Anyone test the throughput of an RB333 in PtP link configuration? What’s the max throughput that you are seeing? I’m specificially interested in the 5.3/5.8 bands.
I have done +70Mbit UDP one way on 5GHz turbo, Nstreme, normal AP->Station WDS setup. Cannot get Nstreme Dual working properly in indoor lab (3-5 Mbit FD)
/Henrik
In the next release the bug with NstremeDual will be fixed and you will be able to get 80Mbps FullDuplex UDP with 5ghz-turbo, framing-policy enabled and without connection tracking.
What can we expect in non-turbo/non-Nstream for UDP and TCP throughput?
IMHO it must be MAX wirelesly possible (and as you know it depends on the frame size)
Heeeyyy.. If this is true i will tatoo the MT logo on a special part of my body ;o)
When is RC4 ready?
/Henrik
If that’s the case, it’s time for 802.11N draft support (Though, haven’t noticed if there is a mini-pci card even made) 
at the MUM USA we demonstrated 802.11n on RB333
Didn’t realize that. When do you expect to have a MT based wireless card and support?
Well RC4 is out. Anyone got 80mb FD 
Ehh. Nope.. Dont know if it have to do with using the swivel antennas (maybe antennas too close to each others) but i cannot get it working properly. I have good signal (30-40dBm), using brand new R52 radios and antennas mounted in another cabinet from RB532 and i see good CCQ. I have done standard Dual N-Streme setup, disabled conntrack etc and only given IP address to N-streme interfaces. I can do around 60-70Mbit UDP on a single interface but using both radios kills it.. I get maybe 15/15 at the top. I only have 18Volts PSU, dont know if that might cause some problems, but it really looks like inteference. As things are right now i can pull more Dual-nstreme speed out of 532A with swivel than RB333. I have used both 2.4 and 5GHz turbo/not turbo and on different freq’s for testing I am not going to put 333 in outdoor boxes and on towers before i get better lab results.
MT guys, please give some info on physical setup and configuration of your 333 tests, this would make a good reference for the rest of us when trying to achieve what you guys have done..
Regards
Henrik
you should adjust the swiwel antennas so you can simulate different polarization for each link (put one antenna at 90degree and other on 0).Also you should lower the tx-power to 5 or 8 using prower mode ‘card-rates’ and you should generate the traffic through the router not between the router. Then you should be able to do 80Mbps FD using UDP Btest.
close to 60 FD tcp tested.
Test environment:
2 pcs rb333
4 pcs r52
4 pcs 22dBi grid
4 pcs slloss pigtail
4 pcs 1,5 m H1000 cable
distance 500m
elevation 35 deg.
urban enviroment
line of sight with some reflection
txA: 5180 turboA
txB: 5800 turboA
separated with polarization
signalA: -40
signalB: -42
tested with independent computer ftp connection
tcp high transfer (1500B)
disable csma=yes
compression=no
encrypt=no
all tests measured in 3 times (average)
with framer policy=none
receive test(high traffic only one direction in time):
A : 59-64 Mbit/sec
B: 68-61 Mbit/sec
cpu: 91-100%
receive-transmit test (high traffic both direction one time)
A: 51-60 Mbit/s
B: 24-30 Mbit/s
cpu: 95-100%
with framer policy=exact-size
receive test(high traffic only one direction in time):
A : 62-70 Mbit/sec
B: 59-67 Mbit/sec
cpu: 98-100%
receive-transmit test (high traffic both direction one time)
A: 64-70 Mbit/s
B: 30-34 Mbit/s
cpu: 100%
But nstreme dual has big problem in rcX not started automatically if the devices powered up nstreme2 connection seems to connected, but no traffic on link. if nstreme2 interface disabled/enabled everything works good. Not stable. Unusable.
But normal nstreme works good
connection 5800 turbo 2 r52 and 2 rb333 same environment
transfer rate 60-70Mbit/s
cpu: 69-75%
(nice)
but rb333 and rc4 also have big problem cpu jumps 1-39% or stand at 30-35% without doing anything (sic!) i dont known whats happen in background. but now 3.0rc4 usable in wireless(also with rb333) environment (and now better than rb532 with rc4)
even with idle router you will get such CPU usage as the Nstreme interface is working - it gives you this CPU usage. But you don’t need to worry, it doesn’t effect so much performance, when board is running at max speed.
btw, you can try out the framer-policy=best-fit and try to increase slowely tramer-limit from 2500 to 4000.
about Nstreme2 bug, could you please tell us more info how to reproduce it, so it could be fixed?