Hacker News Reader

Comment by 6gvONxR4sf7o | Hacker News Reader

Wow, so he writes an article about people being assholes about him not writing free extra stuff, and this other article saying literally “nvidia- fuck you” (and its “shitty consumers”) for not writing this other free extra stuff. He says that’s treating Linux users like shit, when a massive part of nvidia’s business is supporting Linux (GPGPU stuff), which works fantastically.

Reading OP, I figured it was righteously angry. Reading your article, I’m now figuring he’s just angry and abrasive. The OP article reads hypocritical after reading yours.

View on HN

Author here.

>so he writes an article about people being assholes about him not writing free extra stuff

I did write that extra free stuff, and I'm arguing that some people are being assholes because they perpetuate the lie that I didn't do the free work, and because feature "foo" doesn't work, project bar must be bad. But feature "foo" does work, and using that falsehood to say project "bar" is bad is a dick move. This is the "horseshit" that's pissing me off.

The Nvidia issue is different: Nvidia goes out of their way to prevent their hardware from being used for this purpose. For example, they use signed firmwares that we simply cannot use for implementing the features we need, relying on cryptography to lock us out of their platform unless we use their driver. Then that driver doesn't support what we need. Nvidia deliberately prevents us from putting in the effort. They have created an artificial monpoloy on the required labor, then refuse to do it.

You and I are reading the criticism differently. There’s genuine assholery, 100%. You’re entirely right that those people just make everything worse, and I’m sorry let you have to deal with them.

The other side of things is what I see differently. When folks are saying wayland isn’t ready, or even that it sucks, I mostly see them as referring to the ecosystem, not wayland itself. It’s an unfortunate quirk of language. Because foo doesn’t work, the bar ecosystem is currently bad.

On the nvidia side, I’m not going to engage further with you on that because you called me a shitty consumer and that pisses me off. Fuck me, you don’t want me in your userbase anyways, right?

> when a massive part of nvidia’s business is supporting Linux (GPGPU stuff)

I've spent the last few months doing GPGPU stuff on Nvidia Jetson devices and my experience has been the complete opposite. I've rarely seen something with such bad support and documentation.

I'd love to hear more. I've only used the datacenter and flagship consumer GPUs, which work great. I hadn't heard of Jetson, but looking it up, it's a bummer to hear that something that looks that cool for hobbies doesn't hold up.

Well, don't get me wrong. I do like my Jetson Nano. For a hobbyist who likes to tinker with machine learning in their spare time it's definitely a product cool and there are quite a few repositories on Github[0, 1] with sample code.

Unfortunately… I'm not a hobbyist, I'm not tinkering and the aforementioned repositories are… about it. There is little documentation about

- how to build a custom OS image (necessary if you're thinking about using Jetson as part of your own product, i.e. a large-scale deployment). Ideally, I would like to do that as part of a CI/CD workflow. What proprietary drivers and libraries do I need to install? Nvidia basically says, here's our custom Ubuntu image with the usual GUI, complete driver stack and everything – take it or leave it. Unfortunately, the GUI alone is eating up a lot of the precious CPU and GPU resources, so using their OS image in production is no option for me.

- how deployment works on production modules (as opposed to the non-production module in the Developer Kit)

- what production modules are available in the first place ("Please refer to our partners")

- what wifi dongles are compatible (the most recent Jetson Nano comes w/o wifi)

- how to convert your custom models to TensorRT, what you need to pay attention to etc. (The official docs basically say: Have a look at the following nondescript sample code. Good luck.)

- how do I automize testing of TensorRT models & code? Why isn't there a compatibility layer which allows executing my models on non-Jetson devices / on a CPU (obviously at lower performance, but I don't mind).

- … (I'm sure I'm forgetting many other things that I've struggled with over the past months)

Anyway. It's not that this information isn't out there somewhere in some blog post, some obscure Github repo or some thread on the Nvidia forums[2]. (Though I have yet to find a wifi dongle that works reliably…) But it usually takes you days or weeks to find it. From a product which is supposed to be industry-grade I would have expected more.

[0]: https://github.com/dusty-nv/jetson-inference

[1]: https://github.com/jkjung-avt/tensorrt_demos

[2]: https://forums.developer.nvidia.com/c/agx-autonomous-machine...

Reply on news.ycombinator.com