Haven't seen the benchmarks for 5060 Ti, but for the 4-series the 4060 also had half the memory bandwidth and that lead to almost directly half the LLM performa

Haven't seen the benchmarks for 5060 Ti, but for the 4-series the 4060 also had half the memory bandwidth and that lead to almost directly half the LLM performance. That said, if you just use it for dabbling, ie not millions of documents and whatnot but more your local ChatGPT, you're talking about 2 seconds vs 4 or whatever. So probably not.