hello everyone, wanted to share my experience here. Briefly summarized I have a round dish and the following experiences made: the starlink router is garbage and the ethernet port brings only half of the bandwidth, hap ac2 has too little cpu power for a proper configuration that the bandwidth really goes over (in my tests only half of the bandwidth with hap ac2). my config where I am now very satisfied::
1.) Mikrotik Audience
2.) Directly plugged in instead of the Starlink Router
3.) On the MT Wan set to DHCP (which would be set by default anyway)
4.) Mesh with 3 additional devices built and everywhere in the house over 200Mbit download and 30Mbit upload.

so no fear - away with the starlink modem and own router to plug and his traffic under control to have, important just a new mikrotik with decent cpu and also no fear despite cgnat I have the default nat active and work normally via the firewall filter rules everything works fine and not different as with other internet connections.

Good to hear.
It would not have surprised me if they would lock it down so only approved Starlink-terminal/gear could be used.
Too bad the power consumption on the Starlink kit is pretty high from what I read. (due to the heated-dish to melt snow).
Or can you disable ? (or does it regulate this automatically?)

Standard "residential" Starlink dish is 50-75W (excluding the router I think).
So for a 24/24 operation on yearly basis these days with our very high tariffs it would be 150-200 euro's to operate.

hap ac2 has too little cpu power for a proper configuration that the bandwidth really goes over (in my tests only half of the bandwidth with hap ac2).

My 0.02€...
hAP AC2 can EASILY handle 200Mb traffic. Unless you did something specific with VLANs, firewall rules etc ?
Adding a more powerful device will push away your bottleneck a bit but most likely your (config ? ) problem might still be there.

Standard "residential" Starlink dish is 50-75W (excluding the router I think).
So for a 24/24 operation on yearly basis these days with our very high tariffs it would be 150-200 euro's to operate.

Custom solar panel setup with battery to compensate for the costs ?

Good to hear.
It would not have surprised me if they would lock it down so only approved Starlink-terminal/gear could be used.
Too bad the power consumption on the Starlink kit is pretty high from what I read. (due to the heated-dish to melt snow).
Or can you disable ? (or does it regulate this automatically?)

Standard "residential" Starlink dish is 50-75W (excluding the router I think).
So for a 24/24 operation on yearly basis these days with our very high tariffs it would be 150-200 euro's to operate.

the snow mode you can turn off - or even set to automatic, in austria currently costs the operation 85Euro and is often cheaper against other providers just what upload concerns. I have with the hap ac2 no longer over 100mbit speed, but must honestly say I have a very large rulset and also ip blocklists that costs performance, so I got the audience fromm warehouse and runs great.

Both routers have the same CPU.

True but Audience has double RAM and 8 times storage.

Doubt he hit a RAM limit.

Concur with znevna,,,,,,,bogus.

Still...

Assume same config
Same processor
Better performance
Then it has to be ram.

What else ?

Without per core CPU, RAM and storage usage comparison you can't blame the hardware and/or software.
The only thing that's left is the config and stars.

I have my Starlink into an RB5009. All works well. Except IPv6 which was once available but now seems to have disappeared.

I have the round dish - Here in Ireland the weather never really gets too cold anyway but the de-icing can be turned off or just set to automatic.

Power usage is typically around 50W. At least going by my UPS (APC 1500) where load increases 3% with the Starlink PoE connector plugged in.

The only thing that's left is the config and stars.

That's why I put ASSUME same config

2.) Directly plugged in instead of the Starlink Router

hello! searching for this information right now. problem just one - to avoid of using WiFi in round starlink (special purpose). also we have MikroTik router and straight hands)
can anyone provide for us simple scheme of swap plz? on this time I no found any of pinout of voltage lines in starlink power unit (massive black module)

I would recommend here that are unfortunately the blackbox from starlink, still have the round dish where own modem goes / on board rj45 - a wlan brdige with mikrotik and from there to build the real network with mt. br alois

Without per core CPU, RAM and storage usage comparison you can't blame the hardware and/or software.
The only thing that's left is the config and stars.

Fundamentally starlink's speed is pretty variable. So not sure single speed test are the best way to judge performance. Starlink does have a some good debug statistics that can be helpful to know for sure, and that's where I'd start looking to know what speed to expect at best case scenario, at least at the instant your looking at the debug stats. You'll quickly notice how quick they can change...

Even on premium one, with a laptop connected to dish, speed could range from 50-300 down, 10-30 up. Obviously all this has to do where on earth you're using these things (and impairment/interference/etc)... with the equator being kinda the worse place for starlink, and improving as you go north/south from there. So that's a big factor here too in what speed and how stable it be.

Anyway, kinda difficult to believe any particular Mikrotik model used, except at the extreme low end (mAPs etc) or old, couldn't handle the starlink at line rate. And certainly good reason to use a Mikrotik, but doubt the speeds be substantially higher simply by swapping a Mikrotik for the include router. In fact, the starlink router seemingly uses some queuing disciplines that might help with the variable speed, more than a Mikroitik without any queues.

So just saying I'd look at those starlink stats via 192.168.100.1 or in their app before looking at your routing hardware.

I have a hAP AC2 acting as the main router for about a dozen customers (each with access to 300-500Mbps). The primary connection back to my network is a gigabit AF60-LR link to an RB4011 (then gigabit back to the core). The secondary/failover connection is a Wireguard link via the host site's Starlink connection (behind their stock router) to a Wireguard aggregation router (RB4011) in my core.

The hAP AC2 can easily handle the 900Mbps over the AF60-LR link, as well as the encrypted Wireguard connection over 150-200Mbps of Starlink (on good days).

I'm actually impressed at how well this solution is working. AC2/AC3 are pretty capable routers for their size and cost.

Another point here...the Starlink CGNAT does have some upper limit on number port/connections per terminal, it was 5000 a while ago, but could easily be higher/lower. The CGNAT port limit would apply regardless if you use a Mikrotik or the included Wi-Fi router. So I guessing the CGNAT limits, would also limit how much a CPU-centric task like firewall connection tracking have to work on the Mikrotik side.

Now @sirbryan's tunnel approach above would avoid this limit since WireGuard only needs one port, and WG-over-starlink may have some benefits to smooth out the traffic flow as the tunnel be an established connection within the CGNAT in the sky.

the starlink router is garbage and the ethernet port brings only half of the bandwidth, hap ac2 has too little cpu power for a proper configuration that the bandwidth really goes over (in my tests only half of the bandwidth with hap ac2)

Well be precise please. Do you mean:
- Starlink router ethernet speed is half of the ethernet bandwidth: 1 Gbps ?
- hAP ac2 is half of the bandwidth ???? Still about ethernet ? Would be 500Mbps on ethernet !?
- This is well above the rates for a non-business Starlink subscription.

hAP ac2 wifi data rate is only half of the wifi interface rate (866Mbps for 80MHz wide channel dual stream or 400Mbps for 40MHz dual stream) .... this is well known for the MT classic drivers.
Audience can run the wifiwave2 drivers (hAP ac2 so far cannot with it's 16MB storage. hAP ac3 can use wifiwave2)
Wifiwave2 can bring 600Mbps where the classic driver only gets 360-380Mbps, for 866Mbps interface rate. (I get 260Mbps with 400Mbps interface rate)

This is not CPU bound, or RAM bound, but driver difference, with at least the different handling of aggregation (A-MPDU size), what defines the wifi overhead in acquiring transmission time.

Starlink melting snow, works well. No obstructions seen here.

Starlink melting snow, works well. No obstructions seen here.

@bpwl, don't dount it...but since your kinda RF expert here... Have you looked at Starlink's Wi-Fi at all? I don't use starlink at my house, only for work (with wAPs or Chateaus for remote mgmt + LTE backup), so don't have some longer duration experience with their wi-fi. But in my limit tests Starlink's Wi-Fi seems on par with other 802.11ac devices. Now I have noticed that a SA shows higher gain on the include SL wi-fi router, than other APs for sure. Be curious if you had any thoughts on them?

Yes I looked at the Starlink wifi, but I don't want it to be used by that community (40 named users in RADIUS, 20 hAP ac2, 200 client devices). Did not manage to turn it off completely, so it is there with an undistributed password, chewing away some of the wifi airtime with its beacons.

Probably is a normal 802.11ac (don't know if country regulatory domain is set correctly). It will be 3dBm stronger than the MT classic driver signal, and probably the full data rate (600Mbps?).