Indoor test with 5dBi omni antenna:

17mbps TCP and 40mbps UDP !!!

RB133 (Bridge 2.4 Turbo)------------->RB133c (WDS 2.4 Turbo)
Everything disabled;only bridge; CPU max 70%, 18v/1A power adapter.

The speed is not too low?

That speed is fine IMO.

That speed is fine IMO.

"fine"?
With 2 x Linksys WRT54GL (worst hardware than routerboard) you can push up 22/23Mbps TCP on 20MHz channel (so, NO turbo)....

Aye but imo the linksys will crap out as they always do , and get hot if your using hacked firmware. They arent very stable, ther are also tweakable settings inthe mt. Are you doing apples to apples or apples to oranges, what cards do you have inthe mt and what mw is it set at ?

That speed is fine IMO.

"fine"?
With 2 x Linksys WRT54GL (worst hardware than routerboard) you can push up 22/23Mbps TCP on 20MHz channel (so, NO turbo)....

Yes "fine". Top ethernet performance on the RB100 series router is 50mbps. He's getting 40mbps UDP over a wireless, so I don't think that's bad at all. Instead of using the RouterOS bandwidth test try using two PCs, one behind each RB100 and see what results you get.

http://routerboard.com/pdf/results_18may2007.pdf

Edit: Ugh I see bridge without a firewall is about 112mbps, but this is a wireless links using WDS. I wouldn't expect too much from the RB100 boards.

the linksys will crap out as they always do

This is your opinion.

and get hot if your using hacked firmware.

This is not true. I installed a lot of WRT54GL "modded" to 180mW outdoor (in Gewiss or Pac Wireless enclosure), and they are rock solid for over a year. No maintenance at all.

They arent very stable, ther are also tweakable settings inthe mt.

I think it depends on the firmware version you used. I tried many third party firmware, and imho the best is dd-wrt.
RouterOS has many option, and it's very tweakable (exception for fragmentation and RTS/CTS).

Are you doing apples to apples or apples to oranges

Mmmhh....no. RB100 + mPCI is a wifi devices, Linksys is a wifi devices. I'm comparing "potatoes and potatoes"

I wrote this not to start a flame, i'm a very big estimator of MT RouterOS (i used ONLY ROS on my new devices since about 1,5 year), and with it i did something that with other firmware where not possible at all.
But MT has produced (imho) some board that are not good for high bandwidth link.

My two cents

Regards, Mirco.

17mbps TCP is acheve with computers at both sides, no btest. And as I said it's not a CPU limitation, CPU is at max. 70%. In Turbo I think that 24mbps TCP is more realistic not 17mbps. Without Turbo the speed is 14mbps TCP.

http://www.mikrotik.com/pdf/test-RB100-wireless01.pdf
it is an older document, but as you see that you can get more than that what you get.

actually i ment on apples to apples was is the mw the same in the cards, certain cards work better than others

I see that a long time ago. Not my fault that it can't no more than that. Anyway 90% trafic is TCP in a network, thats why you should post speed in TCP but you didn't do that because they are just too slow. And I repeat the CPU is only max. 70% .Signal -53. It's good but I expected more.

Our setup was like:

RB133 {bridged ether2 & wlan2} {wlan2 mode = bridge, wds dynamic}

|
|
v

RB133c {bridged ether1 & wlan1 } {wlan1 mode = station wds}

(We followed this tutorial: http://wiki.mikrotik.com/wiki/Transpare ... o_Networks as it seems that you CAN'T bridge an ap client - anyone knows why?)

The bandwidth test was done from Computer to Computer, one on the RB133 side and one on the RB133c.

As my buddy InoX said, we couldn't get more than 40Mbps on UDP, in 2.4Ghz-Turbo, without Nstreme/Polling. In TCP we got around 16Mbps (by copying a file from a computer to another over SAMBA - Windows Sharing, btest result was even lower, around 14Mbps).

When enabling Nstreme our results were even slower! We tested with the default frame size. We've been told that we should try a frame size of 3200... We will do these days... We will also try routeros v3.

We didn't have any routing, no firewall and no conntrack.

1. Is the performance so low because we're using WDS and Station WDS?

2. Why can't we bridge ether + wlan when wlan is in mode=station? It sounds really silly. I mean all cheap hardware Access Points have this feature, but MT wich is worth a lot more can't do this simple task? I know it's not a linux kernel limitation, because many hardware access points that run linux do work like this. Even OpenWRT does this!

3. How were those tests in the PDF made? If you can't bridge when mode=station?! Please explain how you can bridge ether+wlan without wds? Can you do that with a level3 license (the one on the client side)?

Thank you.

natedogg104 wrote:
the linksys will crap out as they always do

This is your opinion.

Nate is right. Linksys will crap out. A very inferior product compared to the router board. I live in an area with alot of thunderstorms and it seems that a slight surge or brownout will many times reload factory defaults or fry the router.

Eric

2. Why can't we bridge ether + wlan when wlan is in mode=station? It sounds really silly. I mean all cheap hardware Access Points have this feature, but MT wich is worth a lot more can't do this simple task? I know it's not a linux kernel limitation, because many hardware access points that run linux do work like this. Even OpenWRT does this!

I understand your disappontment, but if you search a little on this forum, you will find many threads talking aboyut the new ap-mode on ROS 3.x: pseudo-bridge. It's the ap-mode which you are looking for.

Nate is right. Linksys will crap out. A very inferior product compared to the router board. I live in an area with alot of thunderstorms and it seems that a slight surge or brownout will many times reload factory defaults or fry the router.

This is not a good comparison term. For example, on the last 1,5 year, i burned by a thunderstorm only one card: an Ubiquiti SR2 mounted on an RB532 (which was not damaged). NEVER bourned a Linksys.
Another example: every time there's a thunderstorm, i burn (often) one of our Cisco 803 ISDN backup router (and it's protected by UPS on the power and phone line! I bourned 4 of that on last year, no other devices was damaged). On the same condition, no zyxel or other brand (economic chinese devices included) were burned....so: is a 700€ Cisco router worst than the other devices we are using?

I can close here my post on this thread (except, if i can, for helping our "friend" on it's bandwidth tests), and i'd like to repeat that, without Mikrotik, we can't made a 304km wifi link!

Regards, Mirco

id also have to say thats passing data thru both devices which is processor conumptive, what do you get it you go from device to device with bandiwith test build into the mt, also id likes to say the 133c isnt what you would want to use for a big link anyhow, you may want to use a 333 or 532a. I wouldnt use it or a linksys for any point to point link. just asking for failures in any case/

I will try locking the data rate and play with the transmit power. I will post the results.

(We followed this tutorial: http://wiki.mikrotik.com/wiki/Transpare ... o_Networks as it seems that you CAN'T bridge an ap client - anyone knows why?)

Its part of the wifi standard afaik.

Anyway these are only little boards intended for clients or small APs, if more throughput is needed buy one of the newest boards.

Then let's hope we get v3 asap

Hello again, and sorry for the double post.

We tested with RouterOS v3rc4 tonight and here are the results:

UDP - from Computer to Computer (same setup with WDS + Bridge): 57Mbps one way, with the bandwidth test.

TCP - from Computer to Computer (same setup, wds + bridge): 37Mbps one way, using FTP. The bandwidth test (btest.exe) was _way_ slower, around 17Mbps.

The CPUs were MAXED OUT at 100% both on the rb133 and rb133c, so I guess you can actually get more with more powerfull CPUs. Ping times were acceptable, around 20ms.

The test was done indoor.

So yes, v3 really kicks the llama's ass!

Btw, I wasn't able to make the station pseudobridge work... I set the rb133 as ap bridge and the rb133c as station pseudobridge, bridged the wlan + ethernet on both sides but the computers weren't able to ping themselves. I'm guessing I did something wrong, but seing there are no available docs for v3 yet... Maybe someone can shed some light on how to actually use this?

Thanks

Any chance for a firmware with overclock options for RB133? At least 200-233Mhz option!

Hello again, and sorry for the double post.

We tested with RouterOS v3rc4 tonight and here are the results:

UDP - from Computer to Computer (same setup with WDS + Bridge): 57Mbps one way, with the bandwidth test.

TCP - from Computer to Computer (same setup, wds + bridge): 37Mbps one way, using FTP. The bandwidth test (btest.exe) was _way_ slower, around 17Mbps.

The CPUs were MAXED OUT at 100% both on the rb133 and rb133c, so I guess you can actually get more with more powerfull CPUs. Ping times were acceptable, around 20ms.

The test was done indoor.

So yes, v3 really kicks the llama's ass!

Btw, I wasn't able to make the station pseudobridge work... I set the rb133 as ap bridge and the rb133c as station pseudobridge, bridged the wlan + ethernet on both sides but the computers weren't able to ping themselves. I'm guessing I did something wrong, but seing there are no available docs for v3 yet... Maybe someone can shed some light on how to actually use this?

Thanks

I came near to this numbers with rb500-399 2.9.46 5ghz nstream routed
*without* turbo. So it makes no real sense to twiddle around wlth 100-boards if you need performance. Spend the few euros to get faster
boards. (I would buy 333 if 3.0 final is there )

Stefan

is that nstreme2? 'cose I find it hard to believe if even routed without turbo...

and I usually get x86's if I need good speed and space is not a problem.

Any chance for a firmware with overclock options for RB133? At least 200-233Mhz option!

you can do it with a resistor but not sure of stability at 200MHz as its pushing memory beyond its rating somewhat.

Bill

And where this resistor may be?