Nvidia might stop being a douchebag about Linux: https://developer.nvidia.com/blog/... Right now, that doesn't change much yet, but it opens a possible road.

We will see how it pans out.

Right now I'm not entirely convinced.

Looked briefly over the repository and the pull requests, it seems to be friendly.

I really dunno what GPU vendor is hot at the moment and leading the market, but Nvidia was always a striking red nope cannon for me...

I think it is a _tiny_ step in the right direction, but given the various promises NVidia made and later broke I highly doubt it will work out.

NVidia has a "magna cum laude" in "pissing people off to the point of no recovery"...

@IntrusionCM I blacklisted Nvidia even for pure Windows machines just to keep a proper potential Linux road, and one swallow does not make a summer.

However, while I certainly wouldn't reward Nvidia by buying anything from them right now, it might be a sign to keep an eye on.

@zlice They might have understood, an eternity after Intel and AMD, that OSS as add-on to selling hardware is one of the OSS business models that actually works. They also might finally have taken notice that Windows gets hacked although it's CS.

@zlice there is 20gb rar file code dump torrent from lapsus but I didn’t go trough it yet, I know there’s drivers folder inside so maybe the code is there but it’s impossible to find if you don’t know what you’re looking for

The keys to enable proper power management to be implemented in Nouveau or it didn't happen. They fucked the Linux community before and there never was a drawback to releasing proper specs and allowing open source driver development. So whatever reason they had to DRM their GPUs probably still exist.

@vane It is the power management/reclocking stuff you would look for - they purposefully included a signature check to prevent Nouveau from using the card's full potential.

@Oktokolo They won't open source everything, just like AMD also relies on binary firmware blobs. However, it could at least eliminate the dependency on Nvidia's good will to compile the driver for every kernel version.

Obviously, they won't open source CUDA/ML because AMD has totally missed the future and is still babbling about catching up in gaming. It's not in Nvidia's interest to give AMD an easy way out of AMD's own stupidity.

@Fast-Nop Not requiring firmware to be signed by nVidia would be enough...

https://nouveau.freedesktop.org/: "Little hope of reclocking becoming available for GM20x and newer GPUs as firmware now needs to be signed by NVIDIA to have the necessary access."

@Oktokolo Wouldn't this problem solve itself now? The firmware isn't OSS anyway, neither with AMD, so you'd have to use the provided blobs either way. Only that now, the new OSS driver can actually use Nvidia blobs without running into copyright issues.

Btw, if you want to know why the Nvidia drivers cause disproportionately many crashes, look at their Github issues: it's because they don't even use any sort of static code checker such as CppCheck. W-T-F.

@Fast-Nop I looked briefly at their drivers repo and folder structure meats the leak so I think they’re just cutting cancer with that and don’t want to do it but forced by se email

fucking corporations

@Fast-Nop Nouveau could have required the user to run a script to download the blob from the nVidia site. So i don't expect it to be just a legal issue.

@Oktokolo @Fast-Nop tinfoil-hat hypothesis here: could they be so afraid of releasing OSS drivers because then people would have proof that they have been throttling GPU cycles to avoid overheat?
It could have been done VolksWagen-style - avoid throttling when a benchmark is detected, but do it on real workloads. And hide under-rendered blob collisions in games under oversaturated lens flare or sparks. Most people wouldn't notice.
It would be coherent with blocking Nouveau's full access to power management.
Dunno, might be paranoia talking. Just saw "the dropout".

@JsonBoa Throttling for avoiding death by overheating is something i would see as a bug if a GPU or CPU _does_not_ do it. Of course GPUs throttle when they get too hot.

And of course GPU drivers detect benchmarks and reduce quality settings in a subtle way to get a better performance score. It really isn't a secret.

Also, they don't need to open-source their driver and firmware. Just open-sourcing the data needed to write and run a custom firmware on their GPUs would actually be enough. The Linux community is not only capable of writing drivers and firmware - i would even go so far to assume them being better at it even on nVidia hardware than nVidea themselves.

@JsonBoa Nearly, but not quite. As pointed out, overheat protection is a necessary feature.

I'd think more in the direction of product line diversification via software only. Maybe also some DRM concerns for movie content.

@Oktokolo it's always funny when someone says "it's not a conspiracy, those practices are shady business as usual"
@Fast-Nop 's explanation is even shadier, though. It would validate those people who ask on the internet how to download an Intel core i7. But would also make a lot of corporate sense.

@JsonBoa That's common industry practice. Remember that hurricane in the US where Tesla used a remote update to enable more battery range for the cars so that people could get out? That implies that the range physically must have been there all the time.

Or the stories about AMD Phenoms where you could enable more cores because AMD didn't have enough CPUs with actually broken cores, but neither could sell enough in higher price tiers.

Or the B450 boards where some intermittent BIOS versions suddenly allowed PCIe 4.0 until AMD stepped in with later updates as to keep that exclusive to the higher priced B550 boards.

@Fast-Nop "Introducing AWS BadPlace, our proprietary solution for your underworldly needs. Compatible with the opensource hell-2.0 API.
Torture more, with AWS BadPlace!"

This fucking world, dude.

@Fast-Nop The Tesla battery thing might just have been an adjustment to the discharge safety margin meant to improve battery life. But when you can choose between a day shorter battery life and being killed by a flying cow today - most are probably happy to choose shorter battery life.

In my opinion, the user should be allowed to manually set the lower discharge safety margin himself. If you are allowed to drive, you should also be assumed to be able to decide between more range and longer battery life.

It's a bit of a tug of war. It'll depend on whether or not Nvidia makes obscenely large amounts of money and how much it stands to benefit from the likes of Microsoft.