ngz's recent activity
-
Comment on How bad are Nvidia GPUs for Linux really? in ~comp
-
Comment on How bad are Nvidia GPUs for Linux really? in ~comp
ngz My position isn’t “Wayland devs should be forced to support Nvidia.” As far as I’m concerned, people should be able to work on developing whatever they’re interested in and/or are paid to work on....My position isn’t “Wayland devs should be forced to support Nvidia.” As far as I’m concerned, people should be able to work on developing whatever they’re interested in and/or are paid to work on.
What I’m saying is that given that they’ve made the choice not to provide that support, claiming that Wayland as an ecosystem is ready for the majority of users is a gross overstatement at best.
It’d be like if I developed an alternative browser, refused to support any machines with fewer than 30 physical cores, and then claimed it was the future and everyone should switch to it. The software might be feature-complete on its supported platforms, but if you explicitly refuse to support a large fraction (if not the majority) of the people you’re suggesting try your software you can’t then claim surprise when they don’t adopt it.
-
Comment on How bad are Nvidia GPUs for Linux really? in ~comp
ngz I’m definitely aware of the background here, but I was specifically referring to the “we don’t support Nvidia, and if you want to use Wayland you should just buy a new GPU, ignoring the fact that...I’m definitely aware of the background here, but I was specifically referring to the “we don’t support Nvidia, and if you want to use Wayland you should just buy a new GPU, ignoring the fact that AMD cards are not viable for a lot of usage scenarios” position of Sway and other Wayland projects. It’s dismissive at best and actively hostile at worst to simultaneously claim that:
- X is legacy and nobody should be using it anymore, as Wayland is ready for almost all use-cases
- Nvidia cards (80%+ of the desktop GPU market, IIRC) are unsupported in many Wayland projects
I get that it’s not Wayland’s fault, but it’s also frustrating to see it claimed as ready for use when in reality it doesn’t have good support for common use cases (GPU compute) or the majority of discrete GPUs. If I’m suggesting someone migrate to Linux, it’s unrealistic to ask them to drop several hundred dollars on a new GPU just because the Wayland community doesn’t like the perfectly-functional card they already have.
-
Comment on How bad are Nvidia GPUs for Linux really? in ~comp
ngz I haven’t really had many issues with the proprietary Nvidia driver (using a 3090 and 2080Ti in the same machine), aside from the periodic attempts to try Wayland that invariably fail utterly due...I haven’t really had many issues with the proprietary Nvidia driver (using a 3090 and 2080Ti in the same machine), aside from the periodic attempts to try Wayland that invariably fail utterly due to stuff being broken and not having the time to go debug it. There’s some occasional weirdness but on the whole it’s been pretty stable. If you’ll mostly be playing games, chances are the Nvidia drivers will work fine - just note that some newly-released stuff might not work until Proton has a chance to patch things, and some Wayland stuff is really hostile to Nvidia users. For example, it took a few months for Starfield to be playable on an Nvidia GPU after release. Older stuff generally works fine (or if it doesn’t, it’s not the Nvidia driver that’s the issue).
That said, it’s not like I have any other option. If you want to do compute work on your GPU(s), you’re essentially forced to go with Nvidia. OpenCL is a joke, the ecosystem around Vulkan compute shaders isn't really usable, and ROCm doesn’t have broad compatibility so you have to be very careful to make sure you buy a card that supports it. Whereas CUDA works on all Nvidia cards, most distros, has a ton of good libraries and tooling, and is supported by most pre-written applications. I wouldn’t mind trying an AMD card because open graphics drivers are a good idea, but that’s not really an option until the GPU compute story improves.
Also, since some other people have mentioned it, this is all based on desktop hardware. I think there was some weirdness around hybrid graphics on laptops, but my knowledge there is almost a decade out of date so there’s a good chance that’s been fixed by now.
-
Comment on YouTube is testing a three-strikes policy for ad blocking in ~tech
ngz I can’t speak for anyone else, but a big part of the issue is for me is around how Youtube treats many of the content creators I like. The continuing issues around demonetization and deranking of...I can’t speak for anyone else, but a big part of the issue is for me is around how Youtube treats many of the content creators I like.
The continuing issues around demonetization and deranking of relatively innocuous videos mean I have zero faith that subscription money would actually go to support content creators. I’m happy to support the content, but “paying Google” is distinct from that.
-
Comment on Permanent archival formats. Do they exist? in ~tech
ngz Just be aware that read/write speeds are extremely slow. I managed to get an LTO5 drive for cheap recently, and that’s one of the biggest things that I’ve run into as I’m trying to integrate it...Just be aware that read/write speeds are extremely slow. I managed to get an LTO5 drive for cheap recently, and that’s one of the biggest things that I’ve run into as I’m trying to integrate it into my backup strategy.
-
Comment on Dijkstra’s In Disguise [2018] in ~comp
ngz Hello! I’ve been lurking for a while during the Reddit exodus, got my invite a few days ago, and have been thinking since then about what would be a good first post (i.e. something substantive and...Hello! I’ve been lurking for a while during the Reddit exodus, got my invite a few days ago, and have been thinking since then about what would be a good first post (i.e. something substantive and interesting, rather than “hi, I’m new”). There’s a good chance I’m overthinking it a bit, but still.
Anyway, as far as I can tell this wasn’t posted here when it was first published - it’s one of my all-time favorite technical blog posts and does a great job explaining some seriously non-trivial ideas, so I figured it’d be a good first contribution.
To be specific, this gave me a new appreciation for how much algorithms concepts can connect and generalize across different applications. It seems obvious in retrospect, but I’m pretty sure that’s just hindsight bias since I distinctly remember sitting in front of the computer blown away and taking several minutes working through that realization when I first came across this.
-
Dijkstra’s In Disguise [2018]
9 votes
That… doesn’t actually change the dynamic I’m talking about at all? It doesn’t matter if it’s X12 or not - it’s about the framing and dynamic. Supported use cases might be fine (debatable, though that’s not relevant to the point), but the issue is that it’s not being framed as the future of the Linux desktop for non-Nvidia users - the qualifier is missing.
For a less Linux-world-focused example, consider Microsoft dropping support in Windows 11 for machines that don’t have TPMs - a large fraction of their target user base. They’ve rightly been criticized for it, and it’s slowing adoption because people have perfectly functional machines they already own. “Tough luck, we’re not supporting half our market anymore, better buy new hardware” is equally bad there as it is in Wayland’s case. People are upset with them about it, for the same reasons.
You could argue that users should be holding Nvidia accountable, and I can see the reasoning there. But you can’t exactly blame the users for resenting and disliking the Wayland devs using them as a political weapon against Nvidia when all they want is to be able to use their machines.