drivers? I just don't see how this holds any weight.

These companies aren't rewriting tens or hundreds of millions of lines of code for Linux. The difference is probably a fraction of a fraction of a percent.

I was very involved in this driver subject for some years. The issue is more complicated than it looks.

Closed source GPU drivers are not ordinary drivers. They have a lot of quirks and workarounds for programs, especially games.

When a dev studio makes some mistake in its engine or the engine doesn't fit well to current workflow of the driver, a new driver is shipped with specific workarounds for this game. These quirks generally add shims between the program and the driver by basically modifying calls to the driver to increase performance.

e.g.:

- A game is calling too many clears while it shouldn't? Pool them in the shim layer and let them through in batches of 1000.

- Is a particular benchmark is overloading the card by forcibly drawing off screen portions and hurting scores? Insert a shim to prevent drawing off-screen parts and chea... ehrm, optimize the performance.

- Is an important legacy application exploiting a quirk in an older driver for something? Introduce that quirk via another shim.

When you try to port these drivers as-is to another platforms, things get awry. Also the drivers were tremendously Windows specific. This is why FGLRX had (or does it still have) various daemons, HAL layers, during its teething years.

OTOH, free/OSS drivers has no shims, workarounds and other stuff. They are just drivers. Plain, simple, vanilla. They are easier to maintain and develop. Radeon driver is such a driver. Also AMD/ATI has some more tricks up their sleeve for more stable, better drivers regardless of their platforms.

- IP separation: All of the NDA bound parts of the GPU (currently it should be the HDCP stuff only) is a different IP module on the GPU. The cards are designed so that even the Radeon driver can work with HDCP module while being completely blob free. ATI/AMD specially altered silicon and GPU architecture to enable FOSS drivers on Linux and I think its effects are huge.

- "Factory Team:" Radeon driver is carried forward by volunteers but AMD also has a dedicated team working on Radeon driver full time. This group cannot communicate with FGLRX team to prevent tainting each-others wells. So, Radeon is blessed by AMD/ATI on hardware, software and management level.

When you publish your specs, dogfood it with another team, and optimize your GPU silicon and board to accommodate these open source exceptionally well, you have a killer combination.

> Closed source GPU drivers are not ordinary drivers. They have a lot of quirks and workarounds for programs, especially games.

The quirks part is a bit of a meme. Mesa contains a few application-specific overrides here and there: https://github.com/mesa3d/mesa/blob/c62996796cc46f1a1406f365..., https://github.com/mesa3d/mesa/blob/a678ec9b8c057311ed7e9697.... The reason Linux drivers, including the one from Nvidia, have fewer quirks is simply that Linux has fewer 3d applications overall. Those quirks also reside in userspace driver parts and should not affect system stability.

> When you try to port these drivers as-is to another platforms, things get awry.

Again, quirks are a property of a relatively high-level userspace part of the driver. There is nothing hazardous about porting them.

> why is AMD able to ship more stable drivers?

Well, they don't — AMD drivers for new hardware famously take a few kernel releases to stabilize. (G)GP specifically noted their stability problem started only 5 months ago, which means at least that issue is not really systemic. The real differentiator here is handling of existing complaints: Linux has mailing lists, Mesa has a normal open for everyone bug tracker, while Nvidia has a user support forum, which they mostly ignore.

>> Linux users do not refresh their hardware too frequently generally.

Linux users don't have to refresh hardware as often. Long-term support is a cornerstone of linux. Linux gamers may spend less per year on gaming hardware but they are not broke. Give them a reason and they will drop that banked cash on anything they can trust to play nice with linux.

> Linux users don't have to refresh hardware as often. Long-term support is a cornerstone of linux.

I'm well aware of that (since I'm one of them too), however I didn't make it clear I presume.

Thanks for highlighting it again.

It's also stems for efficient utilization of the hardware, too.

When I was younger, I thought there was a choice between picking free software for ideological reasons, and picking whatever software did the job for pragmatic reasons. But I have mostly come to conclude over time is that this is a false choice; over a majority of use cases and time scales, open source is just pragmatic choice anyways.

In times past, open source was a much harder sell. You pretty much had to do things the Unix way or deal with a very limited program that mimicked its Windows counterpart. That has changed.

I would argue that open source can be more pragmatic these days. A lot of software offers the features people need. Managing a system based upon open source software is significantly easier since there is no reason to impose restrictions on use or redistribution. To choose an example relevant in this case, open source drivers can ship with a Linux distribution so you have functional drivers the moment you boot from the installation medium. In the case of proprietary drivers, there is typically an extra step once the installation is complete.

Personally I would love to switch to OSS tomorrow, but I still see many showstoppers:

- Desktop: Linux vs. Windows - worse support for gaming (e.g. anti-cheat), no support for major creative software like Photoshop/Affinity or Ableton/FL Studio

- Mobile: FOSS Android distro vs. Google Android - no guarantee that apps I currently use will work without Gapps, no paid apps or in-app purchases

I have really been thinking a lot about this in the past months, because I would like to improve my privacy. The only real play seems to be moving to Apple's ecosystem, and buying a console for the games (and paying a lot more and losing customizability and repairability in the process).

The event which persuaded me that open source software was the pragmatic choice happened in 1999 (though it took a few years afterwards for me to articulate my advocacy in those terms). That's when the Gibson Guitars, having acquired Opcode Systems previously, abandoned development on Opcode's flagship sequencer application, Studio Vision — leaving its users high and dry.

Many music users today are locked into proprietary applications analogous to Studio Vision and their proprietary formats. If the company goes under, or if management decides to kill the product, there goes all your work.

Theoretically proprietary software can support open file formats — but there are such strong incentives for companies to leverage network effects and lock in users that at any moment you are one upgrade away from a policy change and silent conversion of your document to a proprietary format. Only FOSS guarantees that your documents will survive the tumult of capitalism.

Two decades later, history has repeated itself, with Flash, Macromedia, and Adobe in the roles previously played by Studio Vision, Opcode, and Gibson...

I have the opposite experience. AMD is effectively unusable for AI. As much as I'd like to switch, I need those NVIDIA drivers because CUDA is the only thing that works reliably. So it's AMD CPU + NVIDIA GPU + Ubuntu sigh

I never had crashes but, having switched from a 1070 (proprietary) to a 6800 (open) this week, the experience is far better. Things that were previously nigh-on impossible to get right, like fixing screen tearing, or freesync without bizarre performance penalties, just work now.

I keep flip-flopping between these as my needs change. I flipped to proprietary last time I tried some Steam games. I will gladly switch back to open source drivers if I'm wrong in thinking game performance on linux will suffer?

probably a whole bunch of depends, but I haven't noticed any performance problems with my 6800xt in linux.

Partly this will be because I'm not playing games like CS:GO not cyperpunk, but I'm capped on 144hz with latest OSS drivers.

Not played anything very intensive yet, but Rocket League and Deep Rock Galactic, each via Proton, are performing way better than they were before, in line with what you'd expect moving from a 1070 to a 6800.

For real. Getting the NVIDIA drivers to work with newer kernels has been a total nightmare.

I've had to test out several combinations before I got a relatively stable config.

Now I'm afraid to apply an automatic update since I don't want to lose several hours to debugging BS driver issues.

>> a relatively stable config.

That's where I was on my last rig (AMD+nvidea). Every driver update was a dice roll. Am I going to play today, or spend three hours googling to figure out why my machine keeps booting into safe mode? Since I switched to AMD+AMD a few months ago: no issues. Maybe nvidea would give me a few more FPS at the same price, but you cannot put a price on "it just works".

Even in windows, nvidia drivers have been bad. One particularly nasty example is the SteamVR frame drop bug they introduced last... May? Still isn't fixed, though they finally posted they found the cause and the next patch will supposedly have a fix.

They're excellent on Windows, which is their main target sadly. I wish AMD did it better.

> They're excellent on Windows

They are 'OK', I wouldn't say they're excellent. I have an RTX 3080 (prior to that I had a 1080Ti) and games crash more often than I would like. Waking from sleep is hit or miss, random driver crashes where the monitor goes blank, driver reinitializes itself, screen comes back on, etc. This is all with the latest and greatest drivers, 4K 160hz monitor (and an older 4k/60hz monitor), etc.

No kidding. Few weeks ago 'apt upgrade' killed my X server and locked up the whole machine. I'm like 99.99999% sure it was the nvidia drivers.

Yes I used to bemoan ubuntu upgrades for not working.

Turns out that it's the NVidia drivers every time.

Even on Windows, the driver batch of 2017 tanked my card performances to the point even old games I used to play on max were stuttering. I hold on the 36x series as long as games let me, but came late 2018 new release started complaining and then stopped starting due the old driver

So I jumped on AMD, missed the rtx train, but my card is still running strong in 2021 and I'm not in a hurry to switch; I'll wait until the release of a midrange ready tracing card from them.

Isn't the benefit of open source that you can just revert back to the 5 month old drivers and wait out the instability?

> I've experienced numerous crashes with my GTX 2080 Super card.

Same with my trusty 1070. The WHQL driver from 2019 didn't crash, but it is now too old for current games.

Do you have links to the actual threads in which these issues are reported ?

And to add insult to injury, the nvidia drivers are what, 500MB? and require you to open an nVidia account to download.

GTFO with that BS.

You can just download them...

The GeForce experience requires an account but you can download the drivers without it. Still obnoxious and clearly confusing to the end user.

You can still manually download them here, without an account: https://www.nvidia.com/en-us/geforce/drivers/

My guess is the huge file size is due to the GeForce Experience tool that's included. It would be much better if it was a separate download, but you can choose not to install it during the setup process.