Page 1 of 1

RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Fri Jun 28, 2013 1:57 am
by roadrunner79
I am using RB911G-5HPnD or RB912UAG-5HPnD boards as CPE in a PtP link in benchmark setup ("on the table" with RF cables and attenuators ~80dB on each RF chain). For AP i have RB411GL+R52Hn.
UL and DL RSSI is about -60dBm on each chain.
In this scenario I'm unable to achieve 100% CCQ value on DL (RB911/912 card receiving) no matter the channel bandwidth. I have 100% CCQ only on UL (RB911/912 transmitting).
There is a slight improvment when using nstreme, rather than nv2.

Any known RF problem in there, or compatibility issues between new atheros chipset of RB911/912 and AR9220?

The above problem is not observed when using RB711G-5HnD boards as CPE (even at -70dBm RSSI i have 100%CCQ).

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Fri Jun 28, 2013 2:52 am
by warwick09
Hmm, did you observe CCQ readings while passing traffic? I.e. when conducting a bandwidth test ... Usually at idle, CCQ readings can be deceptive.

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Fri Jun 28, 2013 8:03 am
by roadrunner79
I have traffic on link, while reading CCQ.
Also max MCS is limited according with CCQ, meaning 78/104mbps in 20MHz.
Maybe there is a problem with QAM64 MCS's.
I have latest firmware 6.1.

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Fri Jun 28, 2013 10:54 am
by roadrunner79
I have used RB912G as AP and RB911G as CPE and not even like this the CCQ is not 100% as it should be. The variation is quite big (from 64% to 98%).

If in the same setup i'm using AR9220 for both AP and CPE (any of the following: R52Hn, R52n-M, RB711G-5HnD, RB711GA-5HnD, SR71-15, etc) the performance is maximum with 100% CCQ (variation of max 2%).

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Wed Jul 03, 2013 7:57 pm
by roadrunner79
Nobody has this problem?

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Mon Nov 18, 2013 10:16 am
by petervogt
We also got this problem. lab-setup with -50 dBm but can't get the ccq to 100% with 2x 911G. Throughput isn't that bad, around 130 MBit TCP (x86 with iperf).

Any ideas? So far we only tested with RouterOS 6.5 and 6.6. Should we downgrade to 5.x?

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Thu Nov 28, 2013 8:42 pm
by roadrunner79
I have observed this behavior in lots of setups (lab, indoor, outdoor), different combinations (802.11n, nv2, nstreme), on all units with new Atheros chipset (912, 911, SXT Lite 5).

I haven't found a solutions.

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Fri Nov 29, 2013 11:06 am
by uldis
roadrunner79, please contact support@mikrotik.com on your setup and how are you testing your wireless link.

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Tue Dec 03, 2013 10:46 pm
by roadrunner79
@uldis - if you want i can arrange a lab setup and give you remote acces via teamviewer to see how it's working.

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Thu Feb 06, 2014 9:15 am
by changeip
Has anyone ever figured this out? I just came to the forums looking for this answer as I am noticing it as well. The past 3 long distance links we installed with 912UAG all show suboptimal ccqs. I am starting to wonder if I swap back to 411 / dbii or sr71 if this will fix it. Interestingly enough I notice ccqs sometimes 'stick' at 67% or 80% ... which seems like a calculation error rather than real life problems.

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Sun Feb 16, 2014 3:38 am
by changeip
okay I have had the chance to swap all our 912 boards for our usual 411gl board setup. In 3 cases we had links > 30km with ccq's all over the place and typically less than 80%. With no other changes other than the routerboard (912 to 411gl) we are now at solid performance, 99-100% ccqs, and no ethernet problems. What a nightmare the past month has been in respect to the 912 experience. Every single one of them is coming off the network - great idea but something is wrong with the design. I think the clocking isnt working - problems with timing on wireless as well as on ethernet I believe. 5.26 - 6.9 with varying results... none of which were good.

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Mon Feb 17, 2014 9:30 am
by pospanko
okay I have had the chance to swap all our 912 boards for our usual 411gl board setup. In 3 cases we had links > 30km with ccq's all over the place and typically less than 80%. With no other changes other than the routerboard (912 to 411gl) we are now at solid performance, 99-100% ccqs, and no ethernet problems. What a nightmare the past month has been in respect to the 912 experience. Every single one of them is coming off the network - great idea but something is wrong with the design. I think the clocking isnt working - problems with timing on wireless as well as on ethernet I believe. 5.26 - 6.9 with varying results... none of which were good.
How about real transfer speeds? Did you maybe tested before and after?

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Mon Feb 17, 2014 6:05 pm
by changeip
yes. all performance after the swap improved, including no more packet loss.

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Tue Feb 18, 2014 11:00 am
by pospanko
Khm,,, Unfortunately I can't use RB411GL because of low MTU - I use MPLS/VPLS. I could use RB711 but they are discounted. Now I'm like fish without water...

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Wed Feb 19, 2014 9:20 pm
by JanezFord
Did any of you guys get a response from mikrotik support staff on this issue? I am planning a project for a customer and rb912 was looking perfect for the job. Now I am not so sure anymore. I also need high mtu so 4xx is out of the question.

JF.

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Sun Apr 27, 2014 10:37 pm
by welan
Any news? Same problem here.

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Thu Jun 05, 2014 7:28 pm
by jmay
Old post but I'm having the same problems with 912's. I've been jackin with these things for months trying to figure out what I'm doing wrong only find this post where apparently everyone is dealing with it.

Frustrating

My CCQ's are all over the place unless I limit them to MCS11 or 12. Throughput sucks. I've got one link that gives me real world traffic of 13 megs down by 85 megs up. Even though ccq says 100%, no cat5 errors. Replaced cat5 cables, feedhorn cables, 911gs with new 911gs. No help. If I reroute traffic through my backup link which is 3 pairs of RB711's, 2 pairs of Motorola PTP 600s, 4 routers, and 4 switches I get 4 times the download speed. LAME!

I love mikrotik routers, I really do. But you guys are gonna have to ditch the wireless game. Heads up, we're buying more and more ubiquiti over here. Only problem I've had with them in 2 years was a lightning strike.

Before mikrotik tells me to submit a trouble ticket I have 6 links using 911gs with similiar problems. And yes, I'm using the latest firmware.

Re: RB911G/RB912G - poor ccq/performance in benchmark setup

Posted: Thu Nov 06, 2014 9:55 am
by roadrunner79
It seems that new AC units suffer from the same problem, poor ccq/performance...