Radio engineers: Why doesn't the AX88U Pro's antenna array outperform the AX86U Pro's?

[aex.perez]

I've wondered about this issue for awhile but put it out of my mind as I couldn't find way to make them equal to keep my OCD in check, it's sort of of why I'm leaning to looking at using the model models for Router and Node when I replace what I have today. I'm not sure if the same can be said for the Pro versions of the AX88u/AX86u I currently have, but looking back at various methods to validate the config options I've used at least on the 2.4Ggz band on the non-Pro versions, all things being equal in the GUI I do see a difference in the devices, antennas/spatial streams. Note on 5Ghz band all the devices are the same.

But even then I see this difference when even when using Wifi Radar / Insight tool in the firmware to confirm and as far as I've seen in the GUI and NVRAM all settings are the same so I'm assuming the difference is in the chipsets or drivers when it comes to the 2.4Ghz band or something in AiMesh I'm not seeing, not having the Pro models to compare against don't know if this is the same for them as well..

All of the devices that can make use of the higher bandwidth and not challenge by range/distaance are either wired or on the 5Ghz band, so the lower bandwith on the 2.4Ghz bandwith / Physical Rate at the nodes for 2.4Ghz has never been a real issue, just a OCD nuissance as I try to keep all things equal (AES vs. AES-CCMP ?).

Thoughts?

 

[ColinTaylor]

Question about security protocol used by RT-ACU86U

These are the screenshots from the Wi-Fi survey page ( /visindex.asp ) for 2.4 GHz and 5 GHz, respectively. My router (shown in read in both cases) is RT-AC86U. I am just curious, what is meant that my router has "AES" (and not AES-CCMP)? As far as I understand, "AES" is just an encryption...

www.snbforums.com

 

[aex.perez]

Question about security protocol used by RT-ACU86U

These are the screenshots from the Wi-Fi survey page ( /visindex.asp ) for 2.4 GHz and 5 GHz, respectively. My router (shown in read in both cases) is RT-AC86U. I am just curious, what is meant that my router has "AES" (and not AES-CCMP)? As far as I understand, "AES" is just an encryption...

www.snbforums.com

OCD Satisfied!

That leaves differences between (3 vs 4) Antennas and Spatial Streams for the 2.4Ghz band between both models and if the same is seen on the Pro models

 

[ColinTaylor]

That leaves differences between (3 vs 4) Antennas and Spatial Streams for the 2.4Ghz band between both models

It's stated in the specs of each model.

and if the same is seen on the Pro models

RT-AX86U Pro:
2.4GHz 3x3
5GHz 4x4

RT-AX88U Pro:
2.4GHz 4x4
5GHz 4x4

 

[Tech9]

I wouldn't be surprised if the "RT-BE7200" Best Buy SKU is exclusively Taiwanese made. Isn't marketed under the RT-BE88U moniker.

This is probably just BestBuy branding, they had this marketing practice before. BestBuy exclusive like RT-AC2900 or RT-AC86U, then RT-AC1750 or RT-AC66U B1, etc. They also had one exclusive RT-AC1900P or RT-AC68U B2 variant. Just adding more to the Asus model line numbering confusion.

 

[aex.perez]

It's stated in the specs of each model.


RT-AX86U Pro:
2.4GHz 3x3
5GHz 4x4

RT-AX88U Pro:
2.4GHz 4x4
5GHz 4x4

I've seen that, I just never equated the physical rate with the streams / antennas / frequency, my mistake
Makes little real if any difference in real life as other factors play a role in overall device throughput over WiFi any way and the 2.4Ghz devices aren't complaing either.

And the external vs internal antenna debate for on the 5Ghz band on the AX88u and AX86u makes no difference there as well (160Mhz Channel Width)...
I suspect the same holds true for the Pro versions as far as this goes, and now my OCD is satisfied

 

[Tech9]

Makes little real if any difference in real life

Call Elon, book a ticket to Mars. No Wi-Fi pollution on Mars, 1148Mbps on 2.4GHz to Space X special edition 4-stream client may be possible.

 

[jasz]

This is probably just BestBuy branding, they had this marketing practice before. BestBuy exclusive like RT-AC2900 or RT-AC86U, then RT-AC1750 or RT-AC66U B1, etc. They also had one exclusive RT-AC1900P or RT-AC68U B2 variant. Just adding more to the Asus model line numbering confusion.

As I said prior, the previous AX58U was Best buy exclusive for US/NA market and "made/assembled" in Taiwan. Everyone else (Amazon, Newegg etc) got RT-AX3000 and shipped from Vietnam. It's the same router per FCC certification. European AX58U was Chinese.

I assume it's just a control/pricing mechanism or a way for ASUS to track issues.

They're obviously the same design, but definitely come from different locations. Prob to avoid Chinese tariffs in all honesty.

 

[aex.perez]

Call Elon, book a ticket to Mars. No Wi-Fi pollution on Mars, 1148Mbps on 2.4GHz to Space X special edition 4-stream client may be possible.

Ha!

Best I see is 500-700 Mbps on 5Ghz with 2 streams at 80mhz, on a iPhone. Not much more than that on the Laptop with a link speed of 2402/2402 also 2 streams but at 160Mhz

1Gb fiber being a limiting factor / wired client 1GbE seeing ~940Mbps in both directions

I wish I could get the 4804 Mbps or even the 1174 Mbps but not possible not even with WiFi 7 and a 10Gb link until I see it and even then…

 

[jasz]

Ha!

Best I see is 500-700 Mbps on 5Ghz with 2 streams at 80mhz, on a iPhone. Not much more than that on the Laptop with a link speed of 2402/2402 also 2 streams but at 160Mhz

1Gb fiber being a limiting factor / wired client 1GbE seeing ~940Mbps in both directions

I wish I could get the 4804 Mbps or even the 1174 Mbps but not possible not even with WiFi 7 and a 10Gb link until I see it and even then…

Sounds packet related.

My throughput drops on my fiber connection unless I run a ONT>OLT speed test to the software backend they use.. It's a weird issue because I can more or less manipulate how wireless performs through ONT side lol. Full 940 WL over 2x2, but it doesn't last long... 5 mins after "reset"? Can avoid losing throughput by running a 4x4 > 4x4 router to router connection.. which will peg full 940, even at lower QAM.

DOCSIS doesn't have these issues, at least with Broadcom D3.1 modem and Spectrum ISP in the US.

Edit: I assume that the PON chip in ONT just routes straight to the NPU portion of the wireless.. I mean 4x4 > 4x4 router in bridge mode ends up working. Can also get a connection to grab on a 2x2 160mhz Client @ 940, but that only works on my main router with QCA 1220 platform. If it was configured backwards to an older AX router, it won't hit the same throughput.

 

[snovvman]

I've had 2 AX86U (non pro) models here with completely different throughput over wireless when placed in same location.

Quality control, manufacturing, variations in component, and/or tolerance issues? If ASUS is delivering the same models that has vastly different performance from one another, I guess it's a crapshoot what the consumer gets out of the box?

Edit: OT for this thread but related to the post--I bought 3 AX86U Pros to see how they might work out. One, out of the box has a weird lag when loading the web UI. All have the same firmware, but one seems to work fine except for delays when loading the UI on login, each and every time.

 

[jasz]

Quality control, manufacturing, variations in component, and/or tolerance issues? If ASUS is delivering the same models that has vastly different performance from one another, I guess it's a crapshoot what the consumer gets out of the box?

Prob just a different FEM selection.. I can't tell you otherwise unless a different style AC adapter somehow impacts throughput.. (One was original ASUS branded, other ACBEL.. ASUS started giving unbranded 3rd parties around 2021-2022... Sold one to a friend so I can't exactly test it).

2x GT-AX6000 and the other AX86U were significantly worse. Pretty happy with my Qualcomm based BE800 (out performs everything I have).. just hate the GUI and lack of customization.

 

[aex.perez]

I don’t think this consumer bade router can do much more anyway, getting 940Mbps of my 1Gb but had to disable anything related to the BWDPI engine, QoS for example and Trend Micro Protections, don’t think there’s enough CPU, remember consumer based

 

[jasz]

I don’t think this consumer bade router can do much more anyway, getting 940Mbps of my 1Gb but had to disable anything related to the BWDPI engine, QoS for example and Trend Micro Protections, don’t think there’s enough CPU, remember consumer based

Have a Qualcomm router with 2.2ghz A73.. its stronger, but has the same general issues for WL purposes.. 1.7ghz NPU. My Asus routers are worse with Fiber.

 

[sfx2000]

Have a Qualcomm router with 2.2ghz A73.. its stronger, but has the same general issues for WL purposes.. 1.7ghz NPU. My Asus routers are worse with Fiber.

I just have to ask, where do you get this "NPU" nomenclature?

IF this refers to the ubi32 NSS offload, it's not an NPU - I would consider Runner from Broadcom to be in the same category.

Nobody in HW/SW development refers to these subsystems as NPU's - seriously, we don't.

It's important to note that NPU does have relevance to the Machine Learning community, as these are the IP blocks focused on device centric AI capabilities...

 

[sfx2000]

Prob just a different FEM selection..

Irrelevant - the FEM's are matched to the radios/basebands, so it really doesn't matter which vendor is used there - these are standard parts, so it's really down to a cost basis on the front-end itself.

As I mentioned earlier in this thread - RF impacts the entire design of the device - from placement of the antennas, the shielding, the RF feedlines to the antennas, even the circuit board itself, as it is the ground plane for the radiated solution.

Other things that can impact RF - the DC/DC converters as the buck things over to AC and back to DC at the desired voltage. It's the AC adapter from the wall, as there are many that are nominally DC, but have a lot of switching noise and ripple. It even goes down to the length and composition of the wires from the wall-wart to the device - it's all part of an engineered solution that has to pass the regulatory requirements for the country the device is sold in.

Please be careful on comments - unless one does this as a day job or a multi-year career, one isn't really qualified to make statements on RF solutions.

 

[jasz]

I just have to ask, where do you get this "NPU" nomenclature?

IF this refers to the ubi32 NSS offload, it's not an NPU - I would consider Runner from Broadcom to be in the same category.

Nobody in HW/SW development refers to these subsystems as NPU's - seriously, we don't.

It's important to note that NPU does have relevance to the Machine Learning community, as these are the IP blocks focused on device centric AI capabilities...

Wherever the network processing is offloaded for wireless. 4x4 BCM radios have their own internal processing on radio and wont even touch the main B53. Not exactly sure how it works on QCA.

Can you explain why a fiber ONT can process higher throughput when internal stats are cleared or an ONT to OLT test is engaged? Impacts wireless throughput, but isn't a wireless issue. Stronger QCA design just functions better, but shares issue.

Shed some light?

 

[jasz]

Please be careful on comments - unless one does this as a day job or a multi-year career, one isn't really qualified to make statements on RF solutions.

I really don't care if you wanna ego me. lol

Why would it matter.. Just making the argument that certain samples do perform better (same channel, same bonding. same firmware.. same everything settings wise). If you don't believe me, thats fine

 

[sfx2000]

Where ever the network processing is offloaded for wireless. 4x4 BCM radios have their own internal processing on radio and wont even touch the main B53. Not exactly sure how it works on QCA.

There is no network offloads on the wireless interfaces - there is crypto offload, and mediatek, qualcomm, and broadcom all do this at the baseband level in the NIC's themselves.

The flow accelerators/offloads, that is on the switch element inside the SoC, and it can be software, hardware, or a combination of both depending on the SoC implementation.

Also on a HW basis - how the wireless NIC's interface to the SoC is important - it can be GMII/RGMII (ethernet), PCI, and more recently PCI-e, again dependent on the vendors - even with in vendor solution, highly integrated chipset will use internal buses to move data around inside the SoC itself (consider IPQ40xx chips for example)

 

[sfx2000]

Where ever the network processing is offloaded for wireless. 4x4 BCM radios have their own internal processing on radio and wont even touch the main B53. Not exactly sure how it works on QCA.

Hate to break it to you - on all modern SoC's - the ethernet ports and wireless interfaces don't touch the CPU - it's all in the switch subsystem.

CPU comes into play with the bridge interface, along with the control plane (let's configure the switch) otherwise it's an application processor that does some of the upper layer processing...

Sorry if you think I'm calling you out - if you believe so, that's on you...

I'm just here to correct the record - I've been doing this stuff for over 30 years now...