43
Some local LLMs tested on an average gaming PC
(github.com)
Performance/vision based tests have tokens/s at the end of tables.
128GB won't make any difference in gaming when compared to, say, 32GB, so it's just to fit larger LLMs into memory. However, 0.5t/s is hardly usable for every day tasks, so 32GB will still be more realistic combination with the RTX4070.