A MoCa mystery of some servers having slower speeds?

[392DCGC]

I have an access point connected via goCoax MA2500D on ATT Fiber (2.5 Gbps into the home). It is a direct coax connection between adapters, and the rest of the coax in the home isn't used/connected to anything. The issue I just stumbled across is my adapter seems to be slowing downloads from certain servers... for example, some speedtest/fast.com servers will download around 300~ Mbps and upload fine, but other servers will pull over 1 Gbps no problem. You'd think it's just a problem with those particular servers, right? Well no, because if I run a speed test on those exact same problem servers from my router instead of the AP, those servers pull full speed through the router where the MoCa was experiencing slower downloads.

OpenSpeedTest local tests pull just fine as well over MoCa... so again, MoCa is somehow a factor, but only certain servers are affected? I also tried physically swapping the AP to a different piece of hardware, and the situation was the exact same.

Example... here is a problem server on the MoCa AP:

Runs fine on the router:

 

[392DCGC]

OpenSpeedTest between MoCa AP and local PC (WiFi 7 device):

 

[degrub]

If you want to test the moca throughput, you should do it between two PCs at each end of the moca connections. You can use iperf2 under windows . Don't use iperf 3 on windows as there are issues. You have to use the -p switch to saturate the moca link. About 10 parallel streams is good.. The two PCs need to be able to keep up with the test as well. Search on this subforum for an example cmd line for the test.

You cannot use a wifi device to test nor a router to test. result is not meaningful for evaluating what the moca modems and the PCs are doing. Going across the internet to another server adds too many variables as the path is not fixed and the server loads vary.

The reason to isolate the moca segment for testing is that if there is any traffic in the other direction across the coax during the period of the test ( from the pc or other users/router/ap, internet, etc) that will influence the result. MOCA is like wifi - it is half duplex. So on average with a MOCA2.5 direct connect setup, you get about 940 Mbit/s both directions. If the flow in in one direction, you can get up to the full available bandwidth of roughly 1.9 Mbits/s.

 

[392DCGC]

If you want to test the moca throughput, you should do it between two PCs at each end of the moca connections. You can use iperf2 under windows . Don't use iperf 3 on windows as there are issues. You have to use the -p switch to saturate the moca link. About 10 parallel streams is good.. The two PCs need to be able to keep up with the test as well. Search on this subforum for an example cmd line for the test.

You cannot use a wifi device to test nor a router to test. result is not meaningful for evaluating what the moca modems and the PCs are doing. Going across the internet to another server adds too many variables as the path is not fixed and the server loads vary.

The reason to isolate the moca segment for testing is that if there is any traffic in the other direction across the coax during the period of the test ( from the pc or other users/router/ap, internet, etc) that will influence the result. MOCA is like wifi - it is half duplex. So on average with a MOCA2.5 direct connect setup, you get about 940 Mbit/s both directions. If the flow in in one direction, you can get up to the full available bandwidth of roughly 1.9 Mbits/s.

I understand the desire to cut other potential variables, but what I'm experiencing doesn't jive with that assessment, so I don't see the need to do that. It is 100% repeatable, and the testing I've done is not strictly over WiFi - even a hardwired device to the AP exhibits the same repeatable low download bandwidth from specific servers. One of the screenshots I posted was a WiFi 7 local speed test from the AP repeatably returning 1.7 Gbps throughput (proving that WiFi is definitely NOT a limitation anywhere near what I'm experiencing from certain speed test servers)... yet as soon as I target the speed servers in question, it is 300~ Mbps download from the AP whether over WiFi or a hardwired. Testing the exact same servers from the router (WiFi or hardwired), this 300~ Mbps download limitation entirely goes away. And again, this only affects certain servers through the MoCa AP - other servers will push full bandwidth no problem through the MoCa AP.

I have seen other posts where people have commented on having similar MoCa throughput issues, but none of them mentioned the degradation only occuring from specific servers.

Put in as simple of terms as possible:

--MoCa AP--
Server A: 300~ Mbps download, normal upload
Server B: normal download & upload
Local: max throughput

--Router--
Server A: normal download & upload
Server B: normal download & upload
Local: max throughput