RouterBoard 133, 532A, 333, 600 Speed tests

Hardware MikroTik hardware general
1

Roc-Noc.com Mikrotik Routerboard Radio Speed Benchmark

Test Date - 12/24/2007

Tested by - Tom Harker - CCNA, USA Mikrotik Distributor

Purpose - to compare radio performance between four different
Routerboard models as well as RouterOS 2.9 vs 3.0 in lab conditions.
Tested models are RB133, RB532a, RB333, RB600.

Test Bench - RB600 at each end running Mikrotik Bandwidth test
to eliminate loading CPU on Routerboard under test. R52 Atheros
cards with Mikrotik indoor swivel omni antennas placed 8 feet apart.

Wireless test parameters - 5180MHz dynamic wds bridge, nstream on,
conntrack disabled.

Network - Point to Point bridged configuration.

RB600 - RB??? - R52 - AC/SWI - - - AC/SWI - R52 - RB??? - RB600

10.10.10.100 - 10.10.10.1 (AP) ---- wireless link ---- 10.10.10.2 -
10.10.10.200

Bandwidth Test tool run from 10.10.10.100 to 10.10.10.200.
Routerboards under test were 10.10.10.1 set as ap bridge
and 10.10.10.2 set as station wds.

Test Results:
udp udp tcp tcp
Routerbd/ver rcv both rcv both

RB133 2.9.50 24.0 15.0 20.3 11.0
RB133 3.0rc13 28.5 16.1 22.7 12.0

RB532a 2.9.50 33.4 18.4 27.9 15.5
RB532a 3.0rc13 32.0 16.9 26.6 15.0*

RB333 3.0rc13 32.5 17.7 28.7 15.6

RB600 3.0rc13 36.4 18.4 32.2 16.9

  • Very unstable. Speed was jumping all over the place.

Comments: This benchmark compares Routerboards in a nearly
ideal environment with very little CPU loading. If you are
only using the Routerboard as a single radio controller then
you should expect similar results. If you are running more
than 1 radio per Routerboard or using the routing and firewall
functionality on the same Routerboard, then expect your radio
speeds to be much slower on the smaller CPU versions such as
the RB133.

Other than those previously mentioned, most settings were left
at their defaults. No attempt was made to tweak the settings
and improve the speed. That will be a future benchmark on
a single radio.

It was interesting that the version 3.0 firmware improved
the wireless connection on the RB133 yet was worse running
on the RB532a. Both have the same processor and run the
same firmware build. It appears that version 3.0 firmware
is not quite ready for production on the RB532a.

I am impressed at how well the RB532a using the older 2.9.50
firmware performs as a radio controller. The RB600 is the
obvious choice for multiple radios or a heavy duty router or
firewall.

copyright 2007 Tom Harker and http://www.roc-noc.com
May be reprinted only in its entirety.

\

http://www.roc-noc.com
Tom Harker, CCNA, ISP, USA Mikrotik Distributor
888-762-5662

2

I have beat you with just an Rb133 ~57mbps UDP ROS 3.xx !!! Use Turbo :laughing:

3

Well, Tom did the tests not me, I’m just helping him\everyone else out.
I’d rather have 36 megs in 20 MHz than 57 megs in 40 MHz… more spectrally efficient. Also not sure how you were able to pull 57 megs wirelessly on a 133.

4

The results are real man. Read this:
http://forum.mikrotik.com/t/bridge-betwen-rb133-and-rb133c/15907/1

5

Doesn’t seem to be much discussion about whether Nstreme is enabled, and if so, which features are enabled.

Makes a huge difference, and our real-world results are a lot different to those reported here.

George

6

With nstreme on RB133 is the same with 2.9.50 and worse with 3.xx . The link is indoor, for testing purposes only!!! Nstreme is for long/heavy trafic links.

7

Inox, my comments were regarding the original post, not your hijack.

The original testing posted by Hammy does not refer to Nstreme, and Nstreme makes a big difference to PtP performance particularly on cards with higher processor capacity.

George

8

Hey George,

I ran the tests. NStream was on. Read the original benchmark post again.

“Wireless test parameters - 5180MHz dynamic wds bridge, nstream on, conntrack disabled.”

I left most of the settings at defaults. This was NOT an attempt to get the most speed out of a Routerboard. That will be a future benchmark test.

This was an attempt to compare the various routerboards in production and the two current RouterOS versions. I am using these settings in many of my PtP links. I have found that turning nstream on generally improves my throughput. I don’t use turbo mode as clean radio spectrum is really scarce here and 40MHz too greedy for my tastes. Many of my PtP links are set to use 5MHz or 10MHz bands.

This was the first time that I ran the MT Bandwidth test on routerboards that were not controlling the radio cards. That left more of the CPU for the radio control. Running the MT Bandwidth test program on the actual routerboards with the radio cards results in much different results.

Tell me more about your real-world results.

Tom

9

Tom,

Assuming you have polling enabled as well as nStreme.

We see about 23Mbits UDP on a 532 at 32km using 2.9.48.

About 35Mbits on a 333 using rc13 with a 20MHz channel.

And around 75Mbits on a 333 using a turbo channel on rc 13.

In all cases nStreme is on with polling enabled with signal in the high -50s to low -60s and good SNRs.

So there is quite generally quite a difference between RBs, (btw 532s are pretty much a waste using a 40MHz channel so we never made the effort to find clear space previously).

We have seen quite a big difference in performance between 2.9.48 and rc13 on a 5MHz channel using an SR9 as well. Around 25% better…

Haven’t tried a 600 yet, only have one and it won’t be on a tower until after the holidays.

George

10

George,

Have you compared RB532a with RB532? Notice that I only tested the RB532a and not the RB532. Since my test showed it about the same as the RB333 and close to what you are seeing on the RB333 then I wonder if the increased CPU speed and increase in RAM on the RB532a (400MHz and 64MB RAM) made the difference?

I guess I’ll have to include the standard RB532 (264MHz and 32MB RAM) in my next benchmark.

Tom

11

We use a small number of 532As, but haven’t found any throughput difference at all from a regular 532.

In addition, all the rev5 532s we have are 400MHz. There is no processor speed difference between them and the 532A. I’m please to see a RAM increase with the 333, but more for smooth running under V3 than anything else.

George