Amazfit Cheetah 2 Ultra – Is The GPS Accurate or Not?

Well, this was a bit of a surprise.

I had dreaded performing this test because I expected the same mediocre result as the Cheetah 2 Pro produced a couple of weeks ago. I’d much rather report good news. It’s just nicer. I assumed that the Ultra model was identical to the Pro but a bit bigger. That’s pretty much true, but the important takeaway is that all the innards are likely identical except for the bigger battery. However, there were a few peripheral expectations for good results – the larger Ultra model will have a subtly different antenna construction (and just a physically bigger one), the weather might be more conducive for upper atmospheric conditions affecting multipath – turns out I had a lightning storm during the test...oops.

The long and the short of it is that today’s result was better than the 2 Pro but not quite as good on the day as the Garmin Forerunner 970.

Test Conditions

This test is one of the key ones in my standard methodology. It follows a predetermined 10-mile route that includes many GPS difficulty points. Results are almost always indicative of general GPS performance in all but extreme conditions. I perform many other tests as well.

The Cheetah 2 Ultra was on test alongside the Forerunner 970. No other comparators are required as this is a longitudinal test covering many devices over many years. Several heart rate monitors were being tested: Fourth Frontier Zone, Fitbit Air, and Polar SENSE, but this article focuses on the GPS. The main review of the Cheetah 2 Ultra will cover HR accuracy in detail.

Test Results

Here are the visual results first, which show the overall route and a couple of points of interest. Cheetah is the purple line.

Cheetah 2 Ultra scored 81% overall, which is good and certainly better than the average for the current range of sports watches overall. For a dual-frequency GNSS chip (which this has), the score is a tad lower than I would hope for, but any difference there is explainable by the vagaries of the day. Meaning it could quite easily score 85% if I retested tomorrow in the absence of a lightning storm, but it would certainly not get a top score of 92%.

Contrast that to Amazfit Cheetah 2 Pro (73%), Amazfit T-Rex 3 (85%), Amazfit Balance 2 (85%), Coros Pace 4 (88%), and Garmin Forerunner 970 (92% – joint top score).

The score is broken down into differently graded and aggregated sections: Easy (91%), Medium (81%), and Hard (69%). Unsurprisingly, the watch found the harder sections, err, harder. That said, the Garmin Fenix 8 and Garmin FR970 counterintuitively performed best on the hardest sections, at least they did on the day of the test.

The lower difficulty score is due to running VERY close to tall buildings, through a long tunnel, and in a narrow alley with high brick walls. Essentially, situations where GPS can’t perform accurately. Amazfit claim to have dead reckoning (another term for GPS Sensor Fusion) but that didn’t work in the tunnel today. The Cut Throat Alley segment was pretty good, and the proximity to a very tall building wasn’t too bad. One medium difficulty area is next to less tall buildings and Cheetah 2 Pro exhibits the tracks being thrown away from the building – this is clear evidence that reflected signals (multipath) are used to determine position rather than discarded, as should be the case with a dual frequency chipset. That sounds bad but EVERY other dual frequency watch does the same. I think the answer is that the point of reflection must be 30 ish metres away. So when I am right next to a wall, the point of reflection is 3m away. If you are running through Manhattan, the point of reflection could be halfway up the Empire State. Those signals would be reflected, detected as such and rejected. Perversely my tests in Kingston are harder than downtown New York. Maybe. At least that’s my theory! – See the GPS accuracy resource on this site, where there are some interesting things I’ve written and deep dives linked from other experts on GPS device testing (eg MikeG a speed sail racer)

In the context of other bike-and-run tests, this result is representative. The 2 Ultra does have small deviations here and there, but it is actually pretty good overall. Essentially, the 81% score from this one test is warranted and validated by other tests.

Can Amazfit Improve?

A: Tricky

To improve performance in difficult conditions, the definition must be harder than getting the basics right. Which is what Amazfit has done here, it’s got the basics right.

We suspect the chipset is the Airoha AG3353. Coros use Airohas and get excellent results. Airoha was used by Garmin, which got pretty good results with it, but now uses Synaptics chipsets to get excellent results. The point being that there might be scope for improving accuracy beyond the standard sometimes achieved by Coros.

The bottom line for normal use in easy to normal conditions, Cheetah 2 Ultra is sufficiently accurate.

Last Updated on 5 June 2026 by the5krunner

the5krunner

tfk is the founder and author of the5krunner, an independent endurance sports technology publication. With 20 years of hands-on testing of GPS watches and wearables, and competing in triathlons at an international age-group level, tfk provides in-depth expert analysis of fitness technology for serious athletes and endurance sport competitors. ID