Are you sure Linux doesn’t support shared GPU memory? I mean if you had an integrated GPU with no strictly reserved memory which is fairly common on cheaper notebooks the GPU has to share the memory with rest of the system. There’s no other way for it to even function.
Pretty “swapping” VRAM to system RAM has been supported for a very long time too. My GPUs can use up to 16GB each of system memory (AMD), and I’d be really shocked if NVIDIA’s proprietary driver doesn’t either because I’m sure the AI workloads need it.
Of course the Steam Deck is a prime example of dynamic CPU/GPU memory allocation as well.
Are you sure Linux doesn’t support shared GPU memory? I mean if you had an integrated GPU with no strictly reserved memory which is fairly common on cheaper notebooks the GPU has to share the memory with rest of the system. There’s no other way for it to even function.
Pretty “swapping” VRAM to system RAM has been supported for a very long time too. My GPUs can use up to 16GB each of system memory (AMD), and I’d be really shocked if NVIDIA’s proprietary driver doesn’t either because I’m sure the AI workloads need it.
Of course the Steam Deck is a prime example of dynamic CPU/GPU memory allocation as well.
If you’re running this GPU under Windows, it’s fine. But good luck doing that under Linux.
https://forums.developer.nvidia.com/t/non-existent-shared-vram-on-nvidia-linux-drivers/260304?page=2
Fair enough, another one for the NVIDIA woes list!