Garmin 645 GPS Accuracy Results & New 2018 Tri Watch Revealed

Best running watch with music Garmin Forerunner 645 Music

Garmin Forerunner 645

Here we look at the GPS accuracy of the Garmin Forerunner 645 Music and the GPS accuracy of 2018’s first new multisport watch; it’s feature-RICH; it’s super CHEAP; and it’s the Amazfit STRATOS. And, with reservations, I quite like it, but that’s another story.

Typically I have found that Garmin GPS watches are feature-full but ‘only’ broadly good when it comes to the accuracy of some of their onboard sensors. The arrival of a new feature-rich and rather pretty smart watch at the £150-$200 kind of price level might make us sit up and take notice if such a cheap watch can go some way to matching what Garmin can stash away in their burgeoning menus of sporting goodness. At least it might make us sit up and take notice if it’s also accurate.

So let’s talk about Accuracy. Specifically GPS Accuracy.

Contention: Runners deserve the same level of accuracy as cyclists.

Cyclists love to have the accuracy of a good power meter as a measure of their outputs. A good level of accuracy for the best power meter seems to be  between +/-1% and +/-2%. Fair enough.

But what is good accuracy for a runner?


Many runners are very interested in knowing the exact distances they have run. The exact distance run is THE goal for their session. Personally that’s not for me but I fully understand that need. The total distance run can be relatively easily validated with a little bit of time, planning and patience.

In my tests, most GPS sports devices over >10 mile routes pretty much all have a +/-1% accuracy level for the overall distance. All the various under- and over-estimates during the route typically cancel each other out over longer distances.

Best running watch with music

Polar M600, Mobvoi Ticwatch S, Garmin 645M, TomTom Runner 3 Music


Other runners might simply be interested in recording a GPS track of where they have been. All the alleyways, bridges, turns and cul-de-sacs. Fair enough. I like to have that track too but it’s usually more for the sake of interest than any kind of training-related reason.

Almost all modern running devices and smartphones will record a GPS track accurate enough for this purpose for most people. The STRATOS and 645 are perfectly fine for this.

Having said that, both DISTANCE and ROUTE accuracy can be adversely affected if your route covers extended periods in more difficult areas of GPS reception – like cities, mountains, forests and the more extreme latitudes.

Garmin 935 Amazfit STRATOS Suunto SPARTAN Trainer

Garmin 935 Amazfit STRATOS Suunto SPARTAN Trainer


Putting relatively new running power meters to one side; PACE/SPEED is often used as a proxy for effort/output by runners and, if you ignore things like gradient and wind, then pace+speed are reasonable enough proxies for most of us.

Clearly a cyclist will not be looking at 1-2% accuracy for their overall effort; they want to know an accurate enough figure for their efforts NOW. A definition of NOW might usually be a 3s to 30 second moving average. For a runner interested to know their PACE/SPEED then, to me, it seems reasonable to expect accuracy over a similar time frame on the flat.

However if speed is derived from GPS (&/or GLONASS) then the many short-term vagaries of the reception of the GPS signal can make short-term averages relatively inaccurate. These vagaries range from the swinging of your arm to the proximity of tree cover and buildings.

So I’m specifically interested in the accuracy of how fast I am running NOW and much of that comes from knowing exactly where I am now – hence GPS accuracy. For those of you interested in accurate distances in difficult conditions then, I strongly suspect, that our interests lie in having the same level of underlying accuracy.

GPS Methodology, Results  & Source Data

All the info on the detail of the methodology I use for a standard test and the source FIT files should be in the link shown below.

Test Route For GPS Devices + Methodology

 The Scores on the Doors

Here are selected results as of 9April 2018.

Assume that there could be a couple of percentage points of error in my results. So the 69% that both devices scored might work their way up to a 74% with a couple of re-tests. BUT there is NO WAY a 69% would ever get above 80% and beat the Suunto/Polar devices shown there.

Superficially I would say that I was surprised that the 645 only scored 69%. It ‘seemed’ better to me than that based on all the runs I’d done before and since the test (I have since re-tested). Maybe the rain/cloud present that day knocked off a few points for some atmospheric reason? But even having said that I believe that I slightly generously scored the 645 at 69% on the day. The same score seemed about what I expected for the STRATOS.

SuuntoAMBIT3 RUN87%
SuuntoAMBIT3 Run81%
SuuntoSpartan Sport81%
TomTomRunner 379%
GarminFenix 379%
AppleWatch 275%
AppleWatch 273%
GarminVivoactive 373%
GarminForerunner 23573%
SuuntoSPARTAN Ultra #273%
LezyneMicro C GPS73%
GarminForerunner 30573%
SuuntoSPARTAN Ultra #269%
GarminFenix 5X69%
GarminFenix 5s69%
GarminForerunner 645 Music69%
SuuntoSpartan Ultra65%
LezyneMicro C GPS62%

If you look in some more detail at the spreadsheet that shows the results, the 645 scored relatively badly in the harder parts of the test. Maybe if I had GLONASS enabled there it would have helped? Maybe.The STRATOS supposedly has GLONASS but also scored relatively badly on the harder sections. As far as I could tell the STRATOS has no way to DISable GLONASS, although I am NOT convinced it actually has it enabled by default as the specs imply. But that’s another story for another day – ie the chip is stated to support GLONASS but I’m not sure that the firmware does!


Real-World GPS Usage

I’ve probably got at least 30 hours worth of workout data for these two watches. Naturally I haven’t spent 30 hours staring at the watches as I run. That’s the beauty of a standardised test. It gathers a lot of info. But it also misses out on some real-world experiences so I will share some of those here as well.

The most notable aspect of running with a 645 and a 935 is that VERY often their autolap beeps were synchronised for  extended periods, sometimes up to 10km. I don’t yet know the GPS chips but I’d assume they were the same or similar because of that. The antennae may be different as the 645 Music has a stainless steel bezel to deal with (it may affect reception to some degree).

The instant running pace of the 645 was notably closer than the Amazfit Stratos to the pace on a 3rd device whose pace came from a properly calibrated footpod. However there were still short-term discrepancies of up to 30 secs/km on the 645 compared to a 3rd device in ‘real-time’. That sort of performance is relatively normal for GPS-derived instant pace; the Amazfit Stratos was worse still with variations of over 45 secs/km.

HOWEVER: Many people might not even notice those VERY significant pace errors. They might only be there for 5 or 10 seconds and such variations could be hidden if the runner was looking at ‘lap pace’ where the variations are averaged out.

HOWEVER: In some running scenarios both the Amazfit Stratos and the Garmin 645 were GREAT at showing a slight variation in instant pace of about +/- 5 secs/km – BUT that was typically at constant speeds, in straight lines and in the open. ie in EASY conditions and probably for up to a kilometre. I’ve not tested either watch on the track; my performances may or may not translate well to the track for you.

If you were to run with a Polar V800 or a Suunto SPARTAN Sport you WOULD see an improvement over the two devices on test.

For cycling usage AND looking at the live display, I noticed no material difference between either the STRATOS or the 645Music when compared to a WAHOO ELEMNT/ELEMNT BOLT or Edge 820.  The STRATOS and the 645 had VERY OCCASIONAL wobblies for a few 10s of metres but absolutely nothing that I would be concerned about. Offline, those wobblies over 10s of metres were sometimes apparent yet the GPS tracks all generally matched closely enough when cycling. Although the 645 did show repeated deviations of over 5m from the correct position.

GPS when swimming?

The 645 is not a tri watch although I have swum in a pool with it a few times.

Furthermore, I’ve not tested the STRATOS’s GPS in open water. It’s too cold right now. #Wimp.

Amazfit STRATOS Garmin 935

Amazfit STRATOS Garmin 935


I suspect that most people don’t really care about GPS accuracy and will never know how accurate or inaccurate their devices really are.

But here YOU are reading this. And here I am writing it. To people like us it IS important. Or at least we think it’s important 😉

Best Amazon price is linked to. Prices approx $400-$450 or £480-£520 and UNlikely for non-Music to fall below $400 in 2019. Buying from Amazon supports this site

It’s pretty obvious that there are flaws in my formal test. There WILL be flaws in any test you can devise too. But if you temper-in real-world experience and if you look at other independent-minded people like @fellrnr and read the various support forums then you’d probably derive the consensus conclusion that what I’m saying, above, is ‘about right’.

To be clear: I have ZERO relationship with Garmin. I’m like you I buy my Garmins with real money. I do make some affiliate revenues from Garmin sales but, hey, realistically, how many Garmin 645 sales do you reckon a 69% GPS ranking will get me? 🙁

My general opinion is unchanged: Garmin devices have ‘broadly good’ GPS accuracy and that’s good enough for most people but not good enough for the prices they charge.

The STRATOS had comparable GPS performance to the Garmin. I suspect Amazfit’s onboard algorithms are quite as good as Garmin’s. How can it be right that a £150 relative newcomer can perform as well as a top-end Garmin. I accept the shortcoming of the STRATOS but not the Garmin 645.

Am I anti-Garmin? Obviously. That’s why I say that the Garmin Forerunner 935 is the best triathlon watch EVER made 😉 and that’s why I use the 935. Or maybe I can just have some great opinions about Garmin as well as some less-than-great opinions?

Hopefully I do my best to say it how I find it.

Oh. And I find the Amazfit Stratos VERY, VERY interesting for the super cheap price. Someone should review it 😉

Oh. I’ve said it before, if you want accurate running pace then buy a footpod. STRYD *IS* the most accurate.


Best Amazon price is linked to. Prices approx $400-$450 or £480-£520 and UNlikely for non-Music to fall below $400 in 2019. Buying from Amazon supports this site

Supporters: I am not a salaried journalist and rely on support from readers to keep the free content coming. If you want to support the work here then ad-free subscription starts at 49p (about 65c). Alternatively buying anything from my partners, below, also helps and may also get you a great discount. Thank you!

Support with 10% code: the5krunner10Support at your local country’s Amazon storefrontSupport with 10% code: the5krunner1010% Discount On Most Power Meters

Leave a Reply here

5 Comment threads
7 Thread replies
Most reacted comment
Hottest comment thread
6 Comment authors
Joopek80Maxthe5krunnerJustin Recent comment authors
newest oldest most voted
Notify of
Robert Black
Robert Black

I’d have ordered the stratos but the lack of sensor support beyond bluetooth heart rate monitors put me off. Is the screen as nice as the press shots? i used to have a nice watch and do miss the chronograph look. Then i got all childish and dove head first into gps watches lol

Robert Black
Robert Black

Do you have any insider info from runscribe? are they making any more of the current generation units, or will the next release be the Pro Plus


The 645 is going to be a good “middle-of-the-road” option for people that have been in Garmin’s ecosystem for awhile now but at a lower price-point and are now looking to do more. What that more might be is up to them, but the options on the 645 are more robust than the vivoactive 3 for example. Now, for a person like me; the 645 has nothing that I don’t already have in the Forerunner 935. The tap pay maybe? Music on-board is superfluous for me as I carry my phone out on runs. Couple that with a splotchy GPS, the 645 will be a transition device for people either wanting more from their watch or working their way up to a more “Robust” device down the line. As for the Amazfit–at $200 with all those options under the hood (But no ANT+ support…which seriously why? Is there something icky about using ANT+?) that watch sounds almost too good to be true. I mean they are even claiming battery life akin to the Forerunner 935, and with nearly the same toolkit! I would like the GPS/Glonass to be at least usable on the Stratos, but as said above; I rely… Read more »


I went from the FR630 to the FR935, accompanied by a Stryd footpod. Without the Stryd any upgrade from the FR630 is questionable, in my eyes, if you only run. Garmin Pay and a few extra metrics from Firstbeat don’t make the FR645 a better watch. Music and the looks could be attractive for some people, for me only the extra memory for custom datafields (32kb instead of 16kb) was a driver to upgrade from the FR630.
Incorporating Galileo would make a real difference to the more serious runners, assuming that it has a high accuracy. Now the Stryd has to be bought to make pace and distance reliable on the newer Garmins. If watches with Galileo would appear, then I would advise people to upgrade.


You have two 935 lines in the table


Amazing how Ambit3 still kicks everyone’s asses 🙂