How do the GPS chipsets (and accuracy) compare between the Garmin Forerunner 920XT, 935 and 945?

I had 910XT, Fenix 3, 935 and 945.

From an accuracy perspective, the 910 was superior. F3 and 935 more or less equal. 945 the worst. From what I have seen I would put 920 between 910 and 935.

Not to be misunderstood: all of them had acceptable GPS accuracy for me as I am more or less only interested in a good track recording. As soon as it comes to pacing, none of the watches can produce reliable results as a Stryd footpod can deliver. So my choice for races and tempo training always is a Stryd and the recording from GPS is nice to have.

I didn't have 910XT, but 920XT and I would put them in order from best to worst: 935, 920XT, 945. YMMV.

Garmins have GPS issues since they switched to new chipset from Sony. I don't think they can do antything with that by updating software. I recenty bought older Vivoactive 3 Music, to avoid GPS issues. But it stopped working after 4 days and I had to sent it to service. 

I can only compare between FR935 and FR945 and the latter is definitely worse. Probably due to the new Sony GPS chipset which underperforms the older Mediatek used in the 935.

Especially for running where you rely on an accurate pace I prefer the FR935. As this is a hardware issue I doubt that any future software update will improve the situation. For the time being it is good to have both, the FR935 when you need exact GPS and the FR945 for anything else.

I recently ran with my old 230 and my 945, the 230 GPS was definitely more accurate. I love the 945, but the GPS is a definite letdown.

My ideal 955 would have much better GPS and smaller bezels on the display. I'm good with everything else (wouldn't complain about better battery life, but it's been ok)

I don't have a Forerunner, but I do have a Fenix 3HR, 5X+ and 6X Pro Solar. Every one of them started out with very poor satellite tracks, kind of wobbly and drink looking, but each of them has improved a lot with multiple firmware updates.

I also have a Polar V800 from 2014, replaced under warranty in 2015. That was my baseline for good GPS tracking and until recently it could not be matched by any of my Fenix watches, but now they can all give it a run for its money, maybe even beat it. So don't think that software cannot improve things. Make sure the firmware is up to date.

Of course, pretty pictures is one thing, track distance and instant pace another. Personally I am not too worried about those things. Distances are close enough from each watch not to bother me, but I'm not a competitive athlete - slow Parkrun for me at most. My local course usually shows nearer 5.1k than 5k, regardless of which watch I use. 

I think pace is probably the bigger concern and I just don't think GPS precision is good enough for that to be a reliable figure from second to second without some fancy accelerometer calculations to smooth the data. It's not a fault of the watches per se, more a fault of the form factor. If you want accurate pace in all environments including tall buildings and woodland trails then a footpod is the answer. I have Stryd. I don't worry about pace, I focus on power, just as I do when cycling.

I'm not sure that the next new watch will have the answer either, so maybe £200 on a good footpod is money better spent than £500 on another watch with, probably, the same constraints on performance.

HUMODDA said:

Probably due to the new Sony GPS chipset which underperforms the older Mediatek

I've read articles where Coros is using the Sony chipset and GPS is working great. It's been suggested that it isn't the Sony chipset, but how different manufacturers are using the chipset in their devices. 

Curious where you are .. I seem to have heard that Galileo may work better in Europe? Could be wrong/urban myth ...

London, UK. I also read that and it seems to be the case for myself.