Trying to understand AiMesh speeds (AX86U Pro with AX3000 node)

[SR-71]

Wondering if someone can help explain this to me. No problem to report as I'm pleasantly surprised by the results. I'm just not understanding the speed behavior (speeds much faster than expected):

Setup:

• AX86U Pro main router in basement with AX3000 (v1) as AiMesh node in my home office on first floor. All on current stock firmware.
• Office is not wired for ethernet or coax, so AX3000 is using WiFi backhaul out of necessity.
• Have laptop connected via ethernet to the AX3000 LAN port 1 in home office.
• Have Comcast 1GB service (1 gig down, 40 mbps up)
• Coax comes in via the basement > connected to Arris SB8200 modem > AX86U Pro's 1-gig WAN port) as you'd expect (Comcast lists the SB8200 as approved up to 957 mbps, close enough to what I need for gig service).
• My office is somewhat close over where the AX86U sits in the basement (1 floor and 1 wall apart, one small room over, maybe 20-25 ft diagonally between the two routers).

Speed Results via speedtest.net to a nearby host (same city) when other household devices aren't actively streaming (to reduce variability in the results):

• When laptop is connected via ethernet to the AX3000 node: repeated speed tests show my laptop is getting ~918 down and ~42 up -- so virtually full speeds for the ISP plan given the above setup
• When laptop is connected to the AX86U Pro via WiFi, getting ~770 down and ~42 up.
• When it's connected to the AX3000 via WiFi, getting ~600-700 down and ~42 up.
• AX86U's internal direct speed test is ~952 down, 42 up with QoS disabled, which shows the max coming in -- very close to the 957 rating by Comcast (it hit 961 one time).

Here's where I'm confused:
It's my understanding that when wireless backhaul is used, the speeds should be roughly half (similar to repeaters, due to the relaying of packets). However, all of the speeds above are more than half, and when using ethernet into the AX3000 node, I'm getting virtually full speeds.

Again, I'm not complaining at all, but just not understanding why. All I can think of is that when the laptop is wired into the AX3000, it's cutting out part of the WiFi transmissions (to/from the laptop to the AX3000 node). That said, the AX3000 still has to relay the packets to the AX86U though, so shouldn't I be seeing some drop off in speed even over ethernet into the AX3000 due to the meshing backhaul over WiFi?

Thanks in advance, just trying to educate myself on how this works since it's my first ASUS AiMesh setup.

P.S. Several weeks back I tried substituting a second AX86U Pro (I posted here when it was on sale) for the AX3000 AiMesh node in my office. Guess what -- I was definitely getting slower speeds by at least 100-200 mbps. This was regardless of how the laptop connected to it and after orienting the new 86U and its antennas in a variety of positions. So I sent it back and put the AX3000 back into service.

Go figure, I thought the second 86U Pro with its 4x4 5Ghz spec and faster CPU would be superior for this use over the AX3000's 2x2, but it wasn't. Maybe the 4th internal antenna wasn't as effective as the AX3000's all external ones? I have the AX3000's two outer antennas oriented straight up for the first floor coverage, and the two innermost at 45 degrees for angling toward the basement 86U for meshing, which is working well. The original 86U is placed high up in the basement, with its antennas angled halfway up (45 degrees) to get the most out of that location. I also tested them being flat and fully vertical, and got the best performance at 45 degrees.

 

[OzarkEdge]

Inspect the wireless backhaul connection details (NSS=4 streams/antennas between router and node) in the Wireless Log to see how the shared backhaul is connecting. In my experience, a 160MHz bandwidth backhaul using DFS (between AX node and router) has less effective range than a 80MHz bandwidth backhaul... lower speed/thoughput at same distance.

OE

 

[SR-71]

Inspect the wireless backhaul connection details (NSS=4 streams/antennas between router and node) in the Wireless Log to see how the shared backhaul is connecting. In my experience, a 160MHz bandwidth backhaul using DFS (between AX node and router) has less effective range than a 80MHz bandwidth backhaul... lower speed/thoughput at same distance.

OE

Thanks, I searched the log for "NSS", but it only had a couple of entries that stated:

avahi-daemon[3788]: WARNING: No NSS support for mDNS detected, consider installing nss-mdns!

No other NSS entries at all. So I'm not sure what else to look for in the log.

I appreciated your comment about 160 vs. 80MHz trade-offs. Yes, as frequency increases, throughput/speed increases but range/stability decreases, particularly at the outer areas.

For this setup, though, I've found that 160MHz gives me better speed performance, especially given the relatively short distance between the two routers. Whenever I've tried 80, I get noticeably slower results. So far, I don't need better range and would rather have the better speed performance at 160. If I had a larger area to cover, I'd probably try 80 again.

Again, I'm really just trying to understand why I'm getting virtually full ISP speeds via ethernet into the mesh node router.

Everything I've read about meshing over WiFi backhaul indicates that I should see roughly half the speed when connecting to a mesh node.

Same question about getting between 63% - 73% of my ISP plan speeds on my laptop when connecting via WiFi to the meshed AX3000 node. That's definitely well above half of the 950+ mbps this cable modem is rated for. It's great that I'm getting that performance, but I don't understand why.

If anyone can shed light on this, I'd appreciate hearing it.

 

[Reflexgti]

Perhaps your signal quality between your router and mesh node is fast enough where you're only seeing a small speed drop when connecting to the AX3000 via WIFI.

 

[bbunge]

Thanks, I searched the log for "NSS", but it only had a couple of entries that stated:


No other NSS entries at all. So I'm not sure what else to look for in the log.

I appreciated your comment about 160 vs. 80MHz trade-offs. Yes, as frequency increases, throughput/speed increases but range/stability decreases, particularly at the outer areas.

For this setup, though, I've found that 160MHz gives me better speed performance, especially given the relatively short distance between the two routers. Whenever I've tried 80, I get noticeably slower results. So far, I don't need better range and would rather have the better speed performance at 160. If I had a larger area to cover, I'd probably try 80 again.

Again, I'm really just trying to understand why I'm getting virtually full ISP speeds via ethernet into the mesh node router.

Everything I've read about meshing over WiFi backhaul indicates that I should see roughly half the speed when connecting to a mesh node.

Same question about getting between 63% - 73% of my ISP plan speeds on my laptop when connecting via WiFi to the meshed AX3000 node. That's definitely well above half of the 950+ mbps this cable modem is rated for. It's great that I'm getting that performance, but I don't understand why.

If anyone can shed light on this, I'd appreciate hearing it.

WIFI backhaul, in your case, is likely limited by the AX3000. While the AX86U Pro has 4X4 radios the AX3000 has 2X2 radios. So the reduced bandwidth is understandable as the AX3000 AiMesh node can not take advantage of the full capacity of the AX86U Pro. A better choice for an AiMesh node would be something with more 5 GHz radios like an AX86S or AX86U.
AiMesh is not designed for speed or bandwidth but improved coverage. If you do not need WIFI in your office make the AX3000 into a Media Bridge so all of its WIFI capacity is used between the AX86U Pro and the AX3000.
Better yet, run an Ethernet cable up to the office. Pre made cables are available in long lengths and not expensive.

 

[rung]

While the AX86U Pro has 4X4 radios the AX3000 has 2X2 radios.

I believe the v1 RT-AX3000 has 4x4 radios. I know my v1 RT-AX58U does.

 

[Tech9]

It's my understanding that when wireless backhaul is used, the speeds should be roughly half (similar to repeaters, due to the relaying of packets). However, all of the speeds above are more than half, and when using ethernet into the AX3000 node, I'm getting virtually full speeds.

Repeater (wireless node) first hop - the speed may vary depending on multiple factors, but you'll see 50% mentioned most of the time for shortness. It's around this number in real life scenarios. Your wired to the node clients have full bandwidth, no retransmissions there. The node is acting like a high-power external wireless adapter for the wired devices. This is not a repeater scenario.

I have 2x wirelessly uplinked UniFi Express units and they also get full Gigabit speeds on the LAN ports when connected upstream with PHY rate 1200Mbps (2x2, 80MHz). They are similar to your node, just don't serve any wireless clients in my use case. If they do - the throughput drops to about 400Mbps. If I want to bump it up further I have to switch to 160MHz uplink and PHY rate 2400Mbps.

 

[SR-71]

Thanks everyone, that helps. To respond to the comments, in order:

bbunge, I had the same thought:

About a month ago, I tried a second brand new AX86U Pro as the mesh node in my office, thinking they'd make a perfect pair. Unfortunately, it was noticeably slower with my laptop connected via ethernet. With its superior specs compared to the AX3000 (faster and additional core CPU, 4x4 5GHz instead of 2x2), it should have at least been as fast or faster than the AX3000. But it wasn't.

I tried all kinds of positioning with the new router and its 3 external antennas, but nothing helped more than a little, so I sent it back. I'm still scratching my head on that one. I had also wanted the second 86U Pro as a backup router in case the primary ever failed, but in the end it wasn't worth it to me. If I were to add another device today, I'm thinking the GT-AX6000 would be a better AX replacement as the prime with the 86U Pro as the mesh node. This video review's side-by-side testing results, while just one review, is pretty eye-opening:

However, with the AX3000 still giving me such good performance after 5 years and still receiving firmware updates, I can't justify spending the money right now -- I'd rather save it for the next tech jump upgrade (eventually "full" WiFi 7 spec with the additional band and capabilities) when my end client devices are able to take advantage of it. I also really like that both the AX86U Pro and AX3000 are nice compact devices that fit into my tighter spots. I don't see the dual-band BE86U replacing either at current pricing. Since it's missing the 6GHz band, the value prop just isn't there for me. Also, with mainly AX and AC devices, I wouldn't likely see much, if any, improvement for the higher price tag, and possibly worse given my recent comparison at the mesh node.

So I'm waiting to see where ASUS goes to fill the gap eventually. The BE92U briefly caught my eye as being a near size format clone of the 86U, but with the third band and 4 external antennas. However, I quickly lost interest when I saw the following specs per band and some reviews/reports about overheating issues:
2.4GHz: 3x3
5GHz: 2x2
6GHz: 2x2

=====

rung, sorry to burst your bubble, but:

ASUS has both the AX58U and AX3000 specced at 2x2 for both bands (verified at https://www.asus.com/us/networking-iot-servers/wifi-routers/asus-wifi-routers/rt-ax58u/techspec/, same as the AX3000 page). The v1 58U is the exact same hardware as my v1 AX3000 (yours is just branded specifically for Best Buy). Still, I appreciated the reply.

=====

Lastly, Tech9, I think you hit the nail right on the head with what I've been wondering about and trying to confirm here:

Your wired to the node clients have full bandwidth, no retransmissions there. The node is acting like a high-power external wireless adapter for the wired devices. This is not a repeater scenario.

That makes perfect sense to me, given the speed numbers I'm seeing. It confirms what I wondered could possibly be the case in my first post above, but not stated nearly as elegantly as you put it:

All I can think of is that when the laptop is wired into the AX3000, it's cutting out part of the WiFi transmissions (to/from the laptop to the AX3000 node).

I appreciated your insightful description/analogy of it being a high-powered external wireless adapter for the laptop (i.e., not a repeater)! That's exactly what helped me understand the difference in performance.

So now I know, if I want to get the best peformance out of the mesh node, wiring into the LAN ports will do it as those devices aren't in a repeater situation. The AX3000 is basically substituted as the wireless adapter for those devices, avoiding a hop. For the other devices using the AX3000 as a repeater, I'm still pleasantly surprised by the higher (more than 50%) speeds. Although, I still don't know why they're that good over the wireless backhaul and why the second AX86U Pro performed noticeably worse (other than the difference in the antenna configuration design).

Thanks everyone, I really appreciate this community helping me learn more about it.

 

[rung]

rung, sorry to burst your bubble, but:

ASUS has both the AX58U and AX3000 specced at 2x2 for both bands (verified at https://www.asus.com/us/networking-iot-servers/wifi-routers/asus-wifi-routers/rt-ax58u/techspec/, same as the AX3000 page). The v1 58U is the exact same hardware as my v1 AX3000 (yours is just branded specifically for Best Buy). Still, I appreciated the reply.

Thanks. I read there was a debate whether the hardware was capable but Asus just decided to only implement 2x2 in software.

In my own defense, here is what the wireless info page says about its connection as a wireless AiMesh node to my AX86u pro:

 

[Tech9]

I read there was a debate whether the hardware was capable

Asus had multiple products in AX5400 and AX3000 class based on the same 4-stream capable hardware. AX3000 class devices were software limited to 2-streams. RT-AX58U V1 was one of them. RT-AX58U V2 though is 2-stream hardware.

That makes perfect sense to me

Realistic expectations for different scenarios with common AX 2-stream client 80MHz channel bandwidth, somewhat clear Wi-Fi environment, dual-band routers with Gigabit ports and 2-stream radios (PHY rate 1200Mbps), including wired clients:

Wi-Fi > ~800Mbps
Wi-Fi - Node/Repeater - Wi-Fi > ~400Mbps
Wi-Fi - Node/Repeater - Wired > ~Gigabit
Wi-Fi - Media Bridge > ~Gigabit

Since the wireless uplink (backhaul) bandwidth is shared between all the clients wired and wireless the above is valid if one single client is using the entire bandwidth. The speed ~700Mbps from wireless node to wireless client is possible only if the node is connected at 160MHz wireless uplink (PHY rate 2400Mbps) and the client supports 160MHz wide channel. Maximum PHY rates and throughput - signal level -54dBm or better, noise -90dBm or better (usually not the case in real life applications).

In general - don't waste your time speed testing Wi-Fi. Best effort shared technology, not guaranteed, may change at any moment.

 

[SR-71]

Asus had multiple products in AX5400 and AX3000 class based on the same 4-stream capable hardware. AX3000 class devices were software limited to 2-streams. RT-AX58U V1 was one of them. RT-AX58U V2 though is 2-stream hardware.



Realistic expectations for different scenarios with common AX 2-stream client 80MHz channel bandwidth, somewhat clear Wi-Fi environment, dual-band routers with Gigabit ports and 2-stream radios (PHY rate 1200Mbps), including wired clients:

Wi-Fi > ~800Mbps
Wi-Fi - Node/Repeater - Wi-Fi > ~400Mbps
Wi-Fi - Node/Repeater - Wired > ~Gigabit
Wi-Fi - Media Bridge > ~Gigabit

Since the wireless uplink (backhaul) bandwidth is shared between all the clients wired and wireless the above is valid if one single client is using the entire bandwidth. The speed ~700Mbps from wireless node to wireless client is possible only if the node is connected at 160MHz wireless uplink (PHY rate 2400Mbps) and the client supports 160MHz wide channel. Maximum PHY rates and throughput - signal level -54dBm or better, noise -90dBm or better (usually not the case in real life applications).

In general - don't waste your time speed testing Wi-Fi. Best effort shared technology, not guaranteed, may change at any moment.

Thanks for laying out the connection scenarios above. The numbers I'm seeing are making sense now that I have that context.

Fortunately, I'm seeing better performance on the Wi-Fi connections, somewhat better than your numbers above as I shared earlier. I think that's due to the following positive factors in my current location (my last one was very heavily saturated from neighbors):

• Both routers are fairly close to each other (under 30 ft)
• Very little signal pollution from neighbors on both bands (the few showing up are very weak)
• Able to use 160MHz for full bandwidth across devices
• PHY rate for 5GHz on the AX3000 node (as reported in the AX86U mesh panel, it varies a little): TX 2000 - 2250 Mbps; RX 2400 - 2880 Mbps (not sure why RX is faster)
• Small number of devices actively streaming/browsing concurrently. More are connected, but their traffic is minimal and some devices are in sleep mode or off until used.
• Thus my laptop gets the lion's share of the bandwidth.

Thanks again, this has been great to understand.

 

[SR-71]

P.S. I didn't know the AX3000 hardware was 4-stream capable but software-limited to 2 streams, that sucks ASUS. I always thought it was just hardware-limited to 2x2 per ASUS's published specs, being a lower mid-range model (also one of their first AX routers) when it was released in 1H 2020). Thanks rung and Tech9 for sharing that.

Being somewhat jaded regarding vendors, it makes me think ASUS crippled it so they could artificially "distinguish" and charge more for the other "higher-end" routers -- unless there was a valid engineering reason behind it (e.g., heat or processor limitation?). After all, if we could have gotten 4x4 performance out of it at the lower price point, why buy the more expensive models unless one needed more range or another "higher end" feature?

Don't get me wrong, it's been a solid performer as my primary router for 4.5 years and got a second life as a mesh node last year. After 5 years of good mid-level performance + still getting fairly regular firmware updates, I seriously can't complain too much. It even outperformed a second AX86U Pro I tried recently as a mesh node in the exact same location in my office. I know that shouldn't be the case, but it was, go figure.

 

[Tech9]

I didn't know the AX3000 hardware was 4-stream capable

RT-AX58U (3000) V1 was actually 2x4:4 configuration on 5GHz radio. This means 2xTx with 4xRx streams and explains why you see >2400Mbps PHY rate @160MHz wide channel. Wide channel also explains your higher Repeater performance, but the client has to support 160MHz wide channel too. So nothing strange, this is how it works.

 

[SR-71]

RT-AX58U (3000) V1 was actually 2x4:4 configuration on 5GHz radio. This means 2xTx with 4xRx streams and explains why you see >2400Mbps PHY rate @160MHz wide channel. Wide channel also explains your higher Repeater performance, but the client has to support 160MHz wide channel too. So nothing strange, this is how it works.

Thanks for the explanation, now those PHY rate numbers make perfect sense.

Yes, my laptop, phones and a tablet all support the AX 160MHz wide channel. A mix of streaming devices means a "no" for some, especially the older ones. But there's plenty of bandwidth to accommodate them, even with their lower Wi-Fi capabilities. I also think the very low interference from neighbors and relatively close proximity within the home helps too, for a relatively clean, strong signal.

This has been a great discussion, I learned a lot. I started off just trying to make sense of why some connections were giving me virtually full ISP speeds through the mesh node while others were producing more than the oft-cited 50% Wi-Fi backhaul/repeater speeds. Understanding this greatly helps me make best use of my hardware and setup. Much appreciated!