Don’t their current beta drivers support like… 10 year old products?
I have a lot of complaints about Nvidia (which is one of the reasons I moved away from their cards), but longevity of support hasn’t really been one of them.
It’s not a perfect comparison, but if we go by the Steam Hardware Survey, the first item I can find on the list that’s not supported with the latest beta drivers is the GT 730, at 0.21% of users. And it’s from June 2014.
Its passmark score is 835, which is lower than the 9 year old Intel HD 520 (867). I somehow doubt though that driver support for Vulkan/Wayland will be the major blocker.
I’ve used a Jetson Nano, so I know enough about how “reliable” Nvidia is. Pooled in money with my college mates for my Bachelor’s final-year project to get an old one in good condition. I will never will never buying anything from Nvidia, because the software experience was dog-shit.
Linux image was stuck on Ubuntu, 20.04, CUDA didn’t work with OpenCV of the box, I had to build it locally, which took an entire day. Enabling the repository would fuck the entire image when I try to download Python libraries, and I would have to reflash it on the SD Card. And well, could I use a different ARM image from, say Fedora IOT? Well yes, but then the CUDA drivers are no longer accessible, because those drivers were never open-sourced, making it useless for computer vision. And what if I accidentally update the kernel? Again, CUDA just stops working. They just abandon their products on their own whims.
I also have a laptop with an Nvidia mobile GPU, and my Linux experience is so utterly terrible with it, I’ve just decided to disable my discrete GPU and use only the integrated Intel UHD driver.
Let me guess, only for newer GPUs?
Their new, open driver only supports 2000+ series, so I guess that also applies here.
desktop icons will only render on 5090 😎
Good to see those Cuda Cores finally getting used.
Don’t their current beta drivers support like… 10 year old products?
I have a lot of complaints about Nvidia (which is one of the reasons I moved away from their cards), but longevity of support hasn’t really been one of them.
I have a 650m from 2013 (almost 10 years) and there are zero working drivers aside from noveau ones.
I have a gtx 670 from 12 years ago and nouveau is incredibly slow for me. I need to use the proprietary driver.
Yeah, but because pricing jumped like someone set a firecracker off under it’s chair people are actually still using vintage GPUs.
It’s not a perfect comparison, but if we go by the Steam Hardware Survey, the first item I can find on the list that’s not supported with the latest beta drivers is the GT 730, at 0.21% of users. And it’s from June 2014.
Its passmark score is 835, which is lower than the 9 year old Intel HD 520 (867). I somehow doubt though that driver support for Vulkan/Wayland will be the major blocker.
I’ve used a Jetson Nano, so I know enough about how “reliable” Nvidia is. Pooled in money with my college mates for my Bachelor’s final-year project to get an old one in good condition. I will never will never buying anything from Nvidia, because the software experience was dog-shit.
Linux image was stuck on Ubuntu, 20.04, CUDA didn’t work with OpenCV of the box, I had to build it locally, which took an entire day. Enabling the repository would fuck the entire image when I try to download Python libraries, and I would have to reflash it on the SD Card. And well, could I use a different ARM image from, say Fedora IOT? Well yes, but then the CUDA drivers are no longer accessible, because those drivers were never open-sourced, making it useless for computer vision. And what if I accidentally update the kernel? Again, CUDA just stops working. They just abandon their products on their own whims.
I also have a laptop with an Nvidia mobile GPU, and my Linux experience is so utterly terrible with it, I’ve just decided to disable my discrete GPU and use only the integrated Intel UHD driver.