Hello,

I have purchased gigabit version of SXT for my ptp link, I tested it in my lab and results are quite good,
I achieved 145-155mb/s real TCP download with ~4mb/s upload at the same time,( I was testing only download)
cpu usage <85-90% peak;
I imagine what rb800 or rb411GL will do,


I will try to achieve these characteristics in production in ~3km range,

Post your results here, please? I'm thinking to upgrade one of my links with these..

This is my lab setup, range 3-4m from each other , tx power=1dbm,

This is pure TCP test? And that ~3mbps upload is TCP ACK traffic?

Yes it's pure TCP, not btest tcp, but actual download from one pc to another,
with Jperf same result,
this throughput screened , was there most time, range was from 145-165mb/s

upload is pure tcp ack + microsoft rdp traffic

Could you post some of your config? At least tell if you use nstreme or nv2?

sure,

I experimented with nstream and nv2 too, what I noticed was that nstream had more stability in throughput then nv2 ,
but it gave ~122mb/s of tcp, nv2 was jumping to ~150-160 but then was dropping to 60mb/s-90mb/s for a second and then up again, and I did not like that--- This was with default config,
after some tuning I achieved nstream like stability for nv2 but higher throughput then nstream,

yes these results from the screenshot are from tuned nv2, but should mention that this was in my lab: range 3m,
I still do not know if this will happen in production network, at least I know that nv2 and SXT CPU is capable of pushing more then 150mb/s real TCP, another story is if I will achieve good wireless characteristics in production link to have same good throughput there,

if you want I can give you more details how I tuned nv2 and wireless settings,

Of course I do And I'll wait for your production-link results.

I recently have set a link up from two SEXTANTS, 5km apart. One of them are on heavy duty tower where all the channels are polluted a lot. Signals are -60/-60 both chains both ways. The best result I managed to achieve was with Nstreme best fit, that was ~70-80mbps in speedtest.net and ordinary http file download. Anyway, I'm looking for some additional ways of improving it.

quite good results, SEXTANTS have 100mb/s port and less power of cpu I think, in your situation it's quite good result,
I have 3km link with nanobridge 25dbi and I have 80mb/s throughput there, clean LOS and free spectral environment, but 8 devices on 5ghz band on the same tower,

how much is peak throughput loading your cpu? I think theoretically you have chance to improve to +10mb/s,

And do not trust speedtest.net often it shows less throughput due to server loading or some other factors then you actually have,

I mostly use Speedtest Mini hosted on my own web server. It has absolutely no load whatsoever.. And 100mbps port shouldn't be problem up to 100 itself, right? The CPU load doesn't seem to be a bottleneck (as pointed in the image attached). But seeing your experience I now see that some fine-tuning might squeeze more juice from that link

As for CPU speed, I think that all S*T devices share the same CPU (400MHz AR7240) and gigabit versions of 711 or SXT doesn't really kick-in up to 100mbps. If SEXTANT had good dual-pol antenna, I would consider swapping RB711-5HnD to RB711G-5HnD, but for now I don't see any benefit of doing that.

Also, I'm thinking to attach this cover to the SEXTANT that is on heavy-duty site:
http://forum.mikrotik.com/viewtopic.php ... 91#p310191

As you see from my previous post there is more then 100mb/s with SXT G version,
G versions of 711 and SXT do have much powerful cpu, no doubt, at least 3x power then SEXTANT and 100mb/s versions,
if in your setup during peak throughput cpu is less then 90%-95% then cpu is not your problem, but if your wireless is well tuned and you look that throughput is high but cpu can't handle it then you must upgrade,

if you want more then 100 consider upgrading to gigabit versions, I can hardly imagine SEXTANT cpu to make 100+ even if it had gigabit port,

for cpu comparison please see file attached, I can confirm these new series AR7242 pushes lot more throughput then ar7240 and 7241 even they have same frequency, you can check throughput tests on http://routerboard.com


shield need to be mounted I agree, and one more thing I looked your screenshot and noticed you have ~20mb/s drop in throughput it does not seem to have enough stability, 60-80mb range I mean, I think when testing with consistent download the amplitude should be no more then 5-10mb/s,
Thanks

About SXT G cpu:
the brochure says: "Atheros AR7241 400MHz CPU" , see it http://routerboard.com/pdf/396/SXTG.pdf
So which one is wrong: AR7242 or AR7241 ?

Normis, anyone ?

Please correct this confussion.