I'm lab testing a wireless bridge on 5ghz.

hardware: routerboard and cm9
software: 2.8.18 with wireless (no legacy)

the best real speed I can get is 13.5mb/s both way. I used speed test esternal to the routerboard to not overheat the cpu.
I got this number using polling and framer policy to dynamic.
During the test the cpu go to around 70%-80%

Enabling Nstreme to speed go down to 1.5-2.5mb/s and the cpu usage doesn't go above 30%.

Any suggestion why there is such decrease in speed?

thanks

Hi

That's about all you can expect on RB's- 24Mbps one way, or what you see - ~13Mbps both ways - it's limited by the CPU (266MHz low power consumption).

Nstreme adds computational overhead, so that's why you're seeing the decrease. However I'm surprised by the low utilisation you see here also.
Check the signal strength both ends, Nstreme only works well at signal strengths better than about -65.

Using fast 1GHz CPUs we get 74Mbps through the same radio cards using MT 2.8.18 at 5.8GHz, 5GHz Turbo, Nstreme enabled.
We've now packaged this up into a nice integrated product in a single case which is "doing the rounds" as a beta and shortly into production.

Regards

Stephen

may you give more info about the 1ghz board?

thanks

Hi there,

We tested a fair number of mini-ITX boards from various vendors.
Results on EPIA V and M series were not good, with poor/erratic throughput on V due to PCI conflicts and lock-ups on M when at full speed.
Our conclusion is that consumer-grade mini-ITX's are not suited to professional environments.

The current configuration running uses industrial mini-ITX Pentium 'M' or Celeron CPUs and is excellent, with sustained 74Mbps wireless throughput in the lab. Shortly we will replace this with lower-power/heat solution using VIA CPUs for a weather sealed external unit. That will have a custom motherboard so no need for all the extenders and mini-PCI adaptors, and a proper 48V telecom power supply.
Right now we have the ITX-form board boxed as an initial product, and as we're an equipment vendor, not a user, that's the form we are going to offer to our customers, all fully tested and with MT loaded.
As this is MT's forum, not a place for our product advertisements, do contact me off-line at stephen.patrick@cablefree.co.uk if you have any interest.

best regards

Stephen

Are you saying that the CPU speed affects the throughput only if using N-Streme or will it also affect throughput using the standard wireless protocols ?

Both.

With 266MHz RB's, 24Mbps is top speed regardless what you do - the CPU is limiting.
With 1GHz CPU, you can push the radio to the maximum and get over 70Mbps.

If you have a slow CPU, turning Nstreme on slows you further as the CPU is the weakest path.
But on a fast CPU, the throughput improves using Nstreme, as the processing overhead doesn't cause a bottleneck.

Hope that helps

Stpehen