Is this it? (Lower than expected wireless throughput)

[KSoze]

Hi:

I have an Asus AC66U router since a couple of years and I've been happy with it. Recently I've started using it for Steam In-Home Streaming between an HTPC connected to the Gigabit port and a PC using a Netgear A6210 Adapter. Router and PC are ~9m apart on different rooms. It seems that the max throughput I can hit is around 220Mbps (tested with iperf, Steam reports ~240Mbps of estimated bandwidth) and it seems to be the normal operation speed for thes router (checked a few benchmarks online).

The thing is, what's with that? Checked a few other routers, like the newer AC88U and they're giving similar results (https://lesterchan.net/blog/2015/11/30/asus-rt-ac88u-router-review/ for example, around 220Mbps too on wired to wireless). So to my question ... is that it? What's with the theoretical speeds? Advertising 3167 Mbps of speed and getting 220Mbps over LAN with expensive USB adapters? Am I missing something? Am I doing something wrong and getting the wrong results? I'm really lost right now.

Thanks in advance!

 

[A.D.]

I have an Asus AC66U router since a couple of years and I've been happy with it. [...] Checked a few other routers, like the newer AC88U and they're giving similar results (https://lesterchan.net/blog/2015/11/30/asus-rt-ac88u-router-review/ for example, around 220Mbps too on wired to wireless). So to my question ... is that it? What's with the theoretical speeds? Advertising 3167 Mbps of speed and getting 220Mbps over LAN with expensive USB adapters?

Assuming you're encrypting your WiFi, the WiFi to LAN performance highly depends on CPU computational power. Even if you don't, packet sizes need to be adjusted, unlike at the (unmanaged) LAN switch where heavy traffic doesn't impress any well-programmed system. On my RT-AC66U, I see CPU load jump from 2 % to 60 % if I use iperf over a 150 MBit/s WiFi connection. And of course, there's a net limit to the utilisation of the WiFi gross bandwidth, as with any data transmission method.

The advertised 3167 MBit/s you mention refer to the gross maximum theoretical bandwidth on both the 2.4 and 5 GHz bands in parallel (AC3100 mode, not possible with two antennas), and it's certainly not advertised for the AC66U.

Nothing personal against lesterchan.net but I cannot see from the test whether it was conducted in a professional way. The pictures are superb but not enough hard facts for my taste. Also, the author claims that this was the most expensive router he ever came across -- not a good sign. I'd prefer a test from a person who has wider experience (e.g. to be surprised about the results and figure the details). I doubt that guy used a four-antenna client. It sounds like he used some MacBook. I think all of them have only two antennas, so the four antennas of the router are pretty much useless. I'd try to find additional references from professioanls if I were you. I am positive the AC88U can handle a lot more throughput.

If you're dissatisfied with your AC66U, there are a few other not immediately obvious aspects, and you didn't supply any details so far.

• Whcih iperf command lines did you use (e.g. window size)
• What's the negotiated WiFi speed
• Are slower devices using the same access point
• What are the signal strenghts
• Is your WiFi channel shared with your neighbour (amd remember that the ASUS may not know this while the PC is disturbed in a relevant way)
• What is the USB data rate of your PC
• Does the PC have to do the WiFi decryption (I would say yes) and how much CPU power is available for that

Etc pp.

Personally, I use the RT-AC66U as an Access Point and Media Server only and it's doing an excellent job there. I wouldn't use it as a router (at least not at the same time) because I see its its performance as too limited. I doubt you'd see full VPN, WiFi and media server performance at the same time. Virtually nobody is testing such scenarios, though -- magically, everybody seems to believe there's one IronMan per feature enclosed in that box. I still feel I got excellent value for the money.

 

[bbsc]

is that it?

No.
RT-AC66U can do much more.
Near 600 Mbps LAN to Wireless in my case.
RT-AC66U, BCM943460MC adapter, near 8 m distance, different rooms.
Sorry for russian langpack on the screenshot but the figures are understandable.

 

[KSoze]

Thanks for the responses.

@A.D.: Thank you very mucho for this response. I linked that page merely as an example of benchmarks I came accross while researching if this was normal. I did the tests with "iperf -s" on the wired client and "iperf -c 192.168.x.xxx" on the wireless client and vice versa. There are no slower devices on the network, just this one. I well know that it won't ever reach advertised speeds but the speeds I'm getting seem a little low. I noticed that cpu spikes to 60-70% when running iperf. The results I'm getting with iperf correlate well with file transfer speeds between these 2 clients (about 30-32 MB/s). This network is using encryption (WPA2+AES). About the USB bus speed: I have the netgear A6210 plugged in to an USB3 interface but I'm afraid it's being limited to USB2 speeds (something to do with the driver + asmedia usb 3.0 controller I think), but, USB2 should at least get to 300mbps right? I do have too an Asus USB AC56 but it was too unreliable in Windows 10 (and I remember I was getting similar speeds with this device too). The adapter connects to ~800Mbps+ in Windows. I'm currently using it as a Wireless Router connected to another router provided by my ISP which is working in bridged mode. Would I notice a significant performance boost if I switch the mode to AP and use the router the ISP provided as a modem-router? I could kill the wifi radio from the modem router and I guess wouldn't notice that much change in how I manage the network (I would have to configure ports and such on the modem router instead of the Asus AC66U but doesn't seem that big of a deal). As for media servers, I have other devices on my network acting as media servers so I only need the AC66U to provide a wireless internet connection and a means to stream content from one device to another.

@bbsc: Thanks. So I should be getting more speed then. Are you using some kind of custom firmware? Settings? Did you try any other wifi adapter? Distance between my router and adapter are too around ~9m with 2 walls in between (room to room with a hallway).

Thank you both for the replies!

 

[bbsc]

Are you using some kind of custom firmware? Settings? Did you try any other wifi adapter?

Using Merlin, 380.57 and 378.56_2.
The settings are mostly default, nothing special.
Cannot say anything about external adapters because I did not try any with this router.
Have tried 2T/2R adapters: Intel 7260ac and Broadcom BCM94352HMB.
Intel was able to do near 55 MB/s (when it was operational at all), Broadcom - up to 65 MB/s.

I would say BCM94352 seems to be the optimal solution for laptop.

 

[KSoze]

@bbsc: 1 last question if you don't mind. Are you using the AC66U merely as an AP or as a router? I'm trying to figure out if I would notice a performance increase if I disable the routing capabilities and just use it as an access point. I was planning on switching to Merlin over the weekend too.

Thanks!

 

[A.D.]

@A.D.: [...]

If the WiFi connection is at ~ 800 with two different wireless USB WiFi sticks and both show the same limited bandwidth, then I'd check USB throughput. No, not all combinations of USB hardware and driver give you maximum performance. You can try another port, also -- my PC's front and rear USB ports are driven by different hardware components. => I suspect it's your PC rather than your AC66U. If there's a problem at all -- many devices will never provide you with more than a quarter of the gross bandwith, even at low speeds. But we're already two in this thread who see a much better percentage.

Regarding iperf, there are options for the TCP packet and window size. Both increase throughput in my scenarios.

Regarding performance, I'd see any modified firmware which is not based on the original ASUS source as suspicious of using the hardware less optimal.

If you find more blogs mentioning the ~ 220 throughput of the AC88U, remember that Bloggers are often just that: Bloggers. It's not possible to conduct a meaningful test without understanding the relevant details. Of course the marketing of the astronomic figures can be criticised. But I'd rather bet that bloggers test a 4 x MIMO against a 2 x MIMO (and don't mention or are even unaware of it) than that ASUS launches a 4 x MIMO device with a 2 x MIMO performance. => Look at the pictures on the blogs, get the figures from the pros (and I don't mean the pseudo pros with their ad-overloaded pseudo sites).

 

[KSoze]

Thanks for the replies A.D.: I really appreciate them. The thing I find is that wireless adapter drivers for Windows 10 seem to be pretty rough. Latest Netgear A6210 drivers have this in the changelog: "Always set U2 mode in ETRON & ASMedia USB 3.0 Host control". I guess it means it is stuck at USB2 performance when in conjunction with an ASMedia USB3 host (my case) so I looked some more into it but couldn't find anything meaningful. A pity I don't have any other device to try.

Sorry if I'm bothering you but you mentioned you use it as an access point. Do you think I would notice a significant performance increase if I switch from router to just AP? I'm currently using stock firmware but was planning on switching to Merlin over the weekend.

And again, thanks for your replies.

 

[bbsc]

Are you using the AC66U merely as an AP or as a router?

Router, but all download tasks were stopped on all clients when testing, of course.

 

[A.D.]

Using Merlin, 380.57 and 378.56_2.

Does the 380.57 work for you? It doesn't drive my LEDs correctly, and elsewhere it appears to lead to some WiFi <-> WiFi problems.

I reverted to merlin 378.56.2. Which hardware revision do you have?

 

[A.D.]

Do you think I would notice a significant performance increase if I switch from router to just AP?

If you never see 100 % CPU usage (given that no USB drive is scanned; that's done on low priority and won't hurt) I wouldn't expect that. If you use VPN, I'd say, maybe. There's a test out there saying that at least the first hardware revision with initial firmware had a poor PPPoE WAN performance, but if your 66's WAN is connected to a router rather than a modem, that may not matter -- it might be that the routing performance in general is poor, or that the PPPoE performance is poor, and the information might be also completely outdated if you have a newer hardware revision and/or firmware.

I'm currently using stock firmware but was planning on switching to Merlin over the weekend.

I recommend merlin 378.56.2 for now. The stock firmware wouldn't spin down my USB HDD. Some other additions are also very nice to have.

 

[KSoze]

Guess I'll stick with router mode, switch to Merlin FW and see if I can play around with the usb controller drivers for now. Thank you all for your replies.

 

[System Error Message]

below 300Mb/s is wireless N speeds, sounds like you werent using wireless AC. Practical wifi speeds of good wifi chips would be up to 50% of the rated link speeds so if you connected using a 2 channel AC wifi client you would expect to see between 400-500Mb/s. I have before gotten up to 90% of the link speed with wifi using the AC68U but thats with forcing everything, overloading the client CPU in order to push more data but not without significant packet losses. The AC87U wouldnt be able to do that even with MU-MIMO because it is bottlenecked internally.

If you have something like the AC88U with 4 MU-MIMO streams and have 2 clients both with 2 MU-MIMO streams you would be able to fully use the wifi and see about half the link speeds. This means you can get around 1Gb/s of practical wifi on it if you analyse the results correctly. For example 400Mb/s from one wifi client to another would mean 800Mb/s passing through the wifi radio which would be almost half of the 5Ghz AC radio.

It is a good thing that the AC5300 can combine 2 ports so it helps to prevent bottlenecks.

Another important thing to note is interference, surrounding wifi, metal cages and so on can decrease speeds and temperature does also affect throughput and stability. A professional test would be done in a metal cage to block all outside interference and professional testers usually put 2 of the same routers in bridge mode just to see the capability of the radio chip and firmware and reduce the number of variables. Auto channels usually interfere with each other and manually set channels. I've seen a number of APs or wifi routers always switching to the same channel i use when there are other free channels and it really decreases throughput.

 

[bbsc]

Which hardware revision do you have?

3.20, AFAIR.
Have no complaints about 380.57. The LEDs.. Really I do not care, but I'll have a look.

 

[A.D.]

The LEDs.. Really I do not care, but I'll have a look.

The LEDs themselves may not be too important, but I strongly dislike both reasons I can think of as the root cause.

 

[A.D.]

Guess I'll stick with router mode

By the way, if there's no reason to operate it as a router, then it would make sense to switch to AP mode. Routing requires more CPU cycles, will not improve latency between subnets, and will also not avoid difficulties with all the (often simply implemented) auto detection stuff in apps (and nowdays even in desktop applications). It's a five port LAN switch in AP mode (unlike some products from competitors).

 

[Puppa]

@KSoze I didn't see you mention a frequency band. You'll see lower throughput on 2.4GHz, and 220Mbps isn't too bad for that. Got a good AC 5GHz client around?

The 68U CPU is only 800MHz, nearly half the speed of the latest ones, unless you were lucky enough to get a recently manufactured one that does 1GHz. (3100, 88, and 5300 are at 1.4GHz).

The 3167 number is marketing hype, and they all do it. The same way my 1TB hard disk is missing about 10% of the capacity of a true 1TB. 3167 represents the sum total of throughput on each wireless band at its highest rate, so you'll never see this on a single client without multiple radios, and fast ones at that. If the tool you are using cannot measure multiple connections, it'll be hard to see the throughput.

Sent from my SM-N920V using Tapatalk

 

[bbsc]

You'll see lower throughput on 2.4GHz, and 220Mbps isn't too bad for that.

It's not a very good result even for 2.4 GHz band.
RT-AC66U can show up to 330 Mbps LAN to Wireless throughout in 2.4 GHz band when working with a good adapter.
I can show a screenshot with 41 MB/s real file transfer speed.

I think, the result is determined by Netgear adapter.
It's 2T/2R adapter but it's too small for "multipath" and that's why it works as a good 1T/1R adapter only.
It's only my opinion, though.

Other tests of A6210 and RT-AC66U on tom'sHARDWARE.

 

[KSoze]

By the way, if there's no reason to operate it as a router, then it would make sense to switch to AP mode. Routing requires more CPU cycles, will not improve latency between subnets, and will also not avoid difficulties with all the (often simply implemented) auto detection stuff in apps (and nowdays even in desktop applications). It's a five port LAN switch in AP mode (unlike some products from competitors).

And if I switch to AP mode, would I have to switch the wired connections currently connected to my AC66U to the ISP router or would it work perfectly as they are (connected to the AC66U I mean) without a hit on performance? I currently have 2 consoles and an rPi connected via LAN to the AC66U and it would be a pita if I have to change/move the wires.

Alas, I can't just switch the modem mode from bridged to router without calling the ISP and waiting 2 days for them to make the switch so just "testing it out" would mean a whole week.

Thanks!

 

[A.D.]

And if I switch to AP mode, would I have to switch the wired connections currently connected to my AC66U to the ISP router or would it work perfectly as they are (connected to the AC66U I mean) without a hit on performance? I currently have 2 consoles and an rPi connected via LAN to the AC66U and it would be a pita if I have to change/move the wires.

As I said, the AC66U will be a 5-port LAN switch in Access Point mode, and this will make the uplink faster in the AC66U. Your cables can remain where they are.

Alas, I can't just switch the modem mode from bridged to router without calling the ISP and waiting 2 days for them to make the switch so just "testing it out" would mean a whole week.

I was somehow thinking your other device was a router already. If it's some no name thingy I'd recommend against changing its mode, especially if you're happy with WAN performance and don't encounter 100 % CPU phases on WAN - LAN - WiFi traffic. I'd rather stick with the current setup. Sorry if my suggestion led to confusion.