What is the latest usable version of ROS for this device ? RB133 not RB133C, if it does matter.

It matters.

RB133c has only 16MB of ram, so it means noway with 6.x. You need to keep 5.26 if you have this model.

RB133 with 32MB of ram runs 6.23 with no problems.

My personal experience is that it maxes cpu when bridging between two wifi cards (R52) at 15-16mbits thru (one wifi card in, other card out) on 6.23. Maybe it would make more with better cards - not tested by me.

Ok, there are problems with 6.23 on RB133 with ethernets. Seems that even autonegotioation happenned at 100FD it runs only on 10HD effectively. Playing a bit with it I am able to get around 50mbits. but the settings are really weird and it is not good to set one side in different way then the opposite side to get it at least somehow working. Maybe ethernet queuing in cpu is the problem - but it was not changed during the playing with autonegotiation.

I made supout, sent to support and downgraded to 5.26 meanwhile. With 5.26 autonegotioation works correctly and is able to transfer around 50mbits in average (with 80-90mbits peaks), limited by queuing in cpu.

It is not possible to use "only-hardware-queue" for these ethernets.

I thought RB133 has some switch chip, but does not seem so. (not listed in switch menu).

I have not realised any other problem with 6.23, using wireless-fp package, snmp v.3, dude management, two bridges, no conntrack, no firewall.

For the prices at which the RB133 is available from previous owners it is still quite good device for places where price is more important than effective throughput.

Thank you for info.
I'd like to buy used and cheap ones with 2.4 & 5 GHz cards just for fun/testing so mentioned problems are not so important.
For dhcp/dns and CAPSMAN they should be more than usable.

I had the same problem with some RB133. Doing a bandwidth test between some other board and an RB133 results in 10Mbit/s throughput without maxing the cpu even though it shows 100Mbit-FD. This is is with v6.x, reverting to v5.26 fixes this problem.

Your welcome. As toys they are good. But older mipsbe devices are not expensive also but much more powerful.

Another findings in comparing 5.26 vs 6.23:
Looking at dude graphs I can say that 5.26 running on RB133 has lower memory consumption, faster response times and bigger cpu utilization (being connected via one ethernet, one wifi card as client and running ap on second card without any clients, no conntrack, no firewall). Also reboot time is significantly shorter.

Hope mikrotik would correct the ethernet behaviour in 6.23, but I do not believe in it. After this correction I would put 6.x back as I also would like to try how the capsman will be working on it.

Just testing the 6.24 on RB133. Looks quite good, but rebooting takes sooooooooo loooooooong and even after reboot the router is not in good state. It takes many and many minutes until cpu utilization drops down and everything start to behave "normal". First surprise was when I swithced on the SNTP. It turned cpu to 100% for at least five minutes immediatelly. After that the same effect was when I set the SNMPv3 and let dude to log in via SNMP.

After one night leaving the RB133 idle (1 ethernet connected, 2 radios in client mode connected, all interfaces running DHCP client, being monitored by dude, no other traffic) I can say that 6.24 is definitelly better on RB133 that 6.23. Even comparing to 5.26, it consumes similar cpu in average, more memory, but the responses (ftp, http, ping and so..) are a bit faster.

Of course, the ethernets are still working at 10FD even reporting 100FD.

I have sent this information to Mikrotik support as addendum to ticket nr:

[Ticket#2014121766000789]

as you can see, it is two weeks old and still without any reply. Agree, taking care about RB133 is not priority these days, but it should be workig correctly anyway. So I hope they will make correction when they have time.

Confirmed by Mikrotik that problem with ethernets is under investigation. Using 5.26 instead 6.x is suggested so far.

Good to know they work on it. And I hope it will be reflected on changelog when it is fixed.

I hope so

Just recieved info that this problem has been fixed starting from v6.25rc7.
Looking forward to test it.

For those who care:
6.25rc7 brings some improvement. With auto negotiation on it is able to pass 50Mbits at maximum in one way, with auto negotiation off (100FD) it passes 64Mbits in one way.

Not perfect, but on the good way. Looking forward for next improvement.

50 and 64Mbit on rb133@175Mhz? What settings?

Ethernets in bridge, no other traffic (just monitoring from dude few oids), dhcp client to get ip, time servers, default route. Otherwise blank configuration.

Tested with udp bandwidth test ended at RB133 device (I know, I know...). Will test traffic passtrhu soon.

I wrote that it is not perfect yet mainly due to higher rate when auto negotiation is off - but it should be the same even when auto negotiation is on. This should not have any influence.

6.25 "killed" my RB133C ... ten, twenty....fourty minutes after reboot there is no sign of life in Winbox.
No MAC detection.

6.25 "killed" my RB133C ... ten, twenty....fourty minutes after reboot there is no sign of life in Winbox.
No MAC detection.

From the MikroTik Download page there is a note:

RB111, RB112, RB133C, RB150 devices are not supported in v6, please use v5.

Please use Netinstall and install v5.26 for this board as v6 will not work on this RB133C.

133c doesn't have enough memory to run 6.x. Minimum is 32mb but it has 16mb only. To netinstall 5.26 is the only option.

Recovered to 5.26. Works.