ASUS RT-BE92U

[taurelio]

So i jumped to the RT-BE92U bandwagon and this is my experience: in terms of wifi coverage it was a litle bit desappointing when comparing with my old RT-AC66U B1, In the same place allmost equal stengh signal except on the 2.4 Ghz with increased coverage, jumped to Merlin beta and all running smooth, dns director is a must to have, this is the reason i use Merlin. On the first floor of my house i placed the RT-AC66U B1 as a Mesh unit, not ideal for some but it works for me, have about 300mbps speed, more than enough for the needs. It has been stable with Merlin and more free ram when compared to latest official stock firmware.

 

[Tech9]

i have a lot more of free ram

Happy to hear you found what works for you, but... what you do with this free RAM?

 

[taurelio]

what you do with this free RAM?

Me personnaly? Nothing, but was one of the notable out off the box differences between stock firmware and Merlins version, don't know the reason but still is about 150 mb less used versus stock. What matters is the system works and its stable

 

[Poul Bak]

. It has been stable with Merlin and more free ram when compared to latest official stock firmware.

I noticed the same, about 10% less than stock.

Happy to hear you found what works for you, but... what you do with this free RAM?

The thing is: What you measure is not PEAK value! When mine shows about 53% in idle mode, it means there's plenty of ram for peak performance.

 

[Tech9]

When mine shows about 53% in idle mode, it means there's plenty of ram for peak performance.

It doesn't really work this way on embedded devices. Whatever is needed and used in your specific configuration is loaded in RAM already regardless of how much data is passing through the router. More free RAM won't give you better performance.

 

[Bo17]

Spoke early, still getting crashes on latest Asus firmware (37563). Seeing some of these right before crashing:

kernel: WLC_SCB_DEAUTHORIZE error (-30)
kernel: WLC_SCB_DEAUTHENTICATE_FOR_REASON err -30

but searching through the forums those just seem to be normal Wifi disconnects.

Return window for the router is coming up so I guess I'll try an exchange in case it's a hardware fault, but I'm thinking it's most likely software. Just a shame that QA has gotten so bad recently, and there's nothing else with comparable features within the price range except TP-Link which has a rather lackluster WebUI. Dream Router 7 probably has more stable software but is 50% more expensive and has a bunch of other compromises that I'd have to work around. I feel like there's just so much tech debt in consumer devices at this point that getting something that works perfectly for every configuration is just luck at this point.

 

[Tech9]

Dream Router 7 probably has more stable software but is 50% more expensive

This is perhaps something location specific. In the US the regular price difference is $19. RT-BE92U is $250 (currently on sale for $220), UDR7 is $279 (almost all the time, Ubiquiti doesn't have many items on sale).

 

[Bo17]

This is perhaps something location specific. In the US the regular price difference is $19. RT-BE92U is $250 (currently on sale for $220), UDR7 is $279 (almost all the time).

It is certainly specific to what I bought. BE-92U is currently $220, and Best Buy recycle for an ancient router got me 15% off, so the UDR7 is almost exactly 50% more expensive right now. And again, there's other complications with the UDR7 since specs aside, I'd need even more things on the network that could also fail just to match the capabilities of an old AC86U running Merlin + Diversion + Skynet.

I have been tracking the UDR7 for a while now, and from their own support forums and others it seems to be quite buggy. Returning that if there's issues would be also be a bit more costly/difficult if it came down to it.

 

[Bo17]

Exchanged the router and slapped on a 4-year Best Buy warranty. Gonna test latest Merlin (3006.102.5) since it seems like Merlin is actually ahead of the latest official BE-92U firmware (37563) now.

- UPDATED: Merged GPL 3006.102_38757 for WIfi 7 devices.

Something I dug into was the Broadcom BCM6765 SoC was supposedly ARM Cortex-A7 (ARMv7, 32-bit) according to wikidevi, which doesn't make sense since the SoC should be ARMv8 and it definitely is running in aarch64 mode. No lscpu and meager info in /proc to verify it, but cat'ing /sys/devices/system/cpu/cpu1/uevent and /sys/firmware/devicetree/base/cpus/cpu@0/compatible both showed cortex-a53.
I also recall seeing

    Line 6963: Dec 31 19:00:37 crashlog: <6>CPU features: enabling workaround for ARM erratum 843419
    Line 6964: Dec 31 19:00:37 crashlog: <6>CPU features: enabling workaround for ARM erratum 845719

in the logs and those errata were definitely for the Cortex-A53 as well.
At least I'm no longer concerned with some extremely unlikely issue of software/CPU incompatibility potentially causing the crashes, or the unlikely case that Asus/Broadcom downgraded the CPU in a new device.

 

[jzchen]

@Bo17 It seems your reference may have been incorrect:

Dual 2x2 320MHz Wi-Fi 7, quad-core ARM SoC | BCM6765

BCM6765 is a dual SoC communications processor that enables low-power, small form factor multi-gigabit access points, routers, repeaters and service provider gateways for residential applications.

www.broadcom.com

Still waiting for mine to arrive from Walmart, (or more specifically D&H Distributing. How dare Walmart say ships and sold by Walmart then all of a sudden switcheroo me to a different/marketplace seller).

 

[Bo17]

@Bo17 It seems your reference may have been incorrect:

Dual 2x2 320MHz Wi-Fi 7, quad-core ARM SoC | BCM6765

BCM6765 is a dual SoC communications processor that enables low-power, small form factor multi-gigabit access points, routers, repeaters and service provider gateways for residential applications.

www.broadcom.com

Still waiting for mine to arrive from Walmart, (or more specifically D&H Distributing. How dare Walmart say ships and sold by Walmart then all of a sudden switcheroo me to a different/marketplace seller).

I mean yes, it obviously was incorrect since the Cortex-A7 isn't ARMv8

Broadcom/SoC

Broadcom

wikidevi.wi-cat.ru

(Ctrl+F "BCM6765")
but unless I'm blind, there's absolutely nothing on Broadcom's page to indicate which ARM core it actually has. "Quad-core ARMv8" could be one of a dozen processors ranging from Cortex-A3# to A7#.

 

[jzchen]

@Bo17 I may be too far out from studying EE, but did this line make any sense, or clarify how many bits the CPU is? I always thought it matches, so is this a quad core, 16-bit CPU? Shrugs...

"16-bit DDR4-3200 and LPDDR4/4x-4267 memory interface"

 

[Bo17]

@jzchen, No that's just the memory interfaces it supports, and 16-bit references the width of the memory data bus. The CPU may support multiple DDR4 chips to get 32-bit (2x16-bit) or 64-bit (4x16-bit) memory, but that's a whole other rabbit hole.

The only thing on Broadcom's BCM6765 page you linked that tells you anything about the CPU is "ARMv8 CPU". ARMv8 is the ISA (Instruction Set Architecture), and if there's no errors in Wikipedia's list

List of ARM processors - Wikipedia

en.wikipedia.org

then aside from the Cortex-A32 which is AArch32 (32-bit), all other ARM-designed ARMv8 cores are 64-bit (AArch64), many of which also support 32-bit (Aarch32) instructions like the A53 cores in the BE92U. Newer cores have generally done away with 32-bit and support 64-bit only.

I SSH'ed into the BE92U and dug around a bit for more info, which is how I verified it actually has Cortex-A53 processors and is running in AArch64 (64-bit) mode.

Most of this doesn't matter, as I was mostly trying to figure out if there was some glaring issue after seeing Wikidevi (incorrectly) list the BCM6765 as having Cortex-A7.

 

[RMerlin]

but unless I'm blind, there's absolutely nothing on Broadcom's page to indicate which ARM core it actually has. "Quad-core ARMv8" could be one of a dozen processors ranging from Cortex-A3# to A7#.

That`s because it`s not a generic ARM core, it's not a Cortex. it`s Broadcom's own design. Possibly B53 or a derivative, the official spec sheet does not specify it.

 

[Bo17]

That`s because it`s not a generic ARM core, it's not a Cortex. it`s Broadcom's own design. Possibly B53 or a derivative, the official spec sheet does not specify it.

I mean unless file contents for post #389 (/sys/ files and boot errata) were added by you, there's not much to go off of other than mentions of Cortex-A53 being scattered around the device files.

 

[RMerlin]

I mean unless file contents for post #389 (/sys/ files and boot errata) were added by you, there's not much to go off of other than mentions of Cortex-A53 being scattered around the device files.

We don't know how close to the Cortex-A53 Broadcom's Brahma design is. But according to cpuinfo, the BCM6765 should be the same CPU core as the BCM4916, only slower:

admin@RT-BE92U-B9B0:/sys# cat /proc/cpuinfo
processor    : 0
BogoMIPS    : 160.00
Features    : fp asimd evtstrm aes pmull sha1 sha2 crc32 cpuid
CPU implementer    : 0x42
CPU architecture: 8
CPU variant    : 0x0
CPU part    : 0x100
CPU revision    : 0

0x100 indicates Broadcom B53. CPU part for a Cortex-A53 would have been "0xd03"

For reference:

ARM CPU Vendor 及 Part ID 映射关系（持续更新）

作者：郑振宇 根据ARM CPU官方技术手册，ARM CPU的CPU型号、Vendor、版本等信息存于MIDR_EL1寄存器中:其中从低至高第0-3 bit表示revision，代表固件版本的小版本号，如r1p3中的p3；第4-15 bit表示part number(id)，代表这款CPU在所在vendor产品中定义的产品代码，如在HiSilicon产品中，part_id=0xd01代表Kunpe

kunpengcompute.github.io

As for the B53 sharing some erratas with the Cortex-A53, here's one of the two erratas you mention, which also applies to the B53:

arm64: apply ARM64_ERRATUM_845719 workaround for Brahma-B53 core (bfc97f9f) · Commits · linux-arm / linux-power · GitLab

The Broadcom Brahma-B53 core is susceptible to the issue described by ARM64_ERRATUM_845719 so this commit enables the workaround to be applied when executing on that core. Since there are now...

us-west-2b-production.gl-awslz.arm.com

 

[bigcid10]

Hello everyone
I'm having a issue with this be92u router
I'm getting these errors in the system log (see pic)
I upgraded from a ax88u which is now part of the mesh
I completely reset the system once already and the error seems to occurring
less in the log but it's still coming up though
I'm using the newest Merlin fw 306.102.5
also is there a reason when I log in using the ddns address
the router page seems soo slow to load versus using the default address log on way
thank you

 

[jzchen]

There appears to be a new firmware available this morning but I have not been able to install it (just woke up)...

 

[bigcid10]

There appears to be a new firmware available this morning but I have not been able to install it (just woke up)...

for the be92?
I don't see it

 

[jzchen]

for the be92?
I don't see it

Just like when I recevied it less than 2 weeks ago, I can't update it through the app nor the web GUI. I did not see a download anywhere either.