MOCA 2.5 Help Needed - Why Are my Speeds Slow in 1 Direction?

[kelllogg9]

I have the ScreenBeam (ActionTec) ECB7500 (moca 2.5)

I have two PCs (both Win11 Pro). Both have the same specs. When i transfer from PC1 to PC2 across the 2.5Gbps network i get 285MB/s. However, when transferring the same file (after clearing the system file cache via RAMMap) from PC2 to PC1 i only get 112MB/s. iperf3 tests are the same (ie. one direction is full, the other direction is half). My guess is the "bonded" feature is working only in one direction since you can see its roughly half the speed. Can't figure out why though. Maybe one of the moca devices is defective??

It used to work fine (speed was 285MB/s in both directions). Nothing has changed on the network.

I hooked both Moca devices directly to each other using a tiny coax adapter so its only inch apart from each other -- eliminating any concerns of distance or a corrupt cable. The same problem persists. I have moca filters but obv in this test they were not needed.

I swapped both devices and the speed and direction loss remains the same and in the same directions, despite using the other moca for a pc. You would think the direction would reverse. Wrong!

I connected my ethernet cable directly to both pcs and the speed returns to 280MB/s in both directions. So its not a Windows config issue.

ScreenBeam sent me their latest (soon to be released)firmware (v. 1.18.7.5) to upgrade to recently to see if it would resolve the issue. It did not. My org version was 1.18.2.0.

Their tech thinks i need to increase my TCP Window on both PCs to resolve the issue. I told him that using my ethernet cable with no moca devices shows the tcp window is not the issue cuz my speeds return to 285 MB/s in both directions. His response:

"Because of the nature of the MoCA connection, it has higher latency than an Ethernet connection. Even if the maximum UDP throughput is the same as Ethernet, high-speed TCP throughput requires more buffer (TCP window size) to reach the same throughput as Ethernet."

If that were true then the speed in both directions would be 112MB/s, which it is not. So latency isnt the issue. But nonetheless, i set on both pcs:

netsh int tcp set global autotuninglevel=experimental
Registry: HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\Tcpip\Parameters!Tcp1323Opts (DWORD) set to "3"

Unsurprisingly, it made no difference.

Any ideas? I'm all out.

 

[degrub]

get into the diagnostic pages for the moca modems to check power levels and sync rates.
The only other thing to look at is the power wall warts. Swap them only and see if the problem moves with the power wart.

 

[45263tec]

Registered here just to post this. Having the EXACT same issue in my case with GoCoax adapters. 2 windows machines with Realtek 8125b cards. All POE filters in place (outside and on cable modem). Web UI shows inter-node link is at full MOCA 2.5 speed. Both machines saturate the 2.5G connection over ethernet.

Speed in one direction about 215MB/s. Speed in the other direction is capped at 50MB/s

@kelllogg9 did you ever find a solution?

 

[tgl]

I'm seeing something similar with a pair of ScreenBeam ECB7250's, but worse: one direction achieves full-rate 2.35Gbps per iperf3, but the other direction can only manage a couple hundred Mbps. I've not yet tried the swap-the-adapters idea. Anyone have a clue what's causing this?

 

[kelllogg9]

I'm seeing something similar with a pair of ScreenBeam ECB7250's, but worse: one direction achieves full-rate 2.35Gbps per iperf3, but the other direction can only manage a couple hundred Mbps. I've not yet tried the swap-the-adapters idea. Anyone have a clue what's causing this?

Hi there. Yea i recall having this very similar problem with my ECB7250s. Another person i spoke to recently in DM on this site has/had the same problem as yours. So you most definitely are not alone. I first discovered it over a year ago. Frustrated the hell out of me. At the time i spoke with tech support each day for a few week straight but they were pretty useless. So i ended up going back to goofing around with Windows settings and doing more experiments. At the time i wrote my problem i think i was using Windows 7 (??). I am now using Windows 11. The problem STILL persists at times. Its random it seems. And its always in one direction that has the issue. The other is always full speed 2.x Gbps.

However, i noticed something interesting. I notice that my speeds WILL spike back up to max speeds (280MB/s) both ways at times and other times only one way will always be 280+ while the other way will drop to a max of 180MB/s. But no matter what, at least one direction will be 280+ MB/s.

I get a sense, at least for my case, it is Windows that is throttling my moca max speeds because i notice if say i am streaming a live video say from Twitch and then i transfer say a 10GB movie file in either direction on my LAN the speeds will be 280MB/s both ways (and that is me clearing the file cache to keep it fair -- using RAMMap app). The same with iperf tests. But if i dont stream from the internet and wait a while then the speed may (and usually does) drop back down to 180MB/s in the one direction. I can't be fully sure if this is a consistent pattern. For example, just now i am on Twitch watching an ethot being an ethot and as i watch i am now testing a transfer on the lan and the speeds are 280+ in both directions.

For the fun of it, give this simple test a try on your side, assuming ur using Windows, and see if your max speeds return. And to be clear, my nic cards on both machines are set to their defaults and they are the same 2.5g branded cards. So it isnt an incorrect setting on my nics. I shared all my settings with their tech support but he saw nothing special that was worth changing. And both pcs are windows 11.

I also swapped both moca devices locations. Made no difference. Interestingly, you would think the slow speed in the one direction would reverse. Nope. It remained slow in the same direction. Also hinting it could be a Windows thing on my specific PC.

I also did a format/reinstall of Windows on the PC. Problem still persisted. I also tried a laptop Windows 11. Same problem.

The only thing i haven't done was get another ECB7250 to swap out to test it, which i was planning to do but lost interest

The tech support did give me an upgraded firmware though. So my firmware is a tad newer. I think because one of my moca devices had a newer version than the other and he was confused what my newer version was since he was never informed of a newer version. Took him days to track it down on his end and then send me the newer firmware so i could upgrade my other moca device. Obviously this did not solve my slow speed in 1 direction problem but at least my firmwares were finally the latest (at the time).

I also changed ethernet cables and coax cables and even used a 1 inch coax coupler to really reduce the length. And i also attached both moca devices directly to each other (no coax cables.)...Didn't make a difference.

Regards

 

[thewifiatemyhomework]

same issue…

 

[tgl]

FWIW, I've done more testing since my last post. Some notes:

• It's not a Windows issue for me: there is no Windows in this house.
• Swapping the two adapters doesn't change anything: the slow direction remains the same.
• Replacing the ScreenBeam adapters with ASUS MA-25 adapters doesn't change anything.
• I don't really notice any problem with 1Gbps equipment connected to the MoCA adapters. It takes 2.5G ports on both sides before it becomes apparent that the MoCA gear isn't performing at advertised speed.
• I do see that tests involving parallel data streams perform better than single-stream tests in the slow direction. It's easy to reproduce this with iperf3 by using "-P5" or so. In my case I could get close to the theoretical 2.3G-ish limit with -P6.
• What does change the results is swapping out the directly-connected equipment. What I've found is that a Ugreen 2.5G USB-to-Ethernet adapter is slow to receive when connected directly to the MoCA adapter, and so is a Zyxel XGS1250-12 switch. A Cisco CBS350-8MGP-2X switch is fine though, and I've also had successful full-rate-single-stream results with a late-model Intel NUC machine with a 2.5G port using an i226-V ethernet controller. I don't presently own any other gear with better-than-1G ethernet ports, so beyond this deponent sayeth not.

The conclusion seems to be that there's an ethernet-level performance incompatibility with certain equipment. I suspect that getting similar results from ScreenBeam and ASUS adapters means that they're using the same ethernet interface chips, but I'm not planning to open them up to try to prove that.

 

[thewifiatemyhomework]

I noticed my TP link switches would not always negotiate 2500 and needed to be replugged multiple times. My neighbor used trendnet and it worked fine…

 

[brimur]

Just wanted to add my +1 to this issue. I posted about it recently on Reddit but nobody had any solutions. I contacted GoCoax who are the makers of mine and they said the would comment on that thread but never did.

 

[tgl]

The conclusion seems to be that there's an ethernet-level performance incompatibility with certain equipment. I suspect that getting similar results from ScreenBeam and ASUS adapters means that they're using the same ethernet interface chips, but I'm not planning to open them up to try to prove that.

I can now positively confirm that. I finally sprung for another Cisco switch, and with those switches on both ends of the MoCA link I get nearly-clean 2.35Gbps iperf3 results in both directions, either single-stream or multi-stream. Performance of the ScreenBeam and ASUS MoCA adapters remains just about indistinguishable.

There is one nagging bit of annoyance: in just one of the two directions, I see occasional short bursts of retries reported by iperf3, with ensuing momentary drops in transfer rate. I suppose this is somehow down to the remaining asymmetry in my test setup, which in the interests of transparency looks like:

Mac M2 mini <-> Cisco switch <-> ethernet run <-> Cisco switch <-> MoCA link <-> Cisco switch <-> Intel Gen 13 NUC

The mini has a 10G ethernet port, the NUC only 2.5G. The weird thing though is that the retry glitches happen when sending data from NUC to M2, which is the direction you'd think would be impervious to bufferbloat or suchlike.

So I'm still a little baffled, but I'm content with my current results; I don't expect saturation of the MoCA link to ever happen under non-test workloads. "Buy another Cisco switch" was a darn expensive solution, but I was wanting to do it anyway just to have a more consistent network setup, and so that I could have the Zyxel switch on the shelf as emergency backup.

 

[degrub]

Does Apple's nic support 2.5 GHz correctly ?
What if you replaced the mini with another NUC for testing ?

 

[tgl]

To be clear, the mini's NIC is 10Gbps (and plugged into a 10G port on the first switch). I don't have another 2.5G machine at the moment, other than a Ugreen USB/ethernet adapter that I don't trust at all for this purpose and some APs that lack enough CPU oomph to sustain 2.5G iperf3 sessions, so it's difficult to try other combinations. I can say though that I get the expected 2.35Gbps-and-no-funny-business behavior if the NUC is plugged into either the first or second switch in the chain --- it's definitely the MoCA link that is the weak spot.

 

[tek872956]

Thanks for the post, I have similar issues with my ScreenBeam ECB7250 MoCA 2.5 Network Adapters. Trying to setup the wired backhaul between ASUS GT-AX6000 routers using 2.5GB ports. If one of them connects to the ASUS router directly and another through Real HD SW5-25G switch to another ASUS router and to Intel NUC 12 Enthusiast, I only get ~600Mbps with iperf3, but if I add 1Gbps router between ScreenBeam ECB7250 the router it goes up to ~950Mbps. And the same setup with replacing ScreenBeam ECB7250 by an ethernet wire gives me 2.37Gbps (unfortunately I can only use the ethernet cable for testing, due to the location of both ASUS routers). As a client I have a PC with a built-in Realtek RTL8125B(G) adapter and latest drivers from Realtek. I've tried UGreen USB-C 2.5G Adapter with the same result. I was about to return ScreenBeam ECB7250 as it looked like they are defective (they sort of are as they ethernet port is peace of garbage) but hopefully there is more cost-effective solution vs getting 2 Cisco CBS350-8MGP-2X switches. So far using 1Gbps switch is a good option as this is better than 600Mbps or wireless backhaul. If I knew this I would probably get the ECB6250 instead.

 

[degrub]

I can now positively confirm that. I finally sprung for another Cisco switch, and with those switches on both ends of the MoCA link I get nearly-clean 2.35Gbps iperf3 results in both directions, either single-stream or multi-stream. Performance of the ScreenBeam and ASUS MoCA adapters remains just about indistinguishable.

There is one nagging bit of annoyance: in just one of the two directions, I see occasional short bursts of retries reported by iperf3, with ensuing momentary drops in transfer rate. I suppose this is somehow down to the remaining asymmetry in my test setup, which in the interests of transparency looks like:

Mac M2 mini <-> Cisco switch <-> ethernet run <-> Cisco switch <-> MoCA link <-> Cisco switch <-> Intel Gen 13 NUC

The mini has a 10G ethernet port, the NUC only 2.5G. The weird thing though is that the retry glitches happen when sending data from NUC to M2, which is the direction you'd think would be impervious to bufferbloat or suchlike.

So I'm still a little baffled, but I'm content with my current results; I don't expect saturation of the MoCA link to ever happen under non-test workloads. "Buy another Cisco switch" was a darn expensive solution, but I was wanting to do it anyway just to have a more consistent network setup, and so that I could have the Zyxel switch on the shelf as emergency backup.

What is the chipset for ethernet on the NUC ?

 

[tgl]

What is the chipset for ethernet on the NUC ?

I've not opened it up to confirm, but the spec sheet says i226-V.

 

[tek872956]

Mine has Intel I225-LM, I haven't had much time to test different configuration yet, I will post more updates as I progress.

P.S. I finally received D-Link DMS-108 8 x 2.5G switch I can play with as I needed more ports and second 2.5G switch, although it's not as advanced as Cisco, unlikely it will help in this case.

 

[degrub]

The I22x based nics seem to have some sort of compatibility issues with some other hardware - driver or chipset - for 2.5Gbit links from what i have read.

 

[airbusacft]

New to SNBforums... As anyone been able to solve this massive mystery?
I'm also in a townhouse (rental), and running a CAT cable is impossible; otherwise, I would have done that. I recently switched from Comcast to Google Fiber (5G) service, hoping to get 2.35 via MoCA to my office on the other side of the townhouse. My MacBook Pro w/ M2 Pro is plugged into a docking station with a 2.5 GbE, but speedtest.net and Google Fiber speed test website says download of 500-550 Mbps and upload of 2350 Mbps. I bought a pair of ScreenBeam ECB7250, and they have the most current firmware displayed on their support page. Xfinity main cable into a Commscope power amp with MoCA filter between mainline and amp. Comcast says an amp was needed for the signal; the living room had a red commscope (SV-2G / 5-1002), one end into the STB and the other into the living XB8 modem. I'm unable to get the model number of the amp, but could see a sticker that said MoCA compatible.

I had subscribed to their Xfinity 1G service (approx 975 DL / 100 UPL). My Macbook showed approx 875-900 DL / 75 UL when I tried using only 1 ECB7250 in the office, as their XB8 acted as the 1st MoCA device. Then, when I switched to GFiber, I noticed a massive difference in DL/UPL speeds and started troubleshooting. I've tried everything I can think of. I even thought about trying different MoCA adapters, but after reading the problem is the same as ASUS and Gocoax, I won't do that. I knew my Macbook would see and operate at higher speeds when I connected a Sonnetech solo 10G between a fiber router and Macbook and saw the full and even slightly higher 5G speeds. It seemed to work fine when XB8 acted as the 1st MoCA device but lost almost all the DL when the 2nd ECB7250 was added to the mix.

I've tried the following:
Factory reset of ECB7250s, and swapped locations of the ECB7250s
Swap out coax cables and different CAT 6/6A cables
Tried Sonnettech USB-C 2.5 GbE (same result)
Different docking stations both rated at 2.5 GbE
Tied #1 ECB7250 directly into coax outlet
Disconnected the living room and office coax from the amp and tied them together using a barrel connector (same result)
Direct from fiber ONT to ECB7250

 

[degrub]

Are there diagnostics pages you can see if you log into the ECB6250 ?
Check the link rates - should be a table with all the moca devices listed and the link rate in each direction.

 

[airbusacft]

Here are the link rates from the ECB7250 in my office. Would it help if I put either 2.5G or 10G PoE switch or PoE injector in front of the switch that is connected to the GFIBER router and between the ECB docking station in my office? There are 2 different PHY Rates. 1 was ECB7250 with no splitter, except for commscope power amp that says MoCA compatible. The 2nd PHY Rate is what it looked like going through the 5-1002 commscope 2-way splitter.