Best way to connect 10GBase-T to DEC740

Started by Tjh3, June 19, 2024, 05:41:04 AM

Previous topic - Next topic
June 19, 2024, 05:41:04 AM Last Edit: June 19, 2024, 04:19:14 PM by Tjh3
Tl;DR is, my ISP's "handoff" involves an XGSPON box that only has an "RJ45" jack that is wired for 10GbE. Idk what is the best way to connect. So far my options are:


  • 10GBASE-T SFP+ module: I tried the 10gtek one with really bad results (unreliable speeds in one direction - it would go down to double digit Mbps). I'm also worried about heat and power usage as the DEC740 is an embedded appliance and is fanless.

  • Media converter

  • Unmanaged switch: qnap qsw-308-1c includes a 10gb combo port.
  • Managed switch: by far the most expensive option
I like the SFP+ module option, as an extra switch and/or converter is one more point of possible failure, but I'm worried about the heat and power usage on an appliance device like DEC740.




I would like, if possible to get reliable speeds with as low latency as possible. The RJ45 cable I'll be using is cat6 of about 13m.

If you were able to either eat the cost or return the 10Gtek module, you could try an FS.com module, I have one of these running in a Netgear AV Line m4250 going to our Tricaster which only had a twisted pair interface. So far so good but I haven't pushed more than about a gigabit worth of connections through it (3 cameras and a few other devices).

Alternate, have you contacted 10Gtek? They might have a different firmware that you need. https://www.10gtek.com/contact.html but looking over the site, this might not work out.

FS.com will get back to you and maybe they would be a better choice. I've talked to sales many times and they were really nice and helpful, can't remember if I ever talked to technical support. Something I need to remember going forward.

If you need a 10gbps switch, I'm happy with my Mikrotik CRS309-1G-8S+, I have 2 of these running as "top of rack" in my XCP-NG system (production and lab), then you might be able to use a DAC cable between switch and firewall, and the 10Gtek module between switch and XGSPON.

The Cat6 should be fine for that distance, but I'm also seeing cat6a for around $35usd in a 15 meter length.

June 21, 2024, 05:25:56 AM #2 Last Edit: June 21, 2024, 05:29:29 AM by Tjh3
i was able to return the 10gtek module and get one from fs.com. Thank you. It is working a lot better. CRS309 is actually the one i'm using as my main switch (it then connects downstream to a netgear switch via sfp+ for Wifi and IOT stuff).

I do have a new problem though: When connecting the fs.com SFP-10G-T and cable from my ISP box directly to my computer i am fully able to saturate the 8Gbps speeds from my ISP via a speedtest. I am also able to get near 10Gbps, and about 6Gbps to DEC740 using iperf3. however I cannot for the life of me go above 6Gbps or so on speedtest through the DEC740 to the internet. Just to re-iterate:

1. Internet connected directly to Desktop using RJ45 cable and SFP-10G-T from fs.com = 8Gbps
2. iperf3 test to router from desktop using DAC cable = 9.5Gbps
3. Internet connected to DEC740 which in turn is connected to router using DAC cable (same as #2) = 6Gbps

why is #3 so low? The oddest thing is #2 is high as i'd expect, meaning the actual connection to the router is perfectly fine. The specs said that DEC740 is able to do upto 8.5Gbps,and aside from 4 NAT rules and Adguard home, i'm not running anything else. And CPU usage never goes above 50% either way.

Unless the firewall portion doesn't get hit when I do the iperf test to the dec740 itself and that device is actually not capable of 8Gbps speeds with my settings? I tried just connecting a cat7 cable, and although i was able to get a bit higher (actually hitting 6 and 6.1), still not able to reach the 8Gbps speeds i saw when directly connected.

Turn off Adguard and see what happens, I wouldn't expect it to drag the speed down much, but maybe.

I did try that. As you suspected, it didn't really make much of a difference. The odd thing is, getting **to** the firewall i can hit line speeds. And i watched all the cores during the speedtest and there were only a few brief peaks of no more than 50% each CPU.