I'm not an expert in UMTS-style cellular networks, but I think the numbers in this article are misguided or incorrect. :-(
E.g., the article seems to conflate "Internet-style" measurements of practical round-trip latency over a network path, with "telecom-style" measurements of round-trip latency across a single unloaded radio hop with immediate schedule availability. E.g. it gives a figure of "50ms" for the typical latency of 4G, and "1ms" for the typical latency of 5G. I don't think this is an acceptable comparison.
"50ms" is a believable ICMP ping time from a commercial LTE service, including delay to get a resource block, to sojourn queues inside the network, etc. (I just tried on T-Mobile LTE and am getting 21-69 milliseconds RTT, mean 36 ms, to the next hop.)
By contrast, "1ms" for 5G sounds like a 3GPP "user plane latency" figure (and even then, only for the "uRLLC" version of 5G that is probably only going to be used for more-exotic purposes). This is a lower-level latency measurement made on a single unloaded radio hop. You cannot compare the two.
I think the more-accurate 3GPP user-plane latency figures would be:
- 5G uRLLC ("Ultra-reliable low-latency communication"): about 0.2-0.7 ms (see above)
The practical IP round-trip time with all of these networks is much larger than the above; e.g. ~21-69 ms for LTE, and 22-76 ms for 5G eMBB services (https://www.lightreading.com/the-edge/how-5g-is-pushing-enve...). And even those figures are in the absence of load to or from the UE; if traffic has filled up the queue of the uplink (on the baseband chip) or downlink (in the UE-specific queue on the base station/eNodeB/gNodeB), users can experience latencies of 500+ milliseconds. This is called "bufferbloat" and it comes from a refusal by the baseband and eNodeB/gNodeB vendors to deploy queue disciplines that queue on a per-flow basis, or mark or drop packets when queues are large. (I have talked to many vendors about this, and the prevailing attitude is somewhere between "we are the phone company; we're legally obligated to transmit 99.999% of all packets to the endpoint, no matter how delayed everything gets" and "if there were an actual problem with latencies over our networks, companies would be coming to us like Facebook did and paying us a bajillion dollars to deploy custom traffic rules to let their traffic skip the line; nobody else has offered us this kind of money for a custom traffic rule, ergo there is no problem.")
Similarly, the article seems to somewhat conflate improvements in theoretical throughputs with improvements in the "typical download speed" a user might experience. The latter is not that strongly related to the air interface standard (e.g. LTE or 5G-NR) because any flow's experience depends heavily on how many other flows are sharing the same constrained resource, which in turn depends on (a) how much spectrum has the carrier licensed in your area, (b) how many base stations has the carrier decided to build/lease in your area, (c) how many other customers does this carrier have, trying to download right now through this base station, on this slice of spectrum, at this moment, and of course (d) is the base station actually close to you [do you have good SNR?] or is it spending all its airtime on slower modulation and coding?
Carriers sometimes use improvements in the air interface (which improve throughput-per-Hz) to save money by deploying less spectrum or larger cells. So even if a newer/fancier air interface is x% faster on a per-Hz basis, the benefits may not fully accrue to the users, if the carrier takes advantage of the improvement to deploy bigger cells, or less spectrum per cell, etc. (I once sat in a meeting with AT&T around the time of the 3G-to-4G transition, and in at least this particular setting, they were planning for roughly 1/3 of the improved spectrum efficiency to accrue towards better service for their users, and the remaining 2/3 towards saving resources for themselves. So, yes, improvements in technology do somewhat improve service for users, but... in some cases that may be a minority of the overall benefit.)
And, obligatory plug! In September, the IAB is holding a workshop on Measuring Network Quality for End-Users (https://www.iab.org/activities/workshops/network-quality/), with submissions due August 2. If you care about this stuff, and maybe you know something or have a perspective that could be helpful to the Internet community, please consider submitting to and participating in the workshop.
There is also aggressive power saving features in both 4g and 5g that make the UE sleep when there is no data (the connection can even sleep during the ping interval period, so its best to use ping -f). if you try to ping in sleep mode, the ue first has to wake up which adds considerable latency.
This impacts ssh sessions too, you can set ServerAliveInterval N (/home/user/.ssh/config on the ssh client machine) to some reasonable number to keep it alive.
I don't understand where the "typical" speeds come from. For me, typical speed on 4G is 100 Mbit/s in Belgium while it is 30 Mbit/s on 4G+ in France. Same plan and phone, just roaming in Belgium (where I live, but I mostly use my French SIM) and in a city centre both times.
So it seems to me to depend a lot on the operator and infrastructure, and I don't see how there is one typical speed to quote.
He is based in the UK and his site almost exclusivly covers UK networks with a UK user base reading the posts - so the "typical" speeds would be for the UK.
Typical speeds for "current G + 1" is always high and mysteriously gets significantly lower when "current G + 2" gets introduced. Rinse and repeat.
I'm not sure how come everyone is so highly interested in this churn of networking equipment, it's especially bizarre that telecoms are interested in it (it's a major expense for them), but there you go.
To me 5G is a bit like IPv6. Technically it's better, technically we need it, but in practice we're good enough already. And this is why still a lot of Internet is on IPv4. But thanks to this constant marketing, 5G doesn't have this issue. I'm neutral about it, but I can't care less if my next phone has 5G or not.
Yeah. Your ISP may connect the BTS by a fiber link, or radio link, or it may have X or Y simultaneous users, etc. There's a lot of elements in the chain.
What would be interesting is that apparently a 5G BTS can deal with a lot more users simultaneously, so ISPs may connect them with 2.5G links and provide decent speed of all of them, alleviating saturation problems.
Meantime the broadband industry continues to exploit the confusion between 2G / 5G (cellular technologies) and 2GHz / 5GHz (WiFi technologies) and tell people (specifically my neighbors for example) that they will get faster speeds if they use the “5G” network exposed by their wifi router. If this article addresses that confusion I didn’t see where.
Hey, kinda complicated. The reason they're telling people to use the "5G" network on their access point isn't originated in confusion about 2G/5G (although for sure some people ARE confused).
The reason they're doing it is that one way people test their internet is go to a download speed tester, and to prevent complaints and confusion, the ISP wants the fastest download possible.
The 2.4GHz band is incredibly noisy, and even a microwave running in an adjacent unit can tank your throughput. 5GHz doesn't penetrate walls as well as 2.4GHz, which means less noise in your house. So an excellent, easy way (for average consumers) to test WAN throughput is by using the 5GHz network with line of sight.
The intent and origination is not 5G confusion, but rather 5GHz's increased bandwidth due to a lower noise environment in the average user's house.
And actually, in a way "get faster speeds if they use the “5G” network exposed by their wifi router" is true: if you live in dense housing, between all of the devices spamming on 2.4 and the fact that walls "quiet" 5GHz reasonably effectively, using 5GHz can actually be a marked improvement for many users!
Not overcomplicating things intentionally. There's other reasons of course...standards, shorter wavelengths, etc. Not referring to priority(second point) but capacity. Tri band routers as I'm sure you're aware dedicate two radios to the 5ghz range, and one to 2.4ghz.
If you have say, gigabit internet, your capable bandwidth on 5ghz will be much, much faster, regardless if there's any 2.4ghz interference.
If anyone who isn't a radios expert cares... I prefer dongknows when reading about networks. A lot of people for some reason don't seem to like him, but he does a great job of both putting things simply, yet giving enough backing technical specs to not treat the reader like a child.
That's true, or in cases of in-network m2m communication of course. 5Ghz is still faster, you just can't utilize it if your internet speed is slow. It's like saying a Ferrari isn't faster than an old VW Beetle because the speed limit is only 25. Technically correct.
If you want analogies, it’s the ISP provides a VW (connection speed), but it’s towing a Tesla Model S (which is way quicker than a Ferrari btw).
Then someone comments about how the ISP is mixing and muddying the two, and some replies start going into all the technical details of how the Tesla is quick and why. Technically correct, but clueless because it’s the VW that is the problem.
Fair rewording, but the point remains. 5Ghz is faster than 2.4Ghz. That a person can't take advantage of it because of their ISP is irrelevant to that statement.
If the statement were, 'hey, if you want faster internet, use 5Ghz', well then it would only sometimes be true and such analogies apply.
I can see where we're getting off track.
Person A) ISPs are telling people use 5Ghz to get faster internet.
Person B) 5Ghz is faster because 2.4Ghz is noisy.
Person C) Well, 5Ghz is faster regardless of noise.
Person D) Not if your internet is slow.
So, seems we're all answering something else. I'm expanding on B alone, and you on A-C, I think.
Yep, you got off track. B doesn’t matter in this picture. It’s 2.4GHz btw not 2.4Ghz, and it’s a range from 2.4 to 2.5 with 2.45GHz as the midpoint, and noise varies according to where in that range you are and what’s around. But again it doesn’t matter as it’s not the bottleneck here, so yeah, off track. Yawn.
I’m talking about cases where the broadband is the bottleneck, obviously.
These are people who do not own a microwave or TV (they’re lucky to have a flush toilet) and their nearest neighbor is far enough away that the neighbor’s WiFi is undetectable, so the issues you mention do not apply. Yet they are told by an installer who does a physical visit to the place to put in a dish that choosing the 5GHz WiFi will make a difference. It won’t.
> tell people (specifically my neighbors for example) that they will get faster speeds if they use the “5G” network exposed by their wifi router
Of course you get higher speeds on 5ghz! How is this controversial?
Most people are still on 802.11ac which only works on 5ghz. It's also impossible to find a free channel on 2.4ghz unless you live in the woods so you'd be sharing airtime with the 10 other APs and all their clients on one of the 3 non-overlapping channels in the 2.4ghz band. On 5ghz you can easily run 80mhz channels instead of just barely getting a stable connection on 20mhz.
Near the edge of a signal’s range, its performance and reliability can plummet as signal loss causes trouble.
5GHz has a much lower effective range than 2.4GHz. Thus, there are areas where the 5GHz network will be available, but perform far worse than the 2.4GHz, even with congested 2.4GHz.
(Still worse, in some devices it can cause persistent problems until reboot. My last laptop, a Surface Book, had a Realtek chip, and I was using it near the edge of a router’s range for hours a day last year, and every few days it’d get into a broken state where packet loss was >20% and latency a few hundred milliseconds, even if I moved right next to the router. Connecting to different networks wouldn’t fix it; sleeping the laptop wouldn’t fix it; disabling and reenabling the thing in Device Manager would normally but not quite always fix it; rebooting would always fix it.)
Having separate SSIDs for the two technologies is stupid in itself. In the majority of cases you're better off using the same SSID (and encryption settings) which means the client device can decide which one to use and whether to switch between the two.
There are cases where the client device stubbornly clings on to the 5Ghz network despite the signal being so terrible that no packets come through (more specifically, the packets from the access point to the device are still received, but the packets from the device to the access point are lost), but unless you've identified that you're in this case and can't mitigate it any other way you're still better off using a single SSID.
They're not exploiting anything, they didn't choose the names for cellular or WiFi technologies. At worst they'll say 5G instead of 5 Gigahertz but they're not attempting to confuse or conflate anything.
Think of your internet connection like the water coming into your home. Your 2.4Ghz connection is like the water dispenser on your refrigerator, and 5Ghz is like the hose bib on the outside of your house. If the water authority says the water pressure is enough to fill up a 5 gallon bucket in 1 minute, someone will call up complaining that it takes 20 minutes to fill up a 5 gallon bucket from the water dispenser on their refrigerator.
Your ISP telling you to use your 5Ghz Wifi to test, is like the water authority telling you to test with the hose bib on the outside of your house.
1000^n is only really used by HDD manufacturers to make their storage look larger. Everyone else uses 1024^n. XiB was invented to solve the confusion of everyone using XB but adoption was low and personally I think XiB is more confusing anyways (given people naturally assume XB is 1024^n).
I think what confuses consumers more is byte vs bit, or more specifically, that the capitalisation of the ‘B’ matters.
- most modern GNOME applications and most GUI applications on desktop Linux
- hard drive manufacturers have been sued over this, and US courts have agreed with hard drive manufacturers that 1 GB = 1000 MB
- the International System of Units (SI) and the International Electrotechnical Commission (IEC) both use the decimal definitions (1 kB = 1000 B)
The last major hold-outs are Microsoft Windows and old command-line applications that want to preserve backwards compatibility with any script that might parse their output.
Personally, I don't find the XiB units confusing, as they are the only units with a consistent unambiguous definition. The XB units are much more confusing because you can't be sure whether the 1000^n definition is intended or not.
The thing is, that’s all relatively recent in computing terms. For the vast majority of computing history it’s been 1024^n because that’s how the hardware works — you can’t have 1000^n in a binary system where bytes are multiples of 8. All the 1000^n values you get are where the raw value in bytes is taken and then divided by powers of 1000 but that leads to values that are technically not accurate (eg values that are not addressable).
And here lies the problem, before HDD manufacturers decided to change things up, computer science was already standardised on using 1024^n. Sure there was some outliers in network theory but it was pretty easy because if you needed a precise value then you knew it was 1024^n and if you just wanted an approximate value you could still divide by 1000 in your head. It worked, everyone understood it and everyone was happy.
This whole “1000^n is more human friendly” only appears so because we now have multiple interpretations and people without a tech background making decisions about it. But frankly, if you can’t wrap your head around 1024^n then you’re already in the group of users who honestly don’t need to worry about the precision of getting the scaling right. Those who it does matter for honestly find 1024^n easier.
> And here lies the problem, before HDD manufacturers decided to change things up, computer science was already standardised on using 1024^n.
And long before that, the SI units defined K as 10^3, G as 10^6 etc.. This is how the prefix is used everywhere for every unit, with the one exception of computer scientist playing it the US way.
I personally prefer 1024 as well, but honestly, the scientific side of that argument is a lost cause.
Different operating systems prefer one way or the other, or a mix. MacOS uses powers of 1000, most Linux desktop environments use the correct symbol (1000 and kB, or 1024 and kiB etc), but Unix tools tend to use 1024 and "K". I believe Windows uses the non-standard symbol.
I much prefer to work with powers of 1000. Running "df" on our storage cluster shows
2675230000214900
Although, since my terminal has this font [1] installed, it actually displays like this, with the 2 underlined:
2͟6752͟3͟0͟0002͟1͟4͟900
It's easier to think about 2.6PB than 2.3PiB (how many 200GB files do I have space for, etc).
That’s fine if storage is your only concern, but memory doesn’t work like that. RAM has to be grouped in powers of 8 and assigned in binary. 1000^n doesn’t make any sense at a low level. It was a convention that followed later when HDD manufacturers wanted to make their drives look bigger.
Well it makes more sense to think about memory in pages or at least words (or cache lines if you think performance), disks in blocks and network in packets anyway, memory is very rarely byte addressable natively... not being picky, just to underline that power of two adressable bytes does not sum it perfectly either.
NAND chips are built in same way, but SSD manufacturers sell 256GB(not GiB) drives. That's partially because some SSD uses the difference (256GiB - 256GB) for reserved area for wear leveling. Of course primary reason is to align with HDD.
So even if your files are 200GB, the actual space they consume are not base 10 as you count on your storage cluster, but 200GB plus the last remaining sector that is used but unfilled as the rest of that sector cannot be used by another file.
You can count your sizes in Base 10 but the actual use of space is still Base 2.
If a file is exactly 1000 bits and your sector sizes are 1024... that 1000 bit file is using 1024 bits of space on that drive.
And no, just because they advertise or display drives as having 1,000,000 Bytes = 1MB... it doesn't change how space is sectored out on that drive itself.
I do use 1000^n, but I agree that most people tend to use 1024^n. 1000^n kind of makes more sense since "kilo, mega" etc. are the actual SI prefixes for multiples of 1000s. I don't know who or what caused this chaos but 1000^n is definitely more human friendly.
I feel the problem may be that, unlike just about every other unit in SI, bits are discrete, not continuous. Except in few very specific subfields of theoretical CS, there's no concept of a fractional bit. You can have kilometers and millimeters, you can have kilobits and kilobytes but not milibits and milibytes.
The nature of bits is that of a base-2 system, so using power of 10s for counting them is only superficially human friendly - in practice it's human-unfriendly, because it flies in the face of how bits are used. All hardware and all software groups them by powers of 2, that's inherent to what bits are.
1000^n might be more human friendly but computers aren’t decimal machines and a byte isn’t 10bits. 1024^n technically makes sense as a unit for binary machines that have 8bits to a byte.
Everyone was happy with 1024^n convention in the 80s. The problem was HDD manufacturers got greedy and switched to 1000^n to make their drives sound like they had more storage. Thats what started the confusion.
RFC 1951 (NTP) was published in 1988 and refers to 56k modems. Does a 56k modem operate at 57344 bits per second or 56000 bits per second? Your claim implies the former, but I'm pretty sure it was always the latter.
> a byte isn’t 10bits
It could be. Historically, the number of bits per byte varied somewhat from machine to machine. Many standards used the term 'octets' to avoid ambiguity.
Historically yes. But even as early as the 60s 8bit was the norm. IIRC C then “standardised” 8bits (though ASCII went some way to doing that prior to C).
> The speed you'll get on 3G (and probably even 4G) today is much slower than you used to get when those technologies were new.
Also keep in mind that back in the early 3G days your phone didn't send megabytes worth of analytics data. Back then if an app merely tried to access the internet you'd get a prompt enforced by the OS (and probably a bill from the carrier) and so bandwidth was definitely not wasted on analytics and other "growth and engagement" crap.
In my experience this is partially true (e.g. Carriers reusing one of the DC-HSPA Carriers as a Carrier for LTE or 5G in Band 1 is pretty common around here in Germany).
I suspect the biggest reason for the speed decrease is increased usage though.
Back when 2G/EDGE was current technology I got 100-200kbps there most of the time. Today it is basically unusable as you get timeouts constantly.
IIRC an EDGE Cell could deliver the full data rate to about 8 stations within the coverage area or thereabouts at the same time.
This worked okay back then when there were only a bunch of nerds per cell tower using but it can't cope with current usage density and patterns at all.
At least around here (germany had famously high rates for data traffic at the time when 3g was recent) smartphones and data plans are a lot more widespread than a few years ago and usage patterns got more data-heavy as well.
Even assuming everything else is equal infrastructure-wise, that's not true. The allocations on older techs will shrink, but so will the number of users. The speed you get could be slower but it could also be faster.
In some countries (DE) this is not true. Here they apply the law of the hole i.e. the new technology is inserted where is a hole in coverage. That has a very good marketing value ( look, we have 5G) and regulatory value (look, we have 98% coverage). For the end user it is a disaster: in a given area only one technology might work. But they say in the future we will all be connected.
As far as I know there's nothing in this article that is "wrong", it's just that ultimately when networks support multiple technologies they can choose where to allocate the available bandwidth, and where not to allocate it, and they usually favor whatever is newest.
Looks like the move from 3G to 4G also meant going from "real speeds are 1/5th of theoretical max" to "real speeds are 1/10th of theoretical max". I wonder if that's due to congestion at the phone-to-tower level, congestion at the tower-to-internet level, or due to newer wireless tech just having higher peak performance at optimal physical conditions (e.g. line of sight) that are becoming increasingly hard to actually accomplish in practice?
> I wonder if that's due to congestion at the phone-to-tower level, congestion at the tower-to-internet level, or due to newer wireless tech just having higher peak performance at optimal physical conditions (e.g. line of sight) that are becoming increasingly hard to actually accomplish in practice?
I believe this is entirely configurable by the carrier. As others have pointed out, the primary driver of each successive generation is to support more clients. Increasing individual client bandwidth is a secondary motivation.
I think the primary driver is higher peak performance. With OFDM on 4G and 5G you can have very wide bandwidths for a given tower, and small, but high QoS allocations for voice, or high bandwidth for video streaming. With 3G you had similar flexibility, but the peak RF bandwidth was ~3.84 MHz compared to ~40 MHz of 4G).
This is doubly so with 5G. The maximum allocation is potentially huge from a time and (RF) bandwidth point of view, but an HD video stream is still limited.
The aim is also to increase capacity because of the ever increasing number of mobile devices. Increasing capacity means increasing the number of devices that may connect to a single base station (aka "tower") at any given time.
This was true when moving from 3G to 4G, and this is also true when moving from 4G to 5G, though with 5G there is an emphasis on a very large number of devices with relatively low data rates (in order to accommodate IOT scenarios).
Despite what they want you to believe, 5G is still very far away; it'll require a full rebuild of some mobile networks and their back-end systems.
A term to look for is "5G Core Network", which is basically "5G native". But there's very few providers that have that - simply because there's few hard- and software providers that can deliver the supplies needed.
Disclaimer: I currently work for a player in the mobile networking space. Our software solution can act as a bridge between older and newer "g" systems and implements various aspects of these mobile networks.
For our 5G offering we've just included cURL as a library, apparently (I'm not involved in our core system, just the configuration interface).
In my understanding the diference between deploying 5G-NR-NSA and SA is mostly an software one. You need to have 5G capable RRU heads anyway.
On the other hand from the bussiness and operations side there are two things: (1) 5G-NR-SA does not have any obvious benefits when you still have to support LTE-only UEs (and even GSM-only ones). (2) Many hardware vendors and carriers combine the 5G rollout with bunch of only tangetially related network changes like getting rid of ZTE/Huawei equipment, various "k8s on the edge" initiatives and such.
Perhaps the 5G prophecy hasn't been entirely fulfilled yet, but I am super impressed by what we have so far. Here in Sydney on Telstra in a pretty low density suburb I am getting 400mbps down while indoors (4/5 bars reception).
The other day in an area with 5/5 bars reception area I got 1164/89mbps, while indoors, which is incredible. Latency is between 10 - 20ms.
One thing I think missing from this article is a discussion on range.
The important thing to note with all EM-based communication tech is that the range decreases as the bandwidth increases, due to the inversely proportional nature of wavelength and frequency.
So it's all trade-offs. 5G requires a lot of buildout, as its range is something like 1000 feet, whereas 4g is something like 50,000ft. Said another way, we could've started with 5G before doing 3g/4g etc, it just would not have made economic sense before everyone had a smartphone that they wanted to stream HD videos on which justifies building out a network that needs a tower every 2000ft.
> 5G requires a lot of buildout, as its range is something like 1000 feet
Only part of 5G has the high frequencies with short range. As the article highlights, there are also frequencies similar to 4G, meaning that it's entirely feasible to build a 5G network in the same time frame.
Long range wireless data is subject to a tons of interference and real life troubles, so in reality you never hit those numbers, there will be so many places without proper coverage that those stats are mostly meaningless.
Not to mention it requires a lot of antennas, hardware, power and area coverage. Results will also vary depending on the phone modem.
Real life wireless latency is also quite problematic. If it's never ideal on optic fiber, I cannot predict it's going to be good on wireless.
Not to mention rural, uncrowded areas, mountains, etc.
You are right. In 5G, data packets goes through less tunneling than it was before. Although there is less tunneling, 1ms latency for "end to end" or from user device to the Internet doesn't sound realistic. I have no real life measurements but if I would have to guess, it should be around 5~10 ms.
All this is further confused by the deceptive marketing from carriers. Frequently calling HSPA+ 4G and now AT&T’s 5Ge branding makes for some very frustrating conversations with folks unfamiliar. “Oh, it looks like I got 5G with a software update!”
This is misleading, 2G """speed""" was 0.0096 Mbit/s half-duplex. It was possible to do some urgent work at terminal server for the ridiculous price, tho.
The further improvements were highly dependent how many time-slots network leases to you, up to ~0.24 Mbit/s. Dumbing things down to that nG naming doesn't work very well...
That's it for the TDMA guts, next week – library subscription fees.
This is meaningless without explaining the contention issue wrt. radio and the different strategies the later standards have for handling that. 5G is a range of tech, most of which has not come to market yet (mmwave, full on MIMO) the network core standards are very helpful for telcos - but the way that modern antenna can be made to behave does change what the tech can do in busy cells, but it doesn't change physics!
those speeds mean that each new generation will be substantially faster than the previous generation, because each generation dramatically underperformed what it promised.
so, it works like this: 2G delivered 1G speeds, but it delivered them better than 1G did.
then, 3G delivered 2G speeds, which was much faster than the 1G that 2G delivered!
and so on. it's called "marketing", it's "aspirational", lather, rinse, repeat
I'd like to know how the average battery life and power requirements are of the different standards. For example, I've got a 5G phone but to spare power I sometimes switch to 3G or 4G even though I don't really know if it's really more power efficient. I suppose 5G has a lot of power efficiency techniques built in as well which are not in 3G or 4G.
I'd like to know this too. I can tell you anecdotally that 4G (LTE at least) is a major battery drain if you have very weak 1-bar signal. I work in a basement and my phone is dead by 5pm unless I remember to turn on airplane mode.
You would do well to enable Wi-Fi calling in that circumstance if your provider offers the service. I had a misconception that that service was somehow inferior to getting a “real” connection to the cell network—not so. The cell network does not care how you access it, and more-or-less treats RATs (Radio Access Technolgies) as interchangeable.
> For instance, a connected car travelling on the motorway at 70mph (110km/h) would travel almost 2 meters in the amount of time it takes for a 4G mobile network to respond. The lower latency of a 5G connection will allow mobile technology to be used more safely in cars.
This doesn't have the answer I expected to see, which is that "2G speed" is often used in terms of throttling, and in that case it usually means 128Kbps. (There's no real consensus on what 3G speeds would be in this context, but carriers tend to aim low.)
This article is written about UK networks, where none of the networks advertise data as being available at "2G speed".
The only thing I can think of in the UK that's similar is, the "Always On" offering from giffgaff, an MVNO (https://www.giffgaff.com/help/articles/what-does-always-on-m...). It offers 80GB of data per month at full speeds, then unlimited data at 384kbps.
The UK networks generally offer either a fixed data limit, or truly unlimited data plans.
So are these the basic advantages? Bandwidth and latency? Some TV ads here tout 5G as "comparable to the invention of electricity", which seems to me mind-numbingly ridiculous, but I was wondering if there's some huge benefit that I'm missing.
A big benefit not mentioned is usage in high density environments—think stadiums. The idea there is to stick multiple mmWave antennas over the stands. Those have very short range and high bandwidth (you can kind of think of them like radar that can communicate with you).
I am personally excited for this use case since the base of the ski area that I frequent during the winter is serviced by one macro cell for an entire region including that ski area. On a busy day, the network is useless by 11am apart from calling and SMS (which are prioritized differently).
Ah, yes, I think I heard of these microcells. That is indeed exciting, as my house gets no reception, it would be great to deploy a cell in the area and have reception again.
I've recently had two 5G phones, a Xiaomi Mi 11 Lite 5G and a Realme GT. The Xiaomi could do 20-30mbit/s (slower than my 4G OnePlus 6T at 35mbit/s), while the Realme GT is around 250-280mbit/s. I've kept the Realme.
E.g., the article seems to conflate "Internet-style" measurements of practical round-trip latency over a network path, with "telecom-style" measurements of round-trip latency across a single unloaded radio hop with immediate schedule availability. E.g. it gives a figure of "50ms" for the typical latency of 4G, and "1ms" for the typical latency of 5G. I don't think this is an acceptable comparison.
"50ms" is a believable ICMP ping time from a commercial LTE service, including delay to get a resource block, to sojourn queues inside the network, etc. (I just tried on T-Mobile LTE and am getting 21-69 milliseconds RTT, mean 36 ms, to the next hop.)
By contrast, "1ms" for 5G sounds like a 3GPP "user plane latency" figure (and even then, only for the "uRLLC" version of 5G that is probably only going to be used for more-exotic purposes). This is a lower-level latency measurement made on a single unloaded radio hop. You cannot compare the two.
I think the more-accurate 3GPP user-plane latency figures would be:
- LTE: about 5-6 ms (see https://communities.theiet.org/blogs/426/444 or https://www.techplayon.com/5g-nr-user-plane-latency/ or https://www.artizanetworks.com/resources/tutorials/req_lte.h...)
- 5G eMBB ("mobile broadband" service ): about 3-4 ms (https://www.techplayon.com/5g-nr-user-plane-latency/ , https://www.itu.int/en/ITU-R/Documents/ITU-R-FAQ-IMT.pdf)
- 5G uRLLC ("Ultra-reliable low-latency communication"): about 0.2-0.7 ms (see above)
The practical IP round-trip time with all of these networks is much larger than the above; e.g. ~21-69 ms for LTE, and 22-76 ms for 5G eMBB services (https://www.lightreading.com/the-edge/how-5g-is-pushing-enve...). And even those figures are in the absence of load to or from the UE; if traffic has filled up the queue of the uplink (on the baseband chip) or downlink (in the UE-specific queue on the base station/eNodeB/gNodeB), users can experience latencies of 500+ milliseconds. This is called "bufferbloat" and it comes from a refusal by the baseband and eNodeB/gNodeB vendors to deploy queue disciplines that queue on a per-flow basis, or mark or drop packets when queues are large. (I have talked to many vendors about this, and the prevailing attitude is somewhere between "we are the phone company; we're legally obligated to transmit 99.999% of all packets to the endpoint, no matter how delayed everything gets" and "if there were an actual problem with latencies over our networks, companies would be coming to us like Facebook did and paying us a bajillion dollars to deploy custom traffic rules to let their traffic skip the line; nobody else has offered us this kind of money for a custom traffic rule, ergo there is no problem.")
Similarly, the article seems to somewhat conflate improvements in theoretical throughputs with improvements in the "typical download speed" a user might experience. The latter is not that strongly related to the air interface standard (e.g. LTE or 5G-NR) because any flow's experience depends heavily on how many other flows are sharing the same constrained resource, which in turn depends on (a) how much spectrum has the carrier licensed in your area, (b) how many base stations has the carrier decided to build/lease in your area, (c) how many other customers does this carrier have, trying to download right now through this base station, on this slice of spectrum, at this moment, and of course (d) is the base station actually close to you [do you have good SNR?] or is it spending all its airtime on slower modulation and coding?
Carriers sometimes use improvements in the air interface (which improve throughput-per-Hz) to save money by deploying less spectrum or larger cells. So even if a newer/fancier air interface is x% faster on a per-Hz basis, the benefits may not fully accrue to the users, if the carrier takes advantage of the improvement to deploy bigger cells, or less spectrum per cell, etc. (I once sat in a meeting with AT&T around the time of the 3G-to-4G transition, and in at least this particular setting, they were planning for roughly 1/3 of the improved spectrum efficiency to accrue towards better service for their users, and the remaining 2/3 towards saving resources for themselves. So, yes, improvements in technology do somewhat improve service for users, but... in some cases that may be a minority of the overall benefit.)
For more rants on this topic:
http://blog.keithw.org/2013/07/3g-and-me.html
https://www.marketplace.org/2010/11/05/tech/4g-networks-dont...
And, obligatory plug! In September, the IAB is holding a workshop on Measuring Network Quality for End-Users (https://www.iab.org/activities/workshops/network-quality/), with submissions due August 2. If you care about this stuff, and maybe you know something or have a perspective that could be helpful to the Internet community, please consider submitting to and participating in the workshop.