At least on Linux you have the Nvidia open source drivers, so I guess as long as someone is willing to maintain them. Then again, newer Nvidia drivers usually mean support for newer games, so if you don't go out of your way to play the newest slop you should be fine coasting on the old ones. Assuming that they don't go stale on Arch at some point in the way they went stale on Windows 7.
I watched a BudgetBuilds video about the RX480 yesterday and the latest AMD drivers apparently had some bugfixes for GCN4.
No fucking wonder AMD still has a foot on the race. AMD works flawlessly on Linux while there is just too much tinkering for Nvidia (and even then its a tossup if it even works).
Yeah, for the older cards AMD is way better than NVIDIA. But the newer ones, I’d say the drivers are about on par with each other. They both just work (you still need to manually install NVIDIA’s, but that’s not actually difficult). My 4090 has been remarkably plug and play in both Gentoo and NixOS.
My biggest issue with Nvidia on Linux is that while there is an open-source driver, a lot of the graphics stack is still closed-source. There's no equivalent to something like radv for open-source vulkan support. This makes a lot of the nvidia stuff much more fiddly than it really should be.
I actually saw a comment the other day a guy was asking about heatsinks on these modules. like fuck off nigger. It's utterly amazing how something so useless distorted CPU cooler design.
Maybe? I think it has shorter trace lengths and should be better than DIMMs in the long run. But CUDIMMs are also pushing speeds up.
Just watch as chinks will start making LP downdraft coolers in CAMM sizes/screw patterns the moment they hit the market.
You can absolutely put decent heatsinks on these CAMM modules and have the total height of them being less than that of DIMM slots and modules. If anything, they'll be way better at cooling as you'll be able to pass cool air through the fins horizontally, in line with the CPU cooler, so you won't have to make them as thick.
There's NVK which still has a long way to go with performance optimizations and modern features like ray-tracing and DLSS but it works for Maxwell cards and newer, even if later 900 series Maxwell and Pascal are stuck with horrible performance since they can't be reclocked.
You can slap in 64GB of RAM into most laptops, don't believe what OEMs are saying, "max supported capacity" is bullshit. Always check in Intel ARC, or equivalent AMD CPU tech spec site for real supported capacity, memory controller lives in CPU.
I still have my old 780 and 970. I was having issues with the 780 after 2 years of use, so I got a 970. 780 probably just needed repasted, but I had no idea you could do that with GPUs at the time. Hate throwing tech out, so I just kept it in case a younger relative needed me to build them a PC.
Intel ARK sometimes undersells what can be used. They claim 16 GB for N100 when it was given 32 GB by device makers at the start, later 48 GB, and recently 64 GB (DDR5 single-channel SODIMM). There may or may not be issues with using that much.
