EDIT2: I did two more tests:
1) I used the ethernet dongle on the desktop, with moca between the desktop and the router. Best I can tell, there's no difference between using it vs the onboard intel NIC.
2) laptop [ethernet]-> switch -> moca -> router -> desktop
~380 Mbps across the LAN, yet >700 for an internet speed test from the desktop. Removing moca saturates the gigabit network.
Aside: I'm able to reproduce the network drop over moca. Leaving iperf running, it takes a variable amount of time for the connectivity to drop (e.g., a few seconds up to about 20 seconds so far) and/or return (e.g. 40 seconds to a couple of minutes), but it always comes and goes. If I bypass moca, it's completely stable.
EDIT: It seems this is an issue between the MA2500D and the desktop (Intel I225-V ethernet controller). If I do nothing but connect my laptop (2019 MacBook Pro with a USB-C ethernet dongle with a Realtek controller) to "switch1", I can push 350+ to the same device. I'm surprised this is a problem even with the gigabit switch between moca and the desktop, but clearly the MA2500D does not like my desktop for LAN traffic.
I also neglected to mention in the original post that I've been seeing random internet drops on any device "after" moca, like the desktop. I can even sometimes trigger it while running iperf.
This seems similar to https://www.gocoax.com/forum/everything-for-moca/ma2500d-network-cable-unplugged-drops
Though "Energy Efficient Ethernet" is disabled on my ethernet adapter, so I can't apply a fix from that post.
My desktop is connected to the internet like:
desktop -> switch1 -> moca -> coax -> moca -> router -> modem I'm able to perform internet speed tests at 700+ Mbps. Both the moca devices are the MA2500D. I have a streaming box (shield tv) downstream from the router:
router -> switch2 -> WAP -> shield I've been running iperf on the shield and my PC, doing speed tests. If moca is in the picture, the throughput is capped at ~340 Mbps when it's ethernet all the way.
desktop -> switch1 -> moca -> router -> switch -> shield If I connect switch1 straight to the router (not via moca), I can saturate 930+ Mbps. When the shield is connected to the WAP, I get ~160 Mbps (with moca) and ~220 (without moca), so it's always causing some kind of limitation. Any thoughts? Wondering if I'm just running into some weird edge case of LAN traffic over moca, or if the hardware is doing the wrong thing.