Scheduled maintenance: Community will be offline Monday June 17th, 1:00 AM - 6:00 AM (PT)
New Member
Posts: 8
Registered: ‎04-27-2019
Kudos: 4

Nanobeam 5AC Gen2 poor throughput with high CPU load

[ Edited ]

First time Ubiquiti buyer and first time setting up a PTP link. I'm setting up a link between 2 buildings, 150ft apart. CINR is 34-35. When transfering files, best I can get is 200Mbs +/- 20Mbs. nanobeam internal speed test is 300-340mbs half duplex/320 combined full duplex 

 

I've tried nanobeam internal speed test, jPerf, file download from windows server, internet speedtest. Same performance and in all cases CPU on RX unit is 100% with 90%+/- 5% sirq. ustatsd load has been as high as 30-40% on some firmwares. TX CPU load is also high, but not quite capped. I've tried ptp 80mhz (full 655mbs link) and ptmp 40mhz (flex new mode). I have even put both units in a basement on poles 25ft apart with nothing in fresnel zone on an isolated network.

 

I've also tried various firmwares, including the beta and as well as diabling the web ui server. while beta+ no webui was a little better, it was not a significant difference.

 

 

Has anyone tried using Openwrt on these units for increased performance?

nanobeamsignal.png
SuperUser
Posts: 42,839
Registered: ‎06-23-2009
Kudos: 8761
Solutions: 1491

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

It's difficult to guess, based on so little information; the

desktop screenshot shows no settings or even summaries.

 

But this seems to be a nearly perfect link--unless there is

some issue with the path. For example, if the radios are

at full power to achieve a -47dBm signal, there could be

some odd attenuation. But there's no information on that.

 

If you're attempting to transfer large block files on a radio

link, it will exhibit relatively slow rates compared to the

theoretical values. Is this the case here?   Dave

 


> HQ in Seacoast region New Hampshire U.S.A.
> Ubiquiti Certified Trainer [UCT] for:
     UBWA [AirMax] / UEWA [UniFi] / UBRSS [routers]
UBNT.NH@gmail.com
New Member
Posts: 8
Registered: ‎04-27-2019
Kudos: 4

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

[ Edited ]

Power on both units currently  is at minimum -4dbm. Current firmware is lastest beta. I've tested several permutations of of every setting I thought made sense. I've played with ampdu, wireless mode, channel width,  output power, TDD modes, killing the web UI, current release and beta firmwares.  

 

For speed test I've downloaded files via http, ran nanobeam internal speed test, and used jPerf. Best case nanobeam/jPerf cap around 300-340 1/2 duplex on isolated link (2 laptops connected to each end). http dl caps around 200mbs. 

nano4.PNG
nano3.PNG
nano2.PNG
nano1.PNG
New Member
Posts: 8
Registered: ‎04-27-2019
Kudos: 4

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

here is client config

nano24.PNG
nano23.PNG
nano22.PNG
nano21.PNG
SuperUser
Posts: 42,839
Registered: ‎06-23-2009
Kudos: 8761
Solutions: 1491

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

You could hardly ask for a more simple configuration, and

I don't see any issues with it. A few possibly minor points:

 

> select a channel in the 5700MHz region [don't use 'auto']

> increase rf output so received signal is -40dBm

> avoid beta when testing; use the latest release

> avoid a path with lots of metal and radio sites with metal

   behind them, to reduce rf refractions

> don't tinker with default settings unless there is a

   special need; they should do as well as any here

 

Dave


> HQ in Seacoast region New Hampshire U.S.A.
> Ubiquiti Certified Trainer [UCT] for:
     UBWA [AirMax] / UEWA [UniFi] / UBRSS [routers]
UBNT.NH@gmail.com
New Member
Posts: 8
Registered: ‎04-27-2019
Kudos: 4

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

[ Edited ]

-5700 had least inference, but it was on auto, made no difference
-i've had it cranked upto -34dbm, made little difference, and alot of people consider anything below 50dbm too hot
-as posted, tried several firmwares, made little difference
-path has no metal anywhere near FS
- i disagree completely with not tinkering, that is how people learn and find Firmware/software Bugs as long as it is done in a controlled way using scientific method. Either way none of my tinkering made much a difference vs stock settings. CPU is appears to be biggest issue.

Point of asking this question was to find if any settings were particularly buggy. or if there was some undocumented hack to reduce cpu load, or a particular firmware version that worked best(newest isn't always best for all cases, depends why/how it was patched) etc beyond the obvious beginner troubleshooting like dbm/alignment etc. Also to find out how normal this much cpu load was versus possible defective chip/heatsink.

SuperUser
Posts: 42,839
Registered: ‎06-23-2009
Kudos: 8761
Solutions: 1491

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

[ Edited ]

I know of no tricks to reduce CPU loading, and I would

not expect to find 100%--so your concerns are sensible.

But I can't correlate it with anything beyond what I advised,

and definitely not to a 'heatsink'

 

Use auto 'distance', but not auto channel.

 

I don't know anything about your background, but in

general, tinkering is a bad idea that teaches nothing.

For many users and for many settings, there is no such

thing as 'scientific' tinkering. Settings are often not fully

documented, and even less often understood by users.

Multiple unrelated setting changes can combine into what

we call scientifically a 'mess'. And when people post only

dashboards and not other settings, it can be very confusing.  Dave

 

 

 


> HQ in Seacoast region New Hampshire U.S.A.
> Ubiquiti Certified Trainer [UCT] for:
     UBWA [AirMax] / UEWA [UniFi] / UBRSS [routers]
UBNT.NH@gmail.com
New Member
Posts: 8
Registered: ‎04-27-2019
Kudos: 4

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

I hear what you what you say and still disagree, but it isn't worth the time to explain or helpful. 

 

What would be helpful is hearing what your experience is with max speed on high quality links using these nanobeams and what the cpu load is using shh->top on known good units.

 

SuperUser
Posts: 42,839
Registered: ‎06-23-2009
Kudos: 8761
Solutions: 1491

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

Believe me, I believe in tinkering / hacking as a wonderful

method of exploration in all kinds of things--including

radio. But these aren't toys--and a lot of people on these

forums treat them as if they were.

 

I have used NBE-5AC-Gen2 all over the place and I

haven't seen this issue. I haven't seen all use cases,

and you may even have a bad radio--although I doubt it.  Dave


> HQ in Seacoast region New Hampshire U.S.A.
> Ubiquiti Certified Trainer [UCT] for:
     UBWA [AirMax] / UEWA [UniFi] / UBRSS [routers]
UBNT.NH@gmail.com
New Member
Posts: 8
Registered: ‎04-27-2019
Kudos: 4

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

If you are here to be helpful, I truly appreciate, but please keep the patronizing to yourself, it is not welcomed. 

SuperUser
Posts: 42,839
Registered: ‎06-23-2009
Kudos: 8761
Solutions: 1491

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

[ Edited ]

Tell me where the patronizing is. I'm not even referring

to you personally--I don't know your background. I do

believe in tinkering as a life-style--I've done a lot of it.

I have also seen multiple cases where people have

treated these radios as their toys and have put them

in violation of both law and good practice.

 

Please don't find layers that don't exist.   Dave

 


> HQ in Seacoast region New Hampshire U.S.A.
> Ubiquiti Certified Trainer [UCT] for:
     UBWA [AirMax] / UEWA [UniFi] / UBRSS [routers]
UBNT.NH@gmail.com
New Member
Posts: 8
Registered: ‎04-27-2019
Kudos: 4

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

I'm here to look for a solution or reason to RMA/return these not to argue with you. If you don't have something constructive to say, kindly find another thread. 

New Member
Posts: 26
Registered: ‎09-04-2018
Kudos: 12

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

@cookiemonsterI have exactly the same problem, with a pair of the NanoBeam 5AC Gen2 performing as a link between 2 buildings, and have had no luck from Ubiquiti after a few attempts of theirs at helping. It has been an ongoing issue and is still unresolved, see here

https://community.ubnt.com/t5/airMAX-Installation/High-CPU-usage-of-sirq-on-NanoBeam-5AC-Gen2-PtP-li...

 

 

Then there is this post

https://community.ubnt.com/t5/airMAX-AC/Getting-advertised-450Mbps-throughput-on-Nanostation-5AC-loc...

about the NanoStation 5AC, which runs exactly the same firmware, based on the md5 checksums listed on

https://www.ui.com/download/airmax-ac/nanobeam-ac-gen2/nbe-5ac-gen2/airos8-wa-board-firmware-v8512

and

https://www.ui.com/download/airmax-ac/nanostation-ac/loco5ac/airos8-wa-board-firmware-v8512

And @UBNT-JoshMcGee has stated on that post that there is an open investigation for the CPU usage issue, I believe it applies to the NanoBeam 5AC Gen2 as well. Josh wasn't clear on which particular models they were looking at.

 

Senior Member
Posts: 3,163
Registered: ‎01-29-2015
Kudos: 523
Solutions: 131

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

From my experience - I've had mine up for going on 2 years now - you won't get better than 260-300 or so mbps, even with a very clean 80 MHz width, so really there's no reason to use 80, since 40 will handle that and the signal is cleaner. I have a 5AC 19 and a 5AC Gen2, and the 19 seems to do a little better on CPU than the 5AC Gen2, but it isn't by much. They simply run out of power at that 300 mbps mark, and don't have the "umph" to move any more. 

 

I think on the older 5ACs that I used to have (gen1) they had the abilty to turn off encryption, but I'm not doing that. I'm betting it would help the CPU issue if you are brave enough to do this.

New Member
Posts: 26
Registered: ‎09-04-2018
Kudos: 12

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

[ Edited ]

@RobbieHexcept these devices are advertised as "450mbps+" devices. There should be some way, even on a lab bench or some special circumstances that allows it to acheive the advertised specs, otherwise they are using false advertising. I have yet to see any information from Ubiquiti on how to acheive their advertised claims.

 

Given that the bottleneck that limits the NB-5AC-Gen2 to about 200mbps is the CPU, then no amount of special radio environment is going to help it more than double that speed. You may be right about turning WPA2 off improving performance, but since that is not a selectable option, then that argument cannot be used for the NB-5AC-Gen2, and even if it was, that would be an extremely dirty tactic. The only ways forward are for Ubiquiti to

  • release new firmware to reduce the amount of CPU usage per mbps
  • release information on how to configure it to do what it is advertised to do
  • or admit that it cannot do what is advertised and start a product recall and free exchange for something that can perform as advertised.

Personally I would like to avoid the last option, unless Ubiquiti is going to also pay for technicians to drive out to our remote site and do the swap for us, as it is a 5 hour round trip drive from where I am. Plus I'm not sure if I have a ladder tall enough to reach that roof.

Ubiquiti Employee
Posts: 1,592
Registered: ‎09-24-2018
Kudos: 315
Solutions: 92

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

WPA2-AES is not handled in the CPU, so disabling encryption would provide no difference in the load. The issue which may be causing CPU overload is under investigation, as I said. It is not limited to particular devices. All radios which may be affected by this are being tested.

UBNT_Alternate_Logo.png
Ubiquiti Networks airMAX Support Team

Check out our ever-evolving Help Center for answers to many common questions!

FREE UBWA Student Guide-Great RF Primer!

Senior Member
Posts: 3,163
Registered: ‎01-29-2015
Kudos: 523
Solutions: 131

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

@UBNT-JoshMcGee Thank you, that works for me.

 

Someone mentioned switching to P2MP AirMAX AC and then change to Flex (NEW). I tried changing my AP to that, but the station would no longer connect. Is there something that has to be changed at the station end as well? Is this even a solution for the issue?

 

New Member
Posts: 26
Registered: ‎09-04-2018
Kudos: 12

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

@RobbieH 

Switching between PtP and PtMP AirMAX AC with the Flex (New) TDD framing has not helped reduce CPU usage on our NanoBeam 5AC Gen2 units.

 

If you do want to switch, at least on the NB-5AC-Gen2 units, the stations do need to match the access point setting. I don't believe there is an auto option in the stations so they will reconnect with the new method. I would imagine this is likely the same on many other models.

Senior Member
Posts: 3,163
Registered: ‎01-29-2015
Kudos: 523
Solutions: 131

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load


@raitdeptd wrote:

@RobbieH 

Switching between PtP and PtMP AirMAX AC with the Flex (New) TDD framing has not helped reduce CPU usage on our NanoBeam 5AC Gen2 units.

 

If you do want to switch, at least on the NB-5AC-Gen2 units, the stations do need to match the access point setting. I don't believe there is an auto option in the stations so they will reconnect with the new method. I would imagine this is likely the same on many other models.


I got mine all switched over, but have not had an opportunity to test yet.

 

The available throughput numbers went up some, now it's saying around 305 mbps, but it seems like the interference is higher.

Senior Member
Posts: 3,163
Registered: ‎01-29-2015
Kudos: 523
Solutions: 131

Re: Nanobeam 5AC Gen2 poor throughput with high CPU load

Yep, not fixed. And this is at less than 200 Mbps.

 

 

NB5AC CPU.png