Sure, the Pixel Watch 5 has a bunch of other new features. Some of them notable, most of them shrug. However, there’s one feature that stands out from the rest: New algorithms that substantially increase the accuracy of the Pixel Watch 5 GPS tracks. So much so that I flew to New York City (arguably the most difficult GPS location on earth), to put it to the test against the Garmin Fenix 8 Pro and the Apple Watch Ultra 3.
So confident in their product and accuracy, the Google team literally handed me the watch (well, a few watches), and told me to go off and basically try and break it. So off I went, hour after hour, trying to find ways to break the Pixel Watch 5 in New York City. Running, cycling, walking, you name it.
And to be fair, I did (barely) manage to break it. But mostly I broke the Garmin & Apple GPS tracks. Over and over again.
So, let’s get right into it, with briefly covering what’s new on the watch overall, and then we’ll GPS geek-out for a bit (because it’s crazy cool), and then after all that I’ll circle back to hands-on with all the other new features.
Pixel Watch 5 – What’s New?
Now obviously, I’m focusing heavily on the GPS pieces, because in my opinion, that’s the most substantial piece here. But there are some other notable pieces, including the inclusion of a new strength workout feature that’s bundled with an expanded workout builder feature.
Here’s the full list of what’s new on the Pixel Watch 5, before I quickly go through some of those other features:
– Increased processor CPU power by 12%
– Increased RAM by 60%
– Claimed 20% faster overall operations
– Added GPS post-processing accuracy gains (for walk/run/cycling-based activities)
– Added weight and rep tracking
– Added workout builder (app)
– Added audio cues for structured workouts
– Added rest timers for structured workouts
– Added new Bedtime mode with face
– Added 11 new glanceable complication styles
– Added 2 new watch faces
– Added new generative watch faces (ok, this is legit cool, see below)
– Added ‘Raise to Gemini’ questions/answers in no-internet situations
– Added Gemini proactive text responses for automatic replies (requires Pixel 11 phone)
– Claimed 15% more accurate sleep tracking
– Added new smart wake alarm option
– Added new overnight sleep breathing quality metric
– Added ‘Health Guardian’ umbrella feature/concept
– Added Breathing emergency detection (will detect persistent loss of oxygen saturation)
– Added new Blood Pressure trends report (app-based, coming to Pixel Watch 3/4/5 this fall)
– Added new insulin resistance trends report (app-based, coming to all Pixel & Fitbit devices this fall)
– Same exact case sizes as before (
– Orders as of today, August 12th, shipping August 20tH
– Also added new Steph Curry special band edition, shipping Sept 3rd
Pricing has increased for the first time since the launch of the Pixel Watch series, now $50 more than last year for the non-LTE models, and a $30 increase for the LTE models, bringing it to:
41mm: $399 for Wi-Fi and $479 for LTE
45mm: $449 for Wi-Fi and $529 for LTE
Steph Curry Special Edition (LTE only): Starts at $579
Here are the color options, shown below (Obsidian aka black, Olive, Fog aka white, and Canyon):
Now, if GPS isn’t your jam, then slide all the way down to the end of this post, where I dive into many of these new non-GPS features in more hands-on detail.
Google’s GPS Changes Outlined:
Ok, GPS geekery time. First, let’s talk about what Google says they’re trying to solve for, then talk about how they’re doing it, and then talk about whether they actually pull it off.
To begin, Google says they’re focused on trying to solve for really three specific GPS issues. To consumers, these ‘issues’ manifest themselves as all assortment of wonky GPS track errors, but to GPS engineers, these are basically three specific issues. They are:
– Poor tracks in tall building scenarios: Commonly known as urban canyons, such as New York City, or really any building, tall buildings will not only block GPS signals, but more critically, as far as bad tracks go, they will literally reflect GPS signals (off glass & metal), which tricks GPS watches into thinking they are somewhere they aren’t. In GPS-lingo, these are called multi-path errors, because the path from the satellite to your watch can take multiple paths to the watch, only one of which is ‘real’/true (or even none of which).
– Terrain blockages: This is a semi-similar bucket to the first one, except for natural terrain blockages, like cliffs, or mountains, or tree cover. These things are less likely to reflect the GPS signal, but can still present challenges due to limited GPS satellite views (remember, a GPS watch will normally be connecting to upwards of dozens of satellites at once).
– Atmospheric issues: This is probably the bucket you think of least, but still a very impactful part of the signal. This is mainly focused on whether, including heavy cloud cover, rain, thunderstorms, fog, etc… Basically, anything in the sky that’s dorking up GPS reception.
Ok, with that out of the way, let’s talk about how they are making the tracks better. And now it’s important to note that they are doing this in post-processing. Meaning, it’s not fixing your exact position in real-time as you go down the street, but afterwards. And critically, it is very much not just road/trail-snapping (as some of their competitors do). In fact, I’ll literally demonstrate that over and over again, purposefully trying to trick it (and where the competitors will route snap). But we’re getting ahead of ourselves.
Google is using three specific fixes to post-process the tracks. They are as follows:
1. 3D Building Models: Here, they are using Google’s massive database of the entire world, mapped in 3D (think Google Earth), which in turn knows the exact building heights of just about everything on this planet. From buildings to mountains to bridges, it knows it all. By knowing there’s a building there, and the specific height, it can determine whether or not you (running on the street or wherever) can actually see the satellite your watch thinks it can see. Or, if instead, that GPS signal is being reflected off of a different building, causing a multi-path error/inaccuracy. By knowing which satellites can and cannot be seen by the watch, it can re-prioritize which signals to trust, and then re-process that file with satellites the watch can actually see. All of this also applies to non-building terrain too.
2. DGNSS Weather Correction: The next step is to use real-time reference stations to understand current weather and atmospheric conditions for your exact location. From there, Google will apply these corrections to your file, attempting to mitigate that rain or otherwise problematic weather.
3. Signal fusion: Finally, Google was already doing various on-device signal fusion like most companies (e.g., looking at accelerometer/gyro/etc to see if you were actually going the direction GPS says you are), but now, they reprocess that all again on the cloud, with the 3D building/terrain models and DGNSS weather correction.
Note that they are doing this for pedestrian-based walking/running/hiking, as well as cycling-based activities. They do not impact/help other sport modes at this time, such as openwater swimming.
So, let’s get into some actual tracks here, before and after differences. Note that for production Pixel Watch 5 units, you’ll never see these before/after tracks (just like you don’t for Apple when they process your tracks, or for Garmin, unless you creatively intercept them before processing). But in my case, I was able to go run/ride with two watches during all my NYC tests. One was a production watch, and one was a development watch, which let me see the files before they were processed.
Thus, in this section only, I’m just going to show Google Pixel Watch 5 before/after. In the other sections, I’ll show production tracks vs the others. Make sense?
Here’s the first one, showing how it processed this section, jumping straight into a building section. Purple is processed/fixed, yellow is raw upload from the watch. You can see how crispy the lines are now, shifting me to exactly where I ran in both sections (on the sidewalk, crossing the street, etc…).
The changes aren’t always massive; many times, they are reinforcing confidence in what you’re seeing matches where you ran. Here, the unprocessed track is wobbling its way along, yet the processed track is spot-on:
Next, this section here, which includes a ‘trick’ test that I did at the far right (more on that later), you can see it cleaned up the yellow to match where I ran, but equally, didn’t remove the trick-loop that I did at the far right. Again, super impressive.
The same is true here in Times Square, as it takes me out of the buildings, and instead, shows me weaving through the open plazas to find my way:
Still, there are some slight imperfections, like here where it has me running through the Times Square stands/steps (even corrected). It’s only a couple of meters away, but again, it’s not absolutely perfect.
Or here, a few blocks later, where it simply can’t overcome the buildings, and has me on the wrong side (and in buildings). In this section (as you’ll see later), Apple got it very correct, and Garmin more correct than Google.
Again, to me though, it’s streets like this. Where it takes a ‘meh-good-enough’ GPS track for 2026, and turns it into a ‘holy-crap-perfect’ GPS track.
I’ll stop doing the before/after for now, and instead, we’ll look at how it compares to the others.
But first, some of you are probably already asking: Will this come to other Pixel Watches? And the answer seems to be yes, though Google hasn’t fully confirmed that. Google (year over year) does this quirky thing when asked about new features coming to previous watches, where they always say some variant of ‘We always try and bring new features to previous devices, and will have more information in a future update’. Which is Google-code for “We’ll release this update to existing watches about 2-3 months after the new watch releases, but won’t tell you that now”.
They noted numerous times that their priority was releasing Pixel Watch 5, but equally noted there was nothing that prohibited it on Pixel Watch 4 (which also has multi-band GNSS/GPS). There are very minor antenna tweaks on the Pixel Watch 5, but practically speaking those don’t have a major difference here (as a reminder: antenna design is arguably the most important part of GPS accuracy these days).
Ok, let’s get into the results.
GPS Accuracy Results:
Ok, let’s dive straight into things. In my case, I’ve got two different runs (focused on two different scenarios & areas of town), along with a ride where I went all over creation. As part of my full in-depth review, I’ll get more into comparisons in tougher terrain.
For the purposes of simplicity in this post, I’m going to focus on a longer run, and a bike ride. There’s also a shorter run in a different section of town available here.
So, to begin on this long run, things are pretty similar between the three watches, without a lot of meaningful differences at first glance. However, it doesn’t take long for the separation to begin. Here we can see the Garmin Fenix 8 Pro on the wrong side of the street, while the Pixel & Apple Watch are on the correct side of the street, with the Pixel watch having me correctly on the actual sidewalk, versus the Apple Watch playing a tiny bit in traffic.
Next, as I enter this park, the Pixel Watch is a couple of meters off the crosswalk that I used. I mention this mostly to be fair, not because it’s a huge deal (since I was only a few meters off above with the Apple Watch).
In any case, more notably though, I did three loops around this small pergola of sorts, which was perhaps 3-4m wide. All three units actually tracked that surprisingly well.
Moments like this show slight differences. That again, might not be super obvious. Apple has me in the buildings and the street. Whereas Google shows a more clear crossing of the street. Garmin’s track isn’t bad, but again, isn’t exactly where I went.
Here on this turn, you can see the real advantage of Google’s multi-path algorithms reducing the errors, whereas Garmin is on the wrong side. While Apple is on the correct side, this is almost certainly automatic snapping to the sidewalk going on here (which Apple will do, more on that later).
Still, the Pixel Watch 5 is not perfect. Below (the vertical street), a perfect example of that. You can see Apple correctly keeps me on the sidewalk on this section, whereas Google put me into the side of the buildings across the street. Of course, one only needs to look at the street before (the horizontal one here) to see that Google nailed that.
Of course, my aim is to try and break things. And that’s exactly what I did below. In this case I stopped for a cold drink, and then decided to walk through under this bridge (not running), go about 25 meters beyond the bridge, then cross the street, and then turn back up the street. Google absolutely nailed this. Garmin got the general gist of things (vaguely), and Apple simply decided I wasn’t stupid enough to actually do that and ignored me entirely.
And as I approach the outer edges of Times Square, you can see bigger buildings equal bigger problems for Apple. For Garmin, they mostly aren’t in the buildings, but in the middle of the road. Whereas Google is exactly on the correct sidewalk side.
As I enter Times Square, with the massively tall buildings on both sides, the units actually don’t do that bad. Of course, I’m here to cause problems. So I do a tight loop around this building in the center. Google is the closest to correct, though still with some light building shaving. Garmin gives the 4yo toddler version. And Apple just makes a mark across the building. To the right of the building, the Apple/Google squiggle is correct, as I stopped to take some photos there. Garmin has me on the wrong side of the street.
And one more moment on this run that I aimed to break things was this circle I ran in the middle of the street (it’s now a closed street). You can see Google perfectly hits that circle (I show it in my video, from the GoPro perspective). Garmin isn’t bad, albeit a bit tight. And Apple again decides there’s not way I actually drew a circle in the middle of the street, and thus puts me on the sidewalk.
But again, Google is not perfect. And just a block later, it throws me into the buildings, while only Apple correctly sails down the street on the right sidewalk.
Anyways, for the rest of the run, this pattern basically continues, though Apple did struggle a bit more at times in the buildings. You can view the whole track here, if you’d like.
Instead, let’s look quickly at a bike ride. Generally speaking, cycling is at a higher speed, and tends to mean that GPS tracks are better, since watches (and bike computers) have a higher sensor confidence that you’re going 20MPH/30KPH in a given direction, and aren’t instantly off to the side some other way.
Now, starting off on this ride, I will briefly nitpick the Google, as Garmin got this turn correct. There was construction on the turn, and I had to go well wide around it.
Meanwhile, for quite a ways after that, the Google & Garmin units are basically the same. Apple is incorrectly snapping me to the sidewalk (I’m in cycling mode), whereas you can clearly see Garmin/Google have me exactly in the bike lane. Some number of blocks later Garmin disassociates itself from that though, putting me on the other side of the street.
Still, for big chunks of this ride, there’s no real meaningful difference:
Such as when I pedaled past the very tall World Trade Center, along with surrounding buildings; there’s no major difference here. A couple of meters one way or the other at most (especially between Garmin & Google).
For example, here I did a loop around this building with tiny streets. Apple overshot very slightly on one side, but so did Google, putting me on the sidewalk. Garmin cut in a bit too tight in one spot. All in all, mostly a wash.
Even when I have literally no visible sky above me, all three units get it roughly correct. Here I am below the highway, riding for a long time below it. Google is in the most correct spot below said highway, but none of them are all that far off.
Some time later, I head across the bridge. While all three watches have me clearly on the bridge, only Apple actually gets the position within that correct. The bike lane here is on the ‘upper’ portion of your screen in this scenario, whereas the red where the Pixel Watch 5 is, is the train lines.
As for the rest of the ride, it was mostly pretty good amongst all of them, with not major issues from anyone (again, cycling speed helps). You can see here how the Apple Watch cuts the corner a bit (I’m going from right, turning left onto this street), whereas both Garmin & Google correctly show me pausing a bit out further, waiting for traffic to clear before turning. After which, Apple puts me on the sidewalk, as opposed to Garmin/Google being on the bike lane.
And ultimately, DesFit and DesFit spent hours sitting in front of a conference room big-screen looking through hours of rides/runs we did together with the geekiest of Google Pixel Watch 5 engineers, comparing tracks from our Garmin/Apple devices, alongside both the Pixel Watch 5 pre/post processed tracks. Collectively, we judged literally every single block of every single track as to who was the winner at every moment on those workouts. On the rare scenarios that Apple/Garmin won, they would readily admit it, and then debate precisely why it probably won that half a block. The same goes for the even more rare scenario that the post-processed track was less accurate than the pre-processed track (it was almost never ‘less accurate’, but usually just ‘offset a different way’, where you’d say “I’d just split the difference instead”).
They will readily admit it’s not absolute perfection, but equally they are very confident the tracks are far better than their competitors. So much so that they are saying it’s “2X More Accurate Route Mapping in Tough Environments” along with “Our most accurate GPS route tracking for exercise, outperforming Apple Watch Ultra 3 and Garmin Fenix 8 Pro”.
How did they get to that 2X number, precisely? Well, with big fancy high-tech backpacks using centimeter-level GPS accuracy (note the head-sized pods sticking out the top). Two years ago I took this photo with that team. Oh, and that’s me at left wearing one.
From there, their claim is based on simply comparing the mean difference between each track point from each company (Garmin, Apple, Google) to the ‘ground truth’. These are the same backpack devices that all of these companies use for their own GPS testing.
Translated again, if they look at a specific point, and the Pixel Watch 5 (with post-processing) is off by 2 meters, and the Apple/Garmin watches by 4 meters, then they compare those differences. For every single point on the track. Which is how they get to the 2X claim.
Of course, I suspect that claim varies heavily based on environment. My guess is the number is higher in places like the buildings of NYC, and lower in places like the farmlands of Kansas.
New Pixel Watch 5 Changes in Detail:
Now, let’s take a closer look at some of the other Pixel Watch 5 changes as I outlined at the beginning. Again, there are a few dozen features, but I kinda want to dive into the more notable ones here:
1) Added workout builder (app)
2) Added weight and rep tracking
3) Added audio cues & rest timer
4) Added new Bedtime mode with face & Smart Wake Alarm
5) Added new generative watch faces
6) Added ‘Health Guardian’ umbrella feature/concepts
Let’s first dive into the workout builder portion, which allows you to create strength workouts in the app itself. You can do this by manually creating each portion via a library of movements (and then assigning reps/weights/times etc…), or you can also just straight-up ask Google Coach (Gemini) to create a given workout focusing on a given area, and boom, the workout is done.
From there, the workout gets pushed to the watch, where it’ll iterate through each set/portion automatically.
Notably, it can also do (new):
A) It’ll give you audio cues (voice instructions)
B) They’ve added a rest timer
Next, once it’s time for bed, there’s a new sleep watch face, similar to what most of their competitors have, to minimize/reduce the chance of blinding you in the middle of the night:
And then with that there’s new Smart Wake alarms, similar to what other companies have in terms of looking at the 30 minutes prior to the alarm time, and your sleep phases/stages, in terms of determining wake-up time.
Next, there are new generative watch faces. Simply put, as the watch (verbally) creates any watch face with any theme you want, and it’ll do it. Des of DesFit decided to make a pizza watch face in our demo, and thus, a pizza watch face was born (it gives you a few options and theme choices):
Last but not least there’s the new Health Guardian bits. This is an umbrella concept covers a bunch of features, including both a new Blood Pressure trends and Insulin Resistance Trends features (which are not medically certified). Those are new reports that’ll come later this fall.
Thankfully, this new Health Guardian feature doesn’t require a subscription or anything else.
Final Thoughts:
There’s no question that Google’s GPS accuracy improvements here are incredibly impressive. In my mind, this is a very similar generational leap in GPS accuracy to what we saw about 3-4 years ago when Garmin introduced multi-band GNSS on their Fenix 7 devices. It would be years before competitors got anywhere in the same ballpark as that (and even then, I’d argue only one or two other watches can get close to Garmin’s current multi-band GPS accuracy). Now, Google has very clearly taken the lead, at least when it comes to post-processed tracks.
Google is almost eliminating the rare ‘edge case’ bad GPS scenarios that Garmin still has left. Again, they are not 100% perfect, but they are really darn close (and that’s in arguably the hardest GPS environment on the planet). Of course, as anyone who has a GPS watch (Garmin or otherwise) can attest to, there are plenty of ‘fun’ places you might know in your area that cause GPS watches to drop a beat. Even the newest and fanciest of models. It’ll take time to see how the Pixel Watch 5 handles all of these special places.
As for the rest of the enhancements, they are interesting, though I’d argue most of them are ones that a good portion of their competitors either have, or long have had (e.g. sleep watch faces, workout builders, rep counting, etc…). Still, there’s some legit cool integration into the Google ecosystem around automatic text replies being more aware of the rest of your phone’s data (potential privacy concerns aside). For bits like the blood pressure trends and insulin tracking, I remain somewhat skeptical. We’ve seen companies like Whoop push the boundaries of what the FDA says is a medical device, with very little (read: absolutely none) medical certification or backing. What Google is doing here seems like basically the same thing.
Again, I’ll have a full in-depth review of the Pixel Watch 5 soon, once I can push it harder in places with more challenging terrain. Stay tuned for that, as that’s what’s on tap for this week here in Newfoundland, Canada.
With that – thanks for reading!
0 Commentaires