920 should be more accurate. I say should because the barometric altimeter on my Fenix 3 gets clogged with salt from sweat and the pool and goes out of calibration from time to time.

I think DC Rainmaker wrote an article on barometric vs gps based elevation. Found it. It’s a bit old but should be helpful:

https://www.google.com/...-device-gps.html/amp

I notice strava always over estimates elevation on any routes I create.

Gps/barometric elevation struggles with fine resolution. GPS introduces large (+/-15m) error but is good overall. Barometers drift with changing weather and can have fairly large random errors (~1m) too. What the Garmin unit tries to do* is re-zero the barometer to a known height when the GPS knows it’s got a good fix, when stationary for example. In order to get nice smooth data, this still needs quite heavy filtering.

What Strava et al will do is use mapping data, which relies on, typically, aerial radar/lidar surveys. These are good, but the data you get to play with for free is quite coarse. The grid of spot heights has some distance (say 10-100m) between points.

Myself I’d tend to go with the map for looking at a ride. For aero testing you need a whole lot better data, but that’s another story. :)

---

Developing aero, fit and other fun stuff at Red is Faster