I changed from an rtl-sdr.com v2 dongle and LNA to an airspy mini. Here is a like for like comparison of the difference between the two when I had my antenna mounted in the loft - There’s about 2 months of rtl dongle data and about a month of airspy before I moved the antenna:
Blue is airspy, red is the rtl dongle. The only difference is the receiver - the same antenna, LNA and coax was used. You can see that there was a very large increase in received messages and the number of aircraft seen as well as a significant increase in maximum range. This is also prior to the improvements made to the airspy decoder later that year.
The airspy is using a pi 4, whereas the rtl dongle was connected to a pi 2, but that doesn’t affect the results really - the pi 2 was easily capable of running the dump1090 decoder and the pi 4 gave no performance advantage. The pi 4 allows running the airspy at 20MHz sample rate which does provide a boost however.
Here is the difference between the rtl-sdr dongle and the airspy with improved decoder and the antenna moved to the chimney - note that the rtl dongle data is from the summer when it’s normally much busier, with the airspy data covering winter scheduling from mid-September until just before the covid lockdown in 2020.
A well optimised rtl dongle will see a maximum message rate of around 1800/s, with a maximum number of visible aircraft around 275. The airspy mini will see a maximum of around 3100 messages a second and easily over 400 visible aircraft. I think @keithma has had over 500 briefly during a period of tropo ducting.
Obviously these figures are prior to covid, so numbers are substantially down on these at the moment.
I think an airspy mini is a pretty decent upgrade - you get performance which is on par with a dedicated receiver like a mode-s beast or radarcape (they may be slightly better, but not much) at a fraction of the cost of those receivers. Getting your antenna up in clear air is by far the most cost effective change you can make if you haven’t done it already though.