Testing 5G: Hype vs. Reality

Performance and speed tests

You may be wondering what to expect from 5G in actual real life use. We were too, so we took the Huawei P40 Pro, the Samsung Galaxy S20 Ultra 5G, and the Realme X50 Pro 5G for a spin around our neck of the woods to see what's what. We chose these phones because each of them uses a different chipset: the Huawei has the in-house Kirin 990 5G, the Realme uses the Snapdragon 865, and our S20 Ultra is an Exynos 990 model. We were curious to see if we could spot any differences in 5G performance between them, subjective as those perceptions might inevitably be.

Our carrier of choice has an NSA Sub-6 5G network, using 100MHz on band 78, which is around 3500MHz. We're going to show you the screenshots of our speed testing results first, and then discuss the umpteen caveats that apply.

First off, here's what we were able to achieve on the P40 Pro with the Kirin 990 5G:

5G speed tests on the Huawei P40 Pro

Following are the results from the Realme X50 Pro 5G, with its Snapdragon 865 at the helm:

5G speed tests on the Realme X50 Pro 5G

And finally, here's what we got from the Exynos 990-powered Samsung Galaxy S20 Ultra 5G:

5G speed tests on the Galaxy S20 Ultra 5G

We also had a Huawei Mate Xs around, and its results are so similar to the P40 Pro's that we felt it would be pointless to include them as well. This, of course, is no surprise, as the Mate Xs uses the exact same Kirin 990 5G chip as the P40 Pro.

Now, in our experience with these phones on 5G, we came to realize that both the Snapdragon 865 and the Exynos 990 seem to be marginally faster overall, on average, than the Kirin 990 5G. Because of varying network conditions and all that, it would've been impossible to do a true objective scientific test, so please understand that this is more a subjective impression than anything else.

Interestingly enough, the Kirin 990 5G makes up for that perceived difference in battery efficiency. The P40 Pro and Realme X50 Pro 5G have very similar battery capacities, but the Huawei phone has better battery life, both when actively using 5G and also when just being connected to the 5G network while using Wi-Fi. There still seems to be a battery penalty to going with a 5G-capable chipset compared to the most efficient SoCs of yesteryear, mind you, but that penalty is lower with the Kirin 990 5G than it is with both the Exynos 990 and the Snapdragon 865. Perhaps it has a lot to do with the fact that the Kirin 990 5G is the only chipset of the three that has the modem fully integrated.

And back to our tests, it's worth noting that compared to 4G, all three phones suffered from an increased attenuation which comes with the territory of the higher band being used. The highest 4G band our carrier uses, like in most of Europe, is band 7, aka 2600MHz. Given that its 5G band is around 3500MHz, the signal is less stable when it passes through walls, glass, and just in general when there are any obstacles between you and the tower. Of course, we're just talking a slight decrease here, things are nowhere near as bad, attenuation-wise, as they are with any mmWave network. This can be mitigated by having a denser tower rollout, which is probably still attainable without huge financial downsides for a carrier.

Let's add some context

Okay, now let's address the fine print, which will be lengthy. First off, we aren't going to be paying much attention to upload speeds, because our carrier still uses 4G for those. Not just that, but the 4G band that handles these uploads is band 20, which happens to be that for which our carrier has the least amount of capacity. There are technical reasons for this which we won't get into, suffice it to say that once uploads will be over 5G, the speeds will increase accordingly. As the situation is right now, the theoretical maximum for uploads in this setup is 75Mbps. With that in mind, the results you see aren't actually bad, some of the time.

For the purposes of having some context, note that when using 4G without carrier aggregation, on the same network, we usually get around 40Mbps downloads. This isn't the best effort sort of result, at 3 AM when the network is practically empty, and in the best possible position, etc, it's just an average speed on an average day at an average time.

On 4G+ (also known as LTE-A or LTE+, or basically 4G with carrier aggregation), we usually get around 80-90Mbps down, again, on average. Peaks hover around 150Mbps, but those are getting harder and harder to find, now that most people do actually have a 4G+ capable handset. If you're curious, note that our carrier aggregates two or three bands (depending on where you are), for a theoretical maximum download of 450Mbps (on paper).

On 5G, the carrier uses 100MHz of bandwidth, which, as things stand right now, on NSA and without any carrier aggregation, should allow for a theoretical maximum download speed of 1.2Gbps. Except that's very much theoretical, like with Wi-Fi "top" speeds.

As you can see from our screenshots above, in real life use, we got over 100Mbps almost every time, while the lower results can be attributed to the 5G rollout still being in a sort of beta testing phase for the carrier, despite what marketing claims may have been made.

Generally, though, unless we were in a spot with bad reception, 200Mbps+ was easily achievable, and 300Mbps+ wasn't out of the ordinary either, with peaks around 450Mbps. Depending on how you want to count, that's at least 2x better than 4G+, and maybe even 3x.

Of course if all things remain the same, then as soon as the 5G network starts crowding with new devices using it, speeds will go down, but that's not necessarily a given for the near future. First off, 5G can simply accommodate more simultaneous connections than 4G, it can better make use of the bandwidth it has, and there's also just a lot more of that bandwidth.

Let's also reiterate that while we're happy with this carrier's current implementation, clearly it still has a lot of kinks to iron out with 5G, and that might make the connections more stable and the speeds faster.

What should you expect?

From our limited experience with 5G NSA, we would expect to get a 2x-3x improvement in download speeds over 4G+. Remember, this is Sub-6 5G (band 78), with 100MHz of bandwidth being used. If your carrier has less bandwidth to allocate to 5G, then obviously both the theoretical maximum speed as well as what you'll see in real life will fall. And if it's more bandwidth, then you can see even better results. And if the network is better optimized than ours, that might contribute to an uptick as well. There are just a lot of variables at work, and that's why it's kind of impossible to simplify this.

In the US, for example, T-Mobile has its "low-band" rollout, using band 71 (a.k.a. 600MHz), and that's great for coverage per tower, but the magenta carrier only has 10 to 25MHz of spectrum to work with (it varies depending on exactly where you are in the country). In the best case scenario with 25MHz being used, the theoretical (emphasis on theoretical) top throughput would be 300Mbps, while if your area only gets 10MHz, then that falls to 120Mbps, or thereabouts. Again, we're talking theoretical maximums here, in the real world the top speeds you'll get will probably hover around one third of that, while the average speeds will be even lower of course.

If you spread this 600MHz "low-band" 5G through any area that isn't rural or barely inhabited, then you get even less from it in real life testing, because it's much easier for a tower that covers more ground (like these are) to reach the point where it just can't serve more people well - because its catchment area is bigger than any "mid-band" rollout. So while this could be a great strategy to use to bring broadband connectivity to rural parts of America, it's not a great idea in cities of any significant population density.

And then there's latency. The improvements here should theoretically be bigger, but that depends a lot on what your carrier uses for communication between its 4G towers, whether it's fiber to the tower or something else like microwave or satellite links. The latter introduce way more latency.

The best ping we've seen so far is around 15ms, which is good but not jaw-dropping for sure, and it's actually not very far from the best pings we've seen on 4G towers that are connected via fiber. There's a lot more room for improvement here, though, what with low latency being one of the biggest benefits of 5G. While 15-ish ms might be enough for some internet-of-things applications, others could require even less - think robotic surgery that's human-assisted over 5G, or, yes, those self-driving cars again.