Mikrotik WLAN & CAPsMAN - Bad download perfomance

RouterOS Wireless Networking
1

The problem is - and I don’t really have any idea what causes it - wireless speed is cut roughly by HALF from what it is supposed to be.
The picture is:
NAS -gigabit> Mikrotik CSS326 -gigabit+PoE> wAP AC -wireless conn 802.11ac 2x2> PC with realtek 8822BE wireless module.
RB3011 is connected to CSS326.

The issue itself? For example, I’m trying to download from NAS a movie - a whole videofile, MKV packed, 30GB size. But instead of 60 MB/s, I have only 30MB/s. And I really don’t know what causing the issue - and that drives me mad. Moreover, connection itself is super-stable - it just looks like something caps it at 30-40MB/s.

Next things I’ve tried:

  1. Wired connection gives me ~900 Mbps. So switch is out of question, as well as maybe RB3011.
  2. wAPs powered by PoE - cable cat 5e, 8 wires/4pairs. Everything is in working condition.
  3. Aside of 8822BE, I’ve tried killer chipset and Intel AC-9560 - all the same, it just stucks at 25-40 MB/s.
  4. Also tried making simulatenous workload for wAP, by downloading the same file on two devices (Killer and Realtek, Intel and Realtek) - speed just divided by half, giving 15-16 MB/s for each device.
  5. Local storages are non-issue here - NAS works with two SATAIII HDDs, united in RAID0. Both client devices have NVMe drives with 3GB/s sequential write.
  6. As well as it’s non computing issue - NAS CPU is barely loaded, not even mentioning the PC’s.

All in all, here’s the screenshot - I guess, it would help to make thing even more clear, than I could writing all the stuff down…


And here’s my expectations, or should I say, just “wish” - Intel 9560 + UBNT AC-SHD
http://piccy.info/view3/13040178/1130322bb7bb25a74abe6d210364e022/

Re: RB4011 internal WLAN - enabling CAPsMAN instantly halves speed
2

How do you decide how fast your wireless speed is supposed to be?

My hap ac2 802.11ac 2x2 speed with capsman and local forwarding is indeed about half of what you would expect from the standard under perfect conditions, until now I’ve been blaming it on cheap hardware and poor drivers. But considering the various routeros wifi bugs, which have been fixed in the past months, I wouldn’t be surprised if that difference between theoretical and practical speed was really caused by some hidden bug. As long as there is no fix we will never know :slight_smile:

3

I suggest my speed from these factors:

  1. Both Tx and Rx have 2 chains working simultaneously in 2 streams - thus making 866 mbps real. wAP AC and Realtek 8822BE - you can find specs for both. 2x2:2, what can go wrong?
  2. Since mikrotik admit their wireless protocols far from ideal, thus output is around 50-65% - it makes 500 mbps real.
  3. All wired conncetion rated at 1 Gbps.
  4. I’ve made channel reconnaissance and setup, so no overlapping channels, NOT EVEN CLOSE.
  5. Registration table reports, device is connected at 866 Mbps rate, with 2 streams up and perfect signal strenght : -30-40 dBm.

So, I think there’s no doubts left that download speed MUST be at least arount 60 MB/s. It better be 100 MB/s really, since 802.11ac states 866mbps for 2x2:2 configuration, and I have nearly ideal/perfect conditions, so I don’t really get why mikrotik can’t provide at least 90% of full 802.11ac potential. :confused:

4

Have you tried local forwarding vs capsman forwarding to see if there’s a difference?

5

Of course I did - and the matter of fact - I ALWAYS use local forwarding enabled, cause capsman forwarding gives me like 10% of wireless perfomance.

I’ve tried disabling other CAPs, both 2.4 and 5, I’ve tried removing all the clients from tested CAP - no use, nothing helped.

6

Personally, I get > 100Mbps using wAP AC + RB3011 running CAPsMAN, local forwarding.

How crowded is the wireless around your place?
Are APs locked to CAPsMAN and using encryption?

7

Could you post a screenshot of winbox CAPSMAN-Channels?

8

But you shouldn’t you expect > 500Mbps with three chains on 5GHz ac? I consider 100Mbps with that hardware pretty slow, that speed is already achievable with 2.4GHz dual chain n.

9

I live in suburban area, big house, closest neighbout is literally ~100m away. Not corwded at all - area is “crowded” by only me, myself and I.
No encryption, and no, CAPs don’t locked on CAPsMAN - does it positively affect the perfomance?



That’s what I’m talking about - and I’ve metnioned it already (no?) - that using dual-chain 2.4Ghz band with 802.11n protocol, I can’t even beat through Internet channel! 100mbps are being choked by poor wireless perfomance. I mean, I’m experiencing overall wifi perfomance drop, not only PC and 5Ghz 802.11ac connection.


And the most desperate thing about all this - I’ve tired reseting all the stuff, leaving onyl default configuration by mikrotik, and making fast provisioning for caps, just ot ensure it’s not my fault by too overcomplicated settings - aaaand all the same, 300 mbps cap in 802.11ac5GHz, and 50-70 mbps cap in 802.11n2.4Ghz .
5 MB/s by air with 802.11n 2stream is waaay to bad.

10

Looking at your first screen shot in your first post, I’d say that’s a decent connection.
You’re connected at VHT MCS 7 with two spatial streams and short guard interval resulting in 650MBps connection rate at 80MHz.
Keeping in mind that actual real-life throughput is roughly 50…65% of the w/l connected rate, you’re in a pretty good shape.
Increasing throughput further from here from here would ideally be a different wireless client supporting three spatial streams which would give you 975MBps @ VHT MCS 7, resulting in roughly 500…650 MBps real-life throughput.

-Chris

11

MAIN2.png

CAPsMAN-Channel. Control Channel under CAPsMAN is what I need to see it should be in auto mode.

12

When I lived in the city, I’ve used to have 1Gbps internet channel. Though, it was roughly 1 Gbps - more like 900 mbps at night (best), and down to 600-700 mbps daily. And asus RT-AC66U as a router - no complicated wifi stuff, just a a swetpot with 3 antennas.
I’ve had 450 mbps on my current iPhone 6S result in speedtest, in 5Ghz ac band. I don’t know if it has 1S or 2S - and I remember that device is 1st wave ac, so 2x2:2 40Mhz at best.
So I don’t feel THIS like “pretty good” at all.

What, do you mean frequency? No way, it gonna throw overlapping ones in.
Or channel width? But is there any difference between I manually set 20MHz + Ceee, or I just disable that tab, leaving Ceee?

13

Ok, so I’ve checked streams, leaving 1 stream per test phase - so far, I’ve managed to discover it IS streams issue.
Thus, when leaving only 1 chain activated and 1 stream flow, I get the same 30 MB/s - 270mbps. So wireless perfomance is stucked at this point no matter what - whether I turn on all chains available (and possible), leaving 3, or 2, or 1 - it looks like it always works within 1 stream. And that’s a real issue.

I’m 99% confident it’s mikrotik’s side flaw - for that matter, I found PCE-AC68 adapter for a good price, so I’ll be able to write down my final verdict soon - but that’s how it is for now.

14

Aaannd…what needed to be prooved. I’ve just received my AC68 unit, plug-it in, up-to-date drivers, of course - and I get the SAME goddamn speeds. With 3 streams up now…

P.S. And don’t tell me that’s ok (if you read this at all, mikrotik support), that it’s only “50%” etc, etc…

15

So MIMO is broken in some way that prevent any speed gain from additional chains? Interesting observation, let’s hope that this is reproducible and fixable by mikrotik.

16

Yeah, seems like it. Furthermore, as I mentioned before (here or in support ticket?), MIMO seems to be broken broad - “at all” - if I can say so - because I really can’t imagine how it can be broken.
Both, with CAPsMAN involved and without it, in regular AP mode, 2.4GHz or 5 GHz band, 802.11 ac or n or b - doesn’t matter, MIMO doesnt work.

It’s even more confusing that I was unable to restore previous state of things even downgrading down to 6.38 versions - does it mean it’s low-level hardware related?

17

So, anyone? 6.45beta, still all the same - can’t be only me experiencing this stuff. Or can I?

18

Hi RobCo,

I really feel your pain, I do.

Although I’m not convinced my post will help you, it may spark some insight in you.

1st thing - are you sure your clients are actually WIRED with 2 separate antennae?

2nd thing - for MIMO to work well one needs a good spatial separation for all spatial streams, i.e. there should be no correlation between the 2 spatial streams. This is hard to achieve with ordinary antennae and LoS between AP and client.

You could also try iw phy from most Linux boxes to check if your client WiFi cards support MIMO.On Windows you can try netsh wlan show wirelesscapabilities from admin CMD.

19

yeah, np mate, any idea is good!
Though I’m not a guru, and really have no idea at this moment whether it’s hardware has it’s flaws and works worse despite all specs available for public, or… it’s just me.
Anyway, I can only relate to software I have - so, as you can get it, you can’t trust it 100% → can’t be sure 100%.

1st, Physically, all chains are active, all antennas wired. The reality is - I don’t know for sure :smiley:
What I see in registration table (screenshot attached), it sees 3 spatial streams though it’s really doubtfull info, suggesting from rates it shows and the real world throughput I get
Regarding the cmd promt - even within administrative mode, it shows almost nothing usefull:
For AC68 usefull info is missing at all - either “value not available” or “not supported”, even if it IS and I know it. But still, no info regarding “number of antennas”, “max number of channels, simultaneously”, no info about regular MIMO at all, etc.

For 8822BE, a little bit more, but still not enough - no info regarding antennas, channels aswell - but at least it shows it does support MU-MIMO (no info regarding MIMO), number of Tx Spatial Streams = 2, Rx spatial streams = 2, concurrent channels = 1.

Regarding 2nd - really no idea about the built quality or engineering idea behind all of this mate - and I don’t know any mean way to tweak these for wireless clients.
LoS is…well, ATM of testings, it’s around 2-3m. No obstacles, no metal construction.
I’ve tried stepping off a little, 10-15m, a thick brick wall 15cm in-between - but no improvements. Moreover, I have feeling like the signal itself is great, it holds those 300mbps and 50mbps really good - for a wireless solution, almost like a wire.
Dunno, maybe I want too much - though the things I want from this hardware are not stellar. And also I’ve decided to check bandwith with iperf so people would get something more understandable, than messy “I download a movie from NAS”
So here it is: AC68
8822BE
Strange, isn’t it? I mean, 1st - no difference between 2x2:2 and 3x3:3 client? I know Realtek chipset is much newer than asus - but asus is way more powerfull device. 2nd - well you can say that 5GHz perfomance is good, but shouldn’t it be much better with 3x3? Like, let’s say 800ns AC1170/2 = 585 mbps at least? Damn, I agree for 500, but 300? And what about 2.4GHz?

20

Yes, you’re totally running @ 1 spatial stream, i.e. 1 chain.

Also, you’re not wanting too much of your HW, you’re just wanting what you paid for, which according to MT’s own page is 1300 (or 900Mbps if you consider the real life). What you get is a far cry from the advertised speeds…

Check this out - https://www.cisco.com/c/dam/en/us/products/collateral/wireless/aironet-3600-series/white-paper-c11-713103.pdf#_Toc383047840

According to the link from this renowned manufacturer of networking equipment above, you’re not even getting 256QAM! 256QAM should yield ~433MMbps.

Honestly, I don’t know what else to suggest for troubleshooting. Had we some fancy test equipment things would have been easy. Maybe if you can temporary get another wAP for testing, you could try setting up a WiFi network between those 2, and see how they play it out?