Well almost 60mbit but man am i happy!
NStream @ 5.8ghz with Parabolic Grids and Cm9 with a link over the top of our city about 9km
Not bad huh?
Wow…very nice.
We’re just about to set up an almost exact link, so good to see this performance.
What CPUs are you using on each end?
How long are your cable runs, and what type of cable did you use?
SMA
30ish ft of lmr400 with Athlon’s for cpu’s
That is mad good!!
Is this a dual Nstreme link or just Nstreme?
Man that would be 120mbps with bonding or 60mbps Full duplex.
Dude your kickin ass!
Just plain NStream, altho we could chuck in some 5ghz filters and another pair of radios to upgrade it to nstream2
There is ~40mbit Over ~65km, with nstreme-dual 
http://www.mikrotik.com/3index.php#nstrm
Is this 5Ghz turbo or regular?
It has to be turbo.
Any fast motherboards you could put in an outdoor case and get the same results?
I can’t come close to duplicating these numbers in a test environment. I have a new Intel Celeron MB with a CM9 connecting to a new Via 2 Gig MB with a CM9. Signal strength is -20. Running 2.9.26 on both, in 5Ghz frequency. Running a bandwidth test from one to the other, if I am using regular 5Ghz, non-NStreme, I get around 12-13mb/sec TCP. With Turbo, I get around 24mb/sec. With NStreme, I get about half of what I get without it.
This mirrors the results I get with RB532’s in the field. How can Beccara be getting over twice what I am getting?
When you are running with Nstreme, do you have polling turned on? We have found Nstreme very slow without polling even on PtP links.
Just one link at random, 5GHz plain, 21 miles, RB532 at both ends, Nstreme with polling enabled gets just over 20Mb TCP in each direction individually, or about 12+12Mb both ways at the same time.
Framer policy is dynamic size, framer limit 3200.
George
I did not have polling turned on. After turning it on and setting the link as you noted, I did see improvement to around 20Mb TCP in regular mode and around 30Mb in Turbo. Thanks for the tip! However, that is still half what was reported above.
We see stange things happen with signals better than -40, Try getting your signal to around -50 to -60 and give it a try, make sure your CCQ values are up around 80+
with the signal level of -20 you are overdriving the receiver on the card. Use manual-txpower with low power values to avoid this.
Absolutely 100% agreed. Don’t run more than a -45 signal or you WILL slow down.
George
That did it! I lowered my signal strength down to around -40 and achieved 66 meg. At least, when the Intel machine was sending the data. The Via machine could only push around 57 meg. Both showed 100 CPU usage when they were sending the data.
OK, other important things to watch for. Using OFDM in particular inside a room sprays RF everywhere and the OFDM engine has to work really hard to reassemble all the junk.
Tests should really be done outdoors between well separated test units with a clean LOS and a max signal strength of no more than -45.
Try that and you should see a pretty decent improvement.
How do you get your CCQ values higher?
As the CCQ is a measurement of your network connection/signal quality, you would have to improve your signal quality. How to do that depends on the situation:
- change frequencies to avoid interference
- better/shorter cables
- better antennas
- using only lower data-rates on the radios, like only up to 24 MBit
- …
Best regards,
Christian Meis