No problem, thanks for your attention. > model number of your desktop It's an ASUS B550-F motherboard, and the NIC is an Intel I225-V. Though the behavior is the same if I use a USB dongle with a Realtek NIC. > You said 'To reiterate, the desktop is able to get full bandwidth to my ISP in this scenario. ', how did you get this test result？ Just using internet speed tests like Xfinity's. I can get 700+ Mbps (am paying for 600). > If you are using iperf, please enlarge the window size by adding an option '-w 2M'. I have a support ticket open with ScreenBeam and they suggested the same thing yesterday. So yes, it seems the buffer window size makes a big difference. Short of it is, -w 2M does help. But the issue doesn't present itself unless I use the Windows machine. Here's what I wrote to them. Sorry, it's long but I figured I'd include everything. -------------------------------------------- The window buffer size flag -w seems to be making a difference combined with the direction. Setup: desktop (windows 10)> eth> router> eth> ECB > coax> ECB > eth> laptop (Mac OS) -w 2m Laptop <- Desktop: >900 Mbps Laptop -> Desktop: >900 Mbps Laptop -> Desktop: >900 Mbps (initiated from the desktop with -R) Laptop <- Desktop: ~348 Mbps (initiated from the laptop with -R) Setup: desktop -> ECB -> router -> laptop -w 2m Laptop <- Desktop: >900 Mbps Laptop -> Desktop: >900 Mbps Laptop -> Desktop: >900 Mbps (initiated from the desktop with -R) Laptop <- Desktop: ~348 Mbps (initiated from the laptop with -R) Best I can tell, which computer the ECB devices are connected directly to doesn't seem to make a difference. ------------------ I know the -w controls the application buffer size, but it does indirectly impact TCP window size. I've been trying to inspect the TCP window sizes as well with tcpdump on the mac. There's one thing that stood out: Laptop <- Desktop: 32768 Laptop -> Desktop: 32768 Laptop -> Desktop: 32768 (initiated from the desktop with -R) Laptop <- Desktop: 32767 (initiated from the laptop with -R) This could be nothing, just a quirk of iperf. On Windows, dumpcap seems to always report a window size of 65k with a 32 scaling factor to 2097120 bytes, even in the "slow" scenario. ------------------ Default window, not setting -w; the speed is always ~330Mbps. TCP windows: Laptop <- Desktop: 53248 Laptop -> Desktop: 53248 Laptop -> Desktop: 53248 (initiated from the desktop with -R) Laptop <- Desktop: 53238 (initiated from the laptop with -R) In Windows, dumpcap was reporting either ~4MB or 256K windows. ------------------ I repeated the tests replacing the desktop with another mac laptop; this is a work mac so I can only use it as an iperf client, not a server. I was only able to repro slow speeds by reducing the socket buffer window to less than ~768K. The TCP window on both machines reported 65535 bytes. It _seems_ to only be a problem out of the box when the windows machine is involved.

Did another test just now. desktop -> router -> laptop was gigabit; desktop -> router -> moca -> laptop was ~330 Mbps. I purchased a competitor's product (ScreenBeam ECB7250) and it's exhibiting the same behavior. So the problem could somehow be a software issue on my end, or both products are using the same network controller and there's a bug.

@goCoax Thanks for your response. No, that's an incorrect understanding; sorry, I didn't make it clear in the original post, there's a lot of info there. The router does not have built-in MoCA, it's a Firewalla Gold and all the devices are connected to each other via cat-5a. This is what it looks like by default: desktop [cat5a]-> switch [cat5a]-> moca - coax - moca [cat5a]-> router [cat5a]-> switch [cat5a]-> WAP To reiterate, the desktop is able to get full bandwidth to my ISP in this scenario. The MA2500D also _has_ been connected directly to my desktop, or an ethernet dongle, or my laptop. I've done a bunch of permutations to try to narrow it down. For example: desktop -> router -> moca - coax - moca -> laptop I've also tried this with an entirely separate router (a Netgear R7000P) just to rule out my primary router. If I remove the MoCA, or connect a device other than the desktop, I can push gigabit through. Fortunately I already did have static IPs set up for the MoCA device. Here are the link rates, they're the same for both devices.

Just following up on this question again, thanks.

Thanks. That's unfortunate. I looked up C vs D and mounting wasn't differentiated anywhere, I can't swap the units out now. Can you point me towards the online manuals for the three products please?