I’ve found it depends a lot on the game. In CP2077, DLSS+frame gen looks great to me with full raytracing enabled. But in The Witcher 3, I found frame gen to cause a lot of artifacts, and in PvP games I wouldn’t use regular DLSS/FSR. In general I’ve found the quality preset in DLSS to be mostly indistinguishable from native on 3440x1440, and I’m excited to try FSR 3 when I get the chance.
IIRC it’s an APU thing, and last I heard it was just a rumor (could be out of date). Either way, non-LTSC is EOL in a year and a half. If you’re putting in a Zen 5 CPU, the best choice is realistically either Linux or Windows 11 Pro, since Pro can turn off all the bullshit through group policy. My Windows machine I have to have is on 11 Pro and it’s basically Windows 10 with a slightly different taskbar. No Copilot bullshit, no ads, no Bing in Windows Search. If you’re ok your taskbar on the bottom of the screen, IMO it’s the best choice as long as you have to use Windows.
Box is (basically) just the way to have memory on the heap. Here's a direct comparison of how to do heap memory in C/++ and in rust:
int* intOnHeap = (int*)malloc(sizeof(int));
*intOnHeap = 0;
MyClass* classOnHeap = new MyClass();
let intOnHeap: Box = Box::new(0);
let structOnHeap: Box = Box::new(MyStruct::default());
There can be a bit more to it with custom allocators etc. but that's the only way most people will use boxes. So Box basically just means "a T is allocated somewhere and we need to free that memory when this value is dropped".
I’d be interested in setting up the highest quality models to run locally, and I don’t have the budget for a GPU with anywhere near enough VRAM, but my main server PC has a 7900x and I could afford to upgrade its RAM - is it possible, and if so how difficult, to get this stuff running on CPU? Inference speed isn’t a sticking point as long as it’s not unusably slow, but I do have access to an OpenAI subscription so there just wouldn’t be much point with lower quality models except as a toy.