Normis You say no speed degradation due to irq sharing?
I did some desk tests and those clearly says that total thruput of RB912 is limited by the cpu/irq issue.
at the beggining of this test done over two 912 connected both wlans simultanously but transmitting
both directions, so using both wlan interfaces. then just one direction. And it proves speed degradation
because wlan were ready to go as high as 225mbps but cpu of 912 was maxed out.
when I started btest within connected 912's speed over wlan1-rx dropped from 225mbps to sad 115mbps
and You clearly see CPU is occupied with IRQ - 92% of CPU usage.
It goes Even worse is when set in real environment. I gave up with 50-60mbps = 100% cpu usage with configuration as follows:
RB CCR1016 < ---gig eth ---> RB912 (ospf to sim fullduplex) =======12km of perfect los/fresnel link======= RB912 or RB493G
test from CCR to RB493G was around 50 mbps one direction -send- with some addition of real traffic in background ( 10/2mbps )
Simple exchange 912 to x86 (d945g-sejt) and speed goes up to real wireless thruput 170-200mbps on a wifi link rated at 270/270
So if you say 912 is intended to have IRQ sharing feature, I can only say 912 is useless for me, and hope for minipci-e 802.11ac wlan card in RB912G
and gigabit thruput is kind of forget about it idea.
Do You really want to convince me that irq sharing has some advantages? if so, please count one.
You do not have the required permissions to view the files attached to this post.