Perf in those type of games are a touchy subject. I've always had second hand pc and always managed to play Ill-something games
TL;DR This is not that simple.
For good performance, for example your card supports some compression methods (hardwired in the GPU), that method must be supported by the device driver (nv/amd/etc.), the framework (DX/Vulkan/OpenGL/etc.), and what's more the game that feeds the data to GPU must support it as well. If a game can't use texture compression on your card, then it must fallback to uncompressed textures, so updating them will eat up memory transfer bandwidth and will also eat up available VRAM -> serious performance degradation.
Simply put, if a game is written in a way to only support one particular manufacturer's proprietary compression method and nothing else, then it will run smoothly even on older, slower, second hand GPUs from that manufacturer, but will be laggish on the latest, top-of-the-line GPUs from another manufacturer.
And this is just the texture, the same goes for the shader compiler. It matters how good it's implemented in the device driver (so how up-to-date your driver is), or if the framework supports it or has to do cross-compilation between DX shader, GL shader, SPIR bytecode etc. The more abstraction required, the slower the resulting shader will be. Also some games just have poorly written shaders in the first place.
If you have lots of VRAM and fast bus and high performance shader processing (like most top-of-the-shelf GPUs), then you simply won't notice any these, because it will just brute force push through everything, even without the manufacturer specific optimizations.
All that being said, a well-written shader with optimized, culled number of triangles does not push the hardware to its limits, so that will run just fine even on older models from whatever manufacturer. So by the end of the day, it's the game optimization that can do the most for a good performance.
Like I've said, not that simple.
