Question about 2.4GHz vs 5.0GHz


#1

Not sure which Category this falls in but it's a general question about differences between 2.4Ghz and 5Ghz.

The obvious things like: higher frequency, less wall penetration, higher bandwidth, etc. of course I know them and for those who don't can look it up on the internet so i don't need to get an answer what's repeated all over the internet already anyway. I'm looking for more depth insight in practical differences.

Both can do 100Mbps easily, so shouldn't be a bottleneck for anything under 20Mpbs I suppose.

Still, here is my setup:
Going from south east Asia over VPN to Europe.
using speedtest.net to Dutch providers.
The internet itself supports (checked with local servers over speedtest) speeds over 40Mbps

I also tested local LAN for 2.4ghz and 5Ghz both easily break the 100Mbps barrier.

So you'd expect the bottleneck be somehow the VPN connection from Asia to Europe and thus there shouldn't be a significant change in speed by choosing either 2.4 or 5 Ghz, right?

Yet I did some testing and found out there are differences, between the testing I kept changing
the 2.4Ghz and 5Ghz to be sure I wasn't doing them all in 1 row (to minimize any effects of
differences between the time the tests were done):

It clearly shows in the upload speed the differences are below anything to break my head over.

But between the download there is a clear difference in speed.

Again, the tunnel to European Server is the same throughout the test. the only thing that varies
is the WLAN frequency.
(yes of course I made sure I was the only one on the WLAN)

How can it be, that both wireless network I can copy locally well over 100Mbps so shouldn't be a bottleneck,
there is still such a huge difference to Europe?

I'm sure there are much more technical detailed differences then just bandwidth and frequencies that have a reason for this. i was thinking, maybe the 5Ghz network changes something in packets which VPN might prefer? As I do know VPN over wireless isn't something VPN likes.

Also I know high ping is a killer for VPN, but the PING is roughly identical as well.

So, anyone with an explanation?

and please no:"no difference, just your ISP" I did this test to find out numbers for something I noticed already
since I moved here. Since I moved it "felt" like 5Ghz was faster then 2.4Ghz but kept waving it off myself as: BS, both 2.4Ghz and 5Ghz can easily do 10 or even 20Mbps so that can't be it.

But it kept nagging me, so that's why I did these tests to get numbers and behold...

Explain it to me please? :slight_smile:

ADD: I did write the numbers nice and clean ordered properly in the edit field. But somehow when I click post, save edit, etc it totally screws up the arrangement so I uploaded a screenshot
results


#2

You don't mention your router here, but aside from the interference issues (where 5 GHz has an easier task) or indoor range (packet retransmission at higher ranges, especially for 5 GHz), multi-core routers might see a variance in terms of which core needs to service which devices (ethernet, wlan cards) while also covering de- and encryption for the VPN uses.


#3

While it's likely the above sentence is in relation to your VPN speed, just in case it wasn't:

  • 20mbit/s is extremely slow (2.5MB/s), as is 100mbit/s (12.5MB/s). 2.4GHz N maxes out at 450mbit/s with HT40 (56.25MB/s), whereas 5GHz AC maxes out at 866.7mbit/s with VHT80 (108.34MB/s) [not sure of the throughput with VHT160, likely ~1.7Gbit/s].
    • If you're strictly using the router for internet access only, and not for any LAN side traffic, only having a 100mbit/s ISP plan, 100mbit/s throughput wouldn't make a difference, but you will notice it with any LAN side traffic.

  • 2.4GHz supports a maximum HT40 channel width, however many run HT20 due to SSID congestion, of which halves the throughput. 5GHz supports a maximum channel width of VHT160, with any 5GHz router manufactured after 2010 supporting a channel width of at least VHT80.

#4

you quote theoretical maximum, effective 100mbps on 2.4 ghz is not on the slow side.


#5

agreed that your numbers suggest there maybe a difference, but i think the variance is too high and the samples not enough to support the conclusion that there is.
especially since you are benchmarking a tunnel that goes around half the globe, throu probably hundreds of routers and other tunnels ...


#6

No, a quality N wireless card will get 450mbit/s (Intel N6300 for example), just as a quality AC card will get 866.7mbit/s (Intel 7200AC for example).

  • 100mbit/s on an N wireless card is extremely slow, whereas for G it would not be, however it's unlikely any users are still using G wireless cards since they were obsoleted <2010.
    • N should be getting ~300mbit/s with 2.4Ghz HT20, ~450mbit/s with 2.4Ghz HT40, and ~600mbit/s with 5Ghz VHT80 (it may be HT40, as I haven't used my N6300 card in a while)
      • 2.4Ghz channels would see the best performance on channels >36 - <72
      • 5Ghz channels would see the best performance on channels >100, preferably in the 150s

#7

Throughput on wireless lan is related to both the modulation, and the congestion. 2.4Ghz is often quite congested, and so you expect to be sharing the air-time with neighbors and soforth. 5Ghz having a much shorter effective range means you have far less congestion. I think this is most likely the source of the effect you are noticing.

I note not a single case of 2.4Ghz having higher measured speed than 5Ghz so I don't think that's really the case here, though it's a good point that variation in measurements makes quantification of these things more difficult. I could run a statistical model here but it'd be a waste of time, 5Ghz is definitely faster in this data.


#8

in your case 2.4 and 5 GHz perform the same when not tunneling to europe, e.g. testing to your regional servers. so there should be no interference issue but rather your device chipset support. you did not mention the hardware specs so assuming your 2.4 is wisoc and 5 pci device it is likely possible there is some load on CPU preventing you from getting full speed. also different drivers have their stack tuned differently which also impacts throughput and other connection factors under different circumstances


#9

Thank you all for your opinions and trying to explain what I experience.

Out of my stomach I'd assume dlakelan might be right. But no one else suggested this, so could others confirm this`?

"Throughput on wireless lan is related to both the modulation, and the congestion. 2.4Ghz is often quite congested, and so you expect to be sharing the air-time with neighbors and soforth. 5Ghz having a much shorter effective range means you have far less congestion. I think this is most likely the source of the effect you are noticing."

Can others agree on this explanation? Because I indeed have over 20 different 2.4Ghz WLAN here but in the 5Ghz range.

So congestion might have a huge impact on data transmission in particular VPN?

This comment by Fuller: ' but i think the variance is too high and the samples not enough to support the conclusion that there is.'

Was actually already thinking myself that the huge variation in bandwidth for either frequency might actually proof the number of tests are below anything to make a proper assessment of the situation and might just all be coincidence.

Also the LTE connection I'm using. I keep pinging the other side of the world now for 72 hours non stop and I noticed ping times can vary from 300ms up to over 3000ms.

The connection in complete is from a linksys WRT1900ACS at home, over an LTE modem,
from LTE goes into the fibre cable between the Island here and the Asian mainland. So from the Island it goes to Manilla, from there to Hong Kong (telstra network) from there it goes to the UK, no idea what's between as it looks like there is a direct cable between hong Kong and UK which I find hard to believe but that's what online tracerts show me and from UK it goes to the Netherlands where the other VPN (WRT3200ACM) is located.

Btw, a direct connection VPN over LAN between 2 routers with highest encryption I'm using I get speeds over 100Mbps with these routers and if used strong intel CPU's highest encryption openVPN i get speeds over 300Mbps but I know these speeds are impossible due to high latency which is a huge killed for VPN speed.

So maybe the fluctuations in latency of the LTE network might also cause some of these effects.

anyway, local here on the Island into the fiber network I get speeds using LTE over 50Mbps.
Local Wireless both 2.4Ghz and 5Ghz I get speeds over 100Mbps
between the 2 routers locally both I get 100+ Mbps
to Europe without VPN both 2.4Ghz and 5Ghz I get speeds over 30Mbps
using VPN over 2.4Ghz gives me at least for a few tests a clear lower performance compared to 5Ghz.

So my best bet is, somehow 5Ghz is more VPN friendly but maybe it's not and the point that congestion might actually have some kind of effect on the VPN performance is the most likely case here apart from Fullers correct mention that maybe if I run 10.000+ tests I find out there's actually no significant difference really.


#10

The big mystery is why local bandwidth can get to 100mbps. If interference is the issue.

VPN requires substantial CPU, and the driver's may well be more cpu efficient on 5ghz compared to 2.4ghz. this seems like the likely answer. It's even possible that interference affects the CPU load, but without the VPN you wouldn't have CPU issues so it doesn't show up on local xfer.


#11

And how accurate is the LUCI statistics on CPU load?

Because neither CPU0 nor CPU1 ever hit over 20% on my router.

Even when using openVPN in a LAN 1Gbps connection the max throughput I get then for these routers is around 100Mbps but CPU load for either CPU never spikes in that scenario over 50% but I think it's fair to assume the bottleneck is the CPU in that case.

And yes I know 50% indicates maybe that's just because you have 2 CPU and openVPN using 1 CPU for 100% and thus you see 50% but the graph actually does show 50% for CPU1 and 3-4% for CPU0 in that scenario so the total load of both cores summed up would be 27-28% of the 100% performance if you count in both cores.

So I'm not sure this CPU load tells me everything correct either.

But yeah, it's obvious 2,4Ghz goes over 1 core and 5Ghz over the other, according to the statistics and indeed maybe VPN is using the same core as the 2,4Ghz so that indeed might be the problem..

Maybe CPU load is indeed not the bottleneck, how well does the CPU on WRT routers interacts with the cache and ram? Maybe that's a bottleneck? Any way I can test this..


#12

top -d 1 and look for "idle %" will tell you how much idle time you have available. if that gets near zero then you're running out of CPU.

installing the irqbalance package might help you balance interrupt load among the two CPUs.


#13

i totally forgot that my last compilation. Indeed I had very good experience with this in the past! Thank you