I think it’s a given that Framework will eventually at some point in time offer a dGPU model of laptop
So ignoring the possibility of MXM(although that would be sweet), which vendor should Framework go with? Nvidia has traditionally been the titan of the industry with the most powerful GPUs on the market but AMD is far better supported in Linux with actually decent open source drivers
That and I freaking love polls so I’m throwing another one out there
NVIDIA GPUs are the only ones that are currently truly supported for machine learning development with CUDA and all. While I still hope the ecosystem of AMD GPU will get better on this subject, for now I would only take a NVIDIA GPU for this specific reason.
Having a local GPU is very useful for CUDA development and all . So I’m actually rooting for modular laptop GPU. Letting the choice to the consumer is always better!
I don’t think it makes sense to go with an amd dGPU first. Nvidia laptops still sell way better in the end unfortunatly. I mean the current framework laptops is designed to target the most lucrative market for latops: The thin and light and low power market with an intel cpu. This doesn’t mean there can’t be an amd version, but it probably won’t come first
Linux crowds are very few compared to the windows crowds. With nvidia GPUs you would aim for a way bigger market. And its not like nvidia GPUs don’t work on linux, even if their support could be better
Recently I’ve taken a preference to AMD even on Windows, because why should I need to sign into an NVIDIA account every few months to check if there’s updates available? The only reason I even had NVIDIA’s software on my last computer was to check for updates, not to optimize game settings or get the latest NVIDIA news. I can do those things without an account, thanks. And AMD’s drivers just download automatically, on Windows and Linux.
Personally, love AMD CPU, hated their GPU, I for the love of god can’t figure out how their configuration UI works and nothing other than games worked for me. Rendering doesn’t support cuda, no machine learning. Too much noise for whatever reason in renders. Ray tracing performance worse nvidia when nvidia is not that good to begin with. Maybe I am just not used to it but AMD GPUs are not good for my mental health.
Do they work? Maybe. The issue is rather that people running Linux tend to be more conscious of “software ethics”, wherein NVIDIA has a much worse track record, and thus those people tend to prefer to not support NVIDIA the company.
If NVIDIA were to suddenly make their drivers open source and actually release unencumbered technical documentation, I’d jump over there in an instant. As things stand right now, I express my displeasure in the only ways I can; posts like this, and with my wallet. I won’t buy a computer with an NVIDIA GPU.
I’ve only used Nvidia GPUs with Linux and have never had issues. But then again, if it comes down to one or the other, I’d be happy with either if their performance is good. Either would be better than not having one at all.
I do a lot of AI and ML development as well and also use NVIDIA GPUs. However I’ve recently started decoupling my training from my computer by using NVIDIA Jetson boards and a cloud gaming machine that has a nice NVIDIA GPU. The benefit is that if I’m training, I can grab my computer and run and not have to interrupt it.
I’m all for both, but AMD is probably the better option I think. Driver support is better, AMD is happier to support laptop manufacturers, and they’re about as capable. It’s the DLSS and CUDA support that people who are clamoring for nVidia. FSR can do a good job of 4k upscaling and works in either card. We just need a CUDA module to plug in for people to not care either way, right?