How Much Total Throughput Can Your Wi-Fi Router Really Provide?

[thiggins]

It's harder than you think for that expensive router of yours to deliver all the throughput the big number on its box promises. We show you why. Read on SmallNetBuilder.

 

[RMerlin]

Regarding Windows 10 band preference with the AX200: it seems to be configurable, although it defaults to "no preference", so who knows what that means for actual behaviour.

 

[thiggins]

Thanks. I don't think octoScope messes with the defaults.

 

[sfx2000]

Would be interesting to see wired thru put with multiple wired up devices...

with and without CTF or other flow offload things

Might find performance concerns with the SoC/Switch setup...

 

[RMerlin]

Would be interesting to see wired thru put with multiple wired up devices...

with and without CTF or other flow offload things

Might find performance concerns with the SoC/Switch setup...

It should be fine. At least in the case of the Asus models he tested, wifi is connected to the rest over PCI-Express, so no bottleneck like the RT-AC87U had when both its 5 GHz radio and its Realtek switch were connected to the rest over RGMII.

The models he tested are also DHD-based, which means wifi SoCs have their own separate CPU.

Flow Cache would be irrelevant to what he actually tested in this article, Flow Cache is a WAN side affair, not LAN side.

(there is no CTF in newer Broadcom SoC, it's replaced with Flow Cache and Runner/Archer depending on whether it's a bcm490x or a bcm675x).

 

[Marin]

Were there any tweaks on the wireless settings or were they left at default settings during testing? Thank you.

 

[thiggins]

Were there any tweaks on the wireless settings or were they left at default settings during testing? Thank you.

From the article:
WPA3 wireless security is used if supported; WPA2 if not. Router defaults were left in place, including standard and "universal" beamforming, airtime fairness (if present) and roaming assist. The only changes I made were to set channels and bandwidth and enable any flavors of OFDMA and MU-MIMO supported.

 

[Easy Rhino]

You know, I don't think I would see an article like this anywhere else. Thanks for taking the time to set it up Tim!

Admittedly, my take-homes from this are simple 1) aggregate throughput is way less than the number on the box and 2) about 1.5gbps is the current max for 5ghz.

 

[Marin]

From the article:
WPA3 wireless security is used if supported; WPA2 if not. Router defaults were left in place, including standard and "universal" beamforming, airtime fairness (if present) and roaming assist. The only changes I made were to set channels and bandwidth and enable any flavors of OFDMA and MU-MIMO supported.

Thank you Tim!

 

[Razor512]

Smart connect has changed over the years in terms of how some routers behave.
For example, Compare the R7000 to the R7800 to the RAX80.

And for extenders/ APs, compare the Netgear EX7000 and the EX8000. The EX7000 will cause issues with some devices, such as the ecobee thermostats and many other devices, as it would try to move them to a different band that they don't support, if you have a bunch of 802.11n wireless security cameras on the 2.4GHz band. But if you do the same setup on the EX8000, it will do a good job of spreading them across the 2.4GHz and 2 5GHz radios, separating them according to WiFi generation, and they will largely not move, and it won't try to steer a 2.4GHz only device to the 5GHz band. It seems like some ongoing development is taking place in how the smart connect software in the router is managing the clients.

 

[thiggins]

2) about 1.5gbps is the current max for 5ghz.

I'm not that positive about that one. The test did not associate two Group 1 (2 stream AX, 0 attenuation) STAs at the start of the test. That would tell you what the max potential throughput was. I'll have to run that test separately.

 

[Amato_C]

The articles was great!
Are there any plans to perform testing on the different OFDMA Resource Units (RU's)? I believe this is the true benefit of 11ax and not the higher speed. For example, devices that support different RU's - how many can attach to a 11ax router? For example, theoretically 9 STA's using 26RU should be able to connect to a 20MHz channel - but does the router do that?

Also, it would be interesting in seeing a mix of different RU's on a single bandwidth and across different channel bandwidths.

 

[digital10]

Not looking good for Netgear, I bet Asus would love to use the above graphs for marketing material.

 

[thiggins]

@Amato_C I don't plan any further OFDMA-specific testing. My conclusion after trying many different methods for about a year is that OFDMA effects are difficult to see without many more STAs than I have been using. It will take years, as it did with MU-MIMO, for chip makers to get airtime schedulers tuned and for there to be a critical mass of AX-capable devices in the field.

 

[thiggins]

@digital10 To be fair, only the RAX120 was a comparable product to the ASUS products tested.

 

[Amato_C]

@Amato_C I don't plan any further OFDMA-specific testing. My conclusion after trying many different methods for about a year is that OFDMA effects are difficult to see without many more STAs than I have been using. It will take years, as it did with MU-MIMO, for chip makers to get airtime schedulers tuned and for there to be a critical mass of AX-capable devices in the field.

Thank you.

 

[Cosmin Cazan]

Very nice test. Would love to see this re-done on 6E.
Same exact test, just add 6E (tri-band AP) and some 6E devices.
Would probably make the best case for a 6E AP Throughput improvement. (and 160MHz is a viable case here)

 

[sfx2000]

Would love to see this re-done on 6E.

performance should be similar...

 

[thiggins]

Very nice test. Would love to see this re-done on 6E.
Same exact test, just add 6E (tri-band AP) and some 6E devices.
Would probably make the best case for a 6E AP Throughput improvement. (and 160MHz is a viable case here)

I plan to have this be part of my standard benchmarks. But I can't run this test on 6E until octoScope upgrades the testbed and, of course, more 6E routers are available.

 

[Krisbi]

What about the CPU utilization while running your tests? Does it affect the results at all?