Low throughput on Asus RoG BE-92 USB adapter

[Jasperx]

Hi, everyone, new here. This looks like the place to be for anything Asus related though, so I figured I'd ask for some opinions here.
Basically I recently got the Asus BT8 WiFi7 router and the Asus RoG BE-92 USB adapter (not the nano version, I already saw the topic for it and what a scam that thing is. I mean a wifi7 adapter with a usb 2.0 interface, seriously? I wanna say wtf. Can I say wtf? But anyway, this one is marketed as BE6500 - with a 5 gig usb connector and admittedly only 160MHz channel support which I was aware of and fine with at the time of purchase). Yesterday I found out however that its maximum ceiling speed is apparently only 2880Mbps, according to Dongknows. So even with MLO active, it still can't fully utilize the usb3 port bandwidth. Now that part is just false advertising imho, but that's not why I'm here. Just leaving it as a side note / customer beware thing in case anyone comes across the thread and is thinking of buying one (it is a $100/€ adapter after all and by the looks of it there are better offerings out there). If anyone actually managed to get more than that with MLO active please do report back. I tried placing the BT8 in the place of my router but the signal was pretty bad and I wasn't breaking 2882Mbits even with MLO active on the official firmware (the actual throughput was 300ish I wanna say? I don't remember but it was bad. Could've even been 30 lol).

But anyway, I'm here because I'm only getting 500-700mbit throughput instead of the expected 1.2/1.4gbit, no matter what I do.

First a little bit of backstory. My internet connection is 2gbits. My ISP router is a ZTE F8648P 4x4 ax ruter and as a freebie I got a ZTE H3601P ax 2x2 repeater which is EasyMeshed to the main unit. I live in an apartment one story above and the signal from the main 4x4 router on 5ghz is horrendous, which is why I need a repeater. The 2x2 ax repeater (mu-mimo capable) is sitting ~2 meters below my pc but there's a ceiling in the way. Windows reports a link speed of 2401/2401 on 5Ghz and the signal strength is excellent but fast.com isn't breaking 600Mbits for the most part, sometimes it dips into the low 700s. So I thought fine, there's a ceiling in the way, I'll just blame it on signal attenuation.

Which leads me to the BT8. To be frank and honest, I only got it because it's the only WiFi7 router with OpenWrt support (in addition to the BananaPi-R4 but that's irrelevant right now). The official Asus firmware doesn't support wireless repeater mode (yes I know it cuts the throughput in half but sometimes it's just the go-to option, like in my case where a cable isn't an option). I totally consider that a cash grab since you're basically forced to buy another unit to form a mesh network (which I don't really need but kinda want anyway, just not in the manner Asus envisioned). I loathe the idea of proprietary tech in general, mesh networks being one of the worse offenders in that regard over the recent years. I mean we already have 802.11s, there's EasyMesh, I don't want to use Asus' AI Mesh (or whatever other vendor/brand proprietary mesh for that matter) they're trying to lock us in to. Eventually I'd like to get a couple 4x4 units and build a 802.11s mesh with OpenWrt but that's besides the point.

Long story short, I placed the BT8 with OpenWrt installed in the hallway directly above the main ZTE in wireless repeater mode with a 5Ghz ax backhaul (the BT8 is 3x3) and I have it broadcasting a 6GHz network in my apartment. Windows reports a link speed of 2882Mbits and the signal strength is excellent (full bars on the taskbar). So I installed iperf3 on the router and checked the numbers.

iperf3.exe -c 192.168.100.100
Connecting to host 192.168.100.100, port 5201
[  4] local 192.168.100.93 port 2148 connected to 192.168.100.100 port 5201
[ ID] Interval           Transfer     Bandwidth
[  4]   0.00-1.00   sec  58.0 MBytes   485 Mbits/sec
[  4]   1.00-2.00   sec  68.8 MBytes   576 Mbits/sec
[  4]   2.00-3.00   sec  72.8 MBytes   611 Mbits/sec
[  4]   3.00-4.00   sec  74.5 MBytes   626 Mbits/sec
[  4]   4.00-5.00   sec  75.6 MBytes   634 Mbits/sec
[  4]   5.00-6.00   sec  75.0 MBytes   629 Mbits/sec
[  4]   6.00-7.00   sec  71.4 MBytes   599 Mbits/sec
[  4]   7.00-8.00   sec  73.8 MBytes   619 Mbits/sec
[  4]   8.00-9.00   sec  73.1 MBytes   613 Mbits/sec
[  4]   9.00-10.00  sec  72.2 MBytes   607 Mbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bandwidth
[  4]   0.00-10.00  sec   715 MBytes   600 Mbits/sec                  sender
[  4]   0.00-10.00  sec   715 MBytes   600 Mbits/sec                  receiver

There's like half a foot of a wall blocking direct line of sight to the adapter so again, thinking it was attenuation I brought the BT8 one meter away from the adapter and ran iperf again:

iperf3.exe -c 192.168.100.100
Connecting to host 192.168.100.100, port 5201
[  4] local 192.168.100.93 port 2155 connected to 192.168.100.100 port 5201
[ ID] Interval           Transfer     Bandwidth
[  4]   0.00-1.00   sec  53.6 MBytes   449 Mbits/sec
[  4]   1.00-2.00   sec  66.6 MBytes   559 Mbits/sec
[  4]   2.00-3.00   sec  67.4 MBytes   564 Mbits/sec
[  4]   3.00-4.00   sec  71.0 MBytes   596 Mbits/sec
[  4]   4.00-5.00   sec  75.8 MBytes   636 Mbits/sec
[  4]   5.00-6.00   sec  70.6 MBytes   590 Mbits/sec
[  4]   6.00-7.00   sec  70.9 MBytes   596 Mbits/sec
[  4]   7.00-8.00   sec  72.9 MBytes   612 Mbits/sec
[  4]   8.00-9.00   sec  74.4 MBytes   624 Mbits/sec
[  4]   9.00-10.00  sec  77.2 MBytes   649 Mbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bandwidth
[  4]   0.00-10.00  sec   700 MBytes   588 Mbits/sec                  sender
[  4]   0.00-10.00  sec   700 MBytes   588 Mbits/sec                  receiver

Exact same numbers more or less. I'd be expecting something in the 1.2/.1.4gbits in virtually ideal conditions. In hindsight, I should've been getting those numbers even with the ISP's repeater since it's mu-mimo capable no? The main router broadcasts at 4.something gbits. The repeater catches that at 2400 and rebroadcasts it at 1200 which my client should be getting. Roughly. Same with the BT8 except the BT connects at 3602mbits to the main router since it's 3x3. But I should be getting at least 1.2 (if not the "full" 1.4 on 6GHz)? Fast.com did spike to 1.1/1.2/even 1.4 but it was for mere seconds in the dozens of tests I've done. 99.99% of the time I'm getting literally half the speeds I should be getting. And yes everything is using 160Mhz channels.

Any ideas? Any owners of the same adapter getting better speeds than me? There's a guy over on openwrt forums that's getting a constant 1.6gbits with a Netgear A9000 (uses a mediatek 7925 chip), also with openwrt firmware installed on his bt8 so it doesn't sound like it's an openwrt thing (his ubuntu pc with integrated qualcomm wifi7 gets 500mbits-1.3gbits, but that can easily be blamed on the linux drivers). And I'm not using a qualcomm's chipset and I'm using Windows 11 24H2, not linux. So I dunno, any ideas? :/

 

[jzchen]

Not sure what is going on (it's the middle of the night here) but there's an interesting document on eUSB2V2 Supplement to USB 2.0 Specification that you might want to read on USB.org.

 

[Jasperx]

Not sure what is going on (it's the middle of the night here) but there's an interesting document on eUSB2V2 Supplement to USB 2.0 Specification that you might want to read on USB.org.

I'm sorry but I don't see how that applies to this case. The adapter in question is usb 3.0, not 2.0. If it were 2.0 I wouldn't be expecting more than 400Mbit/s (that's a filthy trick they pulled with the nano version). The one I bought should be pushing 1.2Gbit/s though...

 

[degrub]

USB, like ISP speeds are "up to ....". Remember that is the hardware link rate which is not the same as the bandwidth.
Never run a speed test on the router itself. It is using cpu cycles and it can easily become saturated. Use a faster PC connected to the LAN and , if you want wifi. At the very least, it is a more realistic metric. The spikes in test bandwidth are likely buffer based and can be ignored. If you are using windows, the iperf site indicates you should be using iperf 2 rather than 3. Also, you should use the -p switch on the invocation with a value of 6-10 parallel streams to try to saturate the link. Are you sure the radios connected as 3x3 ? Just because the spec says they can, doesn't mean they did.

 

[jzchen]

I'm sorry but I don't see how that applies to this case. The adapter in question is usb 3.0, not 2.0. If it were 2.0 I wouldn't be expecting more than 400Mbit/s (that's a filthy trick they pulled with the nano version). The one I bought should be pushing 1.2Gbit/s though...

As of (roughly) August 2024 USB 2.0 can support up to 4.8 Gbps PHY. (If I didn't see "filthy trick" I probably would have ignored the 400 Mbps concern)...

4x4, 3x3, 2x2, I'm afraid I have poor understanding I confess...

 

[Jasperx]

USB, like ISP speeds are "up to ....". Remember that is the hardware link rate which is not the same as the bandwidth.
Never run a speed test on the router itself. It is using cpu cycles and it can easily become saturated. Use a faster PC connected to the LAN and , if you want wifi. At the very least, it is a more realistic metric. The spikes in test bandwidth are likely buffer based and can be ignored. If you are using windows, the iperf site indicates you should be using iperf 2 rather than 3. Also, you should use the -p switch on the invocation with a value of 6-10 parallel streams to try to saturate the link. Are you sure the radios connected as 3x3 ? Just because the spec says they can, doesn't mean they did.

I get it. The hardware link rate is "supposed to" be 5.0Gbit/s limited by the USB 3.0 bus (plus overhead, hopefully needless to say). The advertised bandwidth is 6500. I'm well aware I won't be achieving near those speeds. Dongknows claims the adapter's ceiling rate however is limited to only 2882 Mbit/s. Fine, let's accept that dick meet ******* move too. WiFi by its nature is half duplex, so I'm only getting 1400 instead of the 2800. And then that 1400 gets further attenuated by obstacles. I get all that. Except there are no obstacles in this case.

Point remains that with the router sitting one meter away from the adapter I'm getting 600s/700s from the speedtest instead of the expected 1.2/1.4Gbit/s I should be/people are getting with normal 6e adapters.

Point taken on not running iperf on the router, and yes, the router and the repeater are connected with a 3x3 link. The usb adapter is only 2x2 though. I'm just saying, under these conditions I should be getting twice the throughput this usb adapter is giving me.

Guess I'm left hoping that mlo for mt76 drivers, once implemented into the linux kernel/openwrt will help me slightly boost the throughput.

I can't even source any 6e adapters locally, best I'm finding are some ax1800 (which are also in the +50€ range) and I don't really want to "downgrade" in hopes of getting better throughput because the chipset or drivers are bananas. Guess I'll give mlo a while longer and see how things play out before I throw more money out of the window on an ax adapter. I don't even really need those speeds, but I really wanted to break the 1Gbit/s barrier, especially when I hear people are getting 4-5Gbit/s over wifi.

But overall? Huge let down for a 100€ adapter. Especially when I hear the Netgear equivalent is pumping a consistent 1.6Gbit/s. Sadly, it's not available in the eu and this was the "best" adapter I could source.

 

[Ripshod]

Asus' ROG site states quite clearly these have 2x2 (ax) and 2x2x2 (be) radios.

 

[jzchen]

@Jasperx You'll probably want to figure out a way to move the router and/or USB so close that they are maybe .3 m apart, then try speed testing. If you're getting abnormally slow speeds then further analysis is necessary...

 

[jzchen]

I just wanted to share that I had no issues getting wireless backhaul between a GT-AXE16000 and GT-AXE11000 through a floor/ceiling. Just remember a floor/ceiling does not equal a wall. They are weight bearing and a lot thicker!

You may need to get another BT8 for the other side of the floor just for backhaul to increase bandwidth...

 

[degrub]

If you are in a condo or other high rise apartment, the floor/ceiling may be 4-6 inch / 0.10 - 0.15 m reinforced concrete ! That will definitely attenuate a 6 GHz or 5 GHz signal ! Wood and gypsum board, not as bad since thinner.
As a footnote, i have seen listed specialized wifi extenders designed specifically for getting through concrete / concrete block walls/ceilings. i always thought drilling a hole would be more economical

 

[John Fitzgerald]

i always thought drilling a hole would be more economical

If you own the single family residence, yes. A hole is a fire breach that would not meet code and could be a hazard in the condo / high rise apt scenario. Different set of rules at play, just saying...

 

[jzchen]

I guess the only guaranteed safe solution: Out one wall, down, in one wall.

I had no idea why 2.5 Gbps backhaul was overtaken by 6 GHz backhaul, two screenshots on the first post...

Is 6E Faster than 2.5 G Backhaul?

My AXE16000 and AXE11000 have been in WiFi backhaul since last night. I tried the optimize function after which I lost both the AXE11000 and RP-AX56 nodes. I had to delete them and readd them. I had to hard reset the RP-AX56 before it could be found and added back. Gonna try to add two...

rog-forum.asus.com

EDIT- That's through 2003/2004 California building code floor, for what it's worth. (House was started in 2003, finished in 2004)...