It's quite possible that this is meaningless to you, especially if you are new to BI.
The rule is that you have to have enough spare RAM to load the model along with space for working storage.
To stress the point we mean physical RAM not swap space.
As guidance 20 million rows is about the point where switching to a "Cube" based product starts to become a realistic requirement as for a model this size you are probably talking about needing 8GB+ of RAM. 20 Million rows is within the bounds of a new desktops and laptops.
Having said that we have run 30 million row models and the performance is acceptable, but not great, and you start having to trim down the number of columns in the model to fit it into memory. Our 40 million row model is at the limit of usability, if you have nothing else, for example a graph takes about 4 seconds to appear.
You can tell when you have not got enough memory, because processing changes from under a few seconds to a lot longer.
Although PerBI benefits slightly from newer, faster processors, in practice it is normally restricted by the amount of spare RAM available.
We can't really give a definitive answer as it depends on the size of your model and the amount of free RAM.
The best suggestion is download the evaluation and try it.
PerBI is a single threaded application so it does not benefit from multi-core processors.
PerBI is a 32bit application so only cheaper modern PCs and Laptops contain less RAM than PerBI can use.