Pocketables

Trendnet TEG-10GECTX PCIe 10GbaseT Ethernet Adapter Review

Following up on our coverage of the TUC-ET5G it’s only fitting that we crank things up to 11(well 10) and take a look at the 10GbaseT TEG-10GECTX targeting a less mobile segment of small business servers, high end desktops and various NAS units. One thing we did have to do differently for testing the TEG-10GECTX was build a desktop PC to host it

Unboxing and first impressions

Opening the packaging for the TEG-10GECTX was fairly straight forward. The art on the box matched the included card with all the critical details on the package. Overall everything was easy to remove and seemed well protected with no real surprises. The adapter is a surprisingly small for 10GbaseT especially when compared to older 10gig network adapters using SFP+ modules. The included heat sink is rather small but seems sufficient and the PCB is a fairly standard green which is generally a non issue but may clash with some color coordinated setups.

What is NbaseT

NbaseT is rather recent evolution in the long history of Ethernet over twisted pair. While 10GbaseT was standardized back in 2006 gigabit installs using commodity category 5e and 6 cabling while 6A was required for transmission of 10Gig communications. 2.5GbaseT and 5GbaseT commonly refereed to together as NbaseT were approved as standards in 2016 allowing for a communications above gigabit speeds without the strict requirements that 10GbaseT required. Most new 10GbaseT equipment(such as the ipolex transceiver used in this review) allows for auto negotiation of link speed based on line capability as well as the capabilities of the connected device. NbaseT has been adopted much more quickly thanks to an increased demand in bandwith for devices such as wireless access points, the ability to re-use existing wiring and the addition of PoE which was missing from the original 10GbaseT standards.(if you want a more in depth article, or video about the full implications of NbaseT let us know in the comments below)

Building a PC for testing

Testing this card actually required building a dedicated test rig. For this purpose an older 3xxx series Thinkcentre M82. We’ll be repurpousing this system afterwards as a gaming system and will have coverage of that afterwards as well. For more details on that build please watch the build video.

Performance testing

Performance testing on the TEG-10GECTX was done using our refurbished M82 with the card mounted in the pcie 16x slot as well as an Ipolex transceiver that was purchased after reading a review on ServeTheHome. In addition my core network switch a Mikrotik CRS328-24p-4s+rm and home server were utilized. The home server is a custom built dual Xeon E5-2470 V2 system attached via a pair of 10gig links ensuring more than ample bandwidth and processing was available. As a validation I did check iperf between my desktop and the server and saw speeds in excess of 9gbit so there isn’t an issue with any of our testing equipment.

Overall performance was spectacular and managed to reach 9.69 Gbit/sec over a length of category 7 cable using iperf. Jumbo frames were enabled for this testing(and should be on links over 1Gbit) but other than that no settings were tweaked. I did test manually at 2.5 and 5GbaseT as well manually setting the connection speed. At 2.5GbaseT the card ran at 2.4Gbit/sec and at 5GbaseT we saw performance reaching 4.91Gbit/sec without issue.

Closing Thoughts

On the performance side there’s no question that the TEG-10GECTX met with it’s specifications leaving only pricing and use case as the remaining considerations. On the price side of things the TEG-10GECTX is in good company at 119.99 it’s a reasonable alternative to used enterprise gear and is in the same pricing bracket as other cards although there are a few that offer better “per port” pricing that is fairly common for single port cards. As for who should be considering this, the same prosumers interested in it’s mobile counterpart will find it a welcome addition to nas units(that support it) and fixed workstations or custom nas units. More standard home users perhaps just trying to replace a failed adapter on a mainboard will generally find cheaper gigabit options more than sufficient however.

Pocketables does not accept targeted advertising, phony guest posts, paid reviews, etc. Help us keep this way with support on Patreon!
Become a patron at Patreon!

Daniel Smith

Daniel Smith is a full time IT administrator at a medium sized private business former FRC coach and technology enthusiast.

Avatar of Daniel Smith

12 thoughts on “Trendnet TEG-10GECTX PCIe 10GbaseT Ethernet Adapter Review

  • Chipset? Is it a high-quality chipset usable to server-use? What about power-consumption? These > 1Gbit NICs use a lot of power.

    Reply
      • It’s all about the chipset. It doesn’t really matter who the manufacturer of the NIC is, they only add a few resistors on a PCB. And the specified power-consumption doesn’t always stand IRL, that’s why it would be nice to have a check on that. If it runs 6-10 watt 24×7 that’s something to consider, perhaps even more, who knows.

        Reply
        • Officially Trendnet can’t comment, unofficially it’s an AQ107 so that 6-10w range is correct, the big determining factor for a 10gbaseT adapter’s power consumption is link speed and cable length, it’ll run lower power on shorter/slower links and does have all the power management features an aq107 would normally have. If you’re really after low power you should consider fiberoptic, they’re traditionally a few watts lower than a 10GbaseT adapter per port and you can purchase a dual port connectx-4 with sfp28(so 25gig capable) for about 20 dollars more on ebay.

          Reply
          • 6.1 watts is the max on this card. 53kWh max per year. $6.99 to run a year at 13.19 cents per kWh. It shouldn’t be pulling the max constantly – be interesting if you could get an idle power and a network power draw reading.

            What I’m seeing on fiber is it’s half the wattage on average for the cards so a max savings of $3.50 a year under full load.

          • oh it’s small on the client side but it adds up on the switch side 3w*24 or 48 is a lot of power and that’s with good adapters, there is the advantage though of the fiber adapters typically being dual port so there’s a density savings there as well

          • Yeah, but that wattage is based on usage is it not? Cost to transmit terabyte. When idle these things should drop down into the sub watt category, especially the switches.

            Price difference between lower end 24 port fiber and 24 port copper 10gb is phenomenal. ($2000 v ~$300) unless I’m missing less expensive fiber options. I can’t see financially it making sense even with the dual port.

            Looking at max wattage on a switch (which all of them seem to max out at about 125 watts) that’s $144 a year in electricity meaning you’d have to expect the switch to last more than 11.8 years to potentially reach savings if I did the math right and found the least expensive 24 port fiber.

            Don’t think anyone other than me runs 10+ year high end switches

          • That’s it’s own special rabbit hole really, old 10gbaset switches were upwards of 12w per port, new ones are usually 1.5-4w with some variance on link length(there’s a minimum power to simply have an established link) as well as usage(most things have idle power states these days as well) I’m not saying you should pick sfp+ over 10gbaset at home just to save power but if you’re sensitive to every watt a dual port sfp+ adapter is going to be better than 10gbaset(especially if you have to have two)

            Who’s got a 24 port sfp+ switch at 300? the cheapest i’m aware of is the crs326-24s+2q+rm which is ~407(granted it has 40gb ports as well) https://amzn.to/3kRmjFo

        • Don’t trust the word of a spec sheet! Especially of a brand of cheap chipset.
          A bit off-topic I know, but my Kingston KC600 SSD use almost 3 x more power in idle and in read mode than the spec sheet tells, reddit

          Reply
          • I unfortunately don’t have a way to meter a pcie device directly at this time, that is part of why my answer was 6-10w based on other AQ107 devices i’ve observed(such as the OWC 10gbaseT adapter which coincidentally must feed both the AQ107 and intel JHL6540 inside of a 15w limit to be bus powered)

            As far as that aside regarding the KC600 that really does look like something is keeping the drive from going into an idle state, knowing what board/cpu was used there would be interesting(it looks like a server board, they tend to have different behavior regarding idle) I may send kingston a note to see if we could get a KC600 to look at in the next few weeks.

          • I agree measure power-consumption on a PCI socket is whole other story, I would say the only practical way is to measure the whole computer, before and after activating the device, but that should also give you an okay power indication. I just use a rather cheap DIN power-meter for this purpose, forums.servethehome.com, you cannot use a “normal” AC power-meter as e.g. my Fluke, as you need to know the power-factor, especially when you measure behind a switch-mode power-supply.
            In regards to the Kingston SSD.
            1: This drive is connected to a standard SATA AHCI controller with no Intel bios raid activated.
            2: I haven’t heard that a server-chipset should use any special SATA protocal or anyting else that would keep the drive in non-idle state
            3: I tested other SATA drives and they go into idle just fine (and don’t wake up until I sent a command to the drive)
            4: I’m using Linux (Debian) to be sure nothing is sent to the disk, unless I want too.

            But it sure would be interesting to get a second confirmation-test :-)

          • In the case of a higher power card(a gpu) a full system power measurement would be acceptable however a pcie device like this is low enough power that other factors(disk and cpu activity triggered by using the 10gig link) could easily skew the measurements, you’d then have to measure and subtract out changes in the cpu and drive(s) as well and at that point if you’re measuring 6-10watts you aren’t likely to have precise enough measurements to really see what’s going on.(I suppose you could put it on a pcie riser and tap into the cables that way but that’s beyond what I have in my studio… this week)

            Going back to the Kingston I’ll make a post when I hear back from them, as far as the server question I’ve seen SAS ports do odd things with SATA drives(SAS is a superset of sata) especially older hardware but if you’re using a standard SATA port you should be fine. Have you tested it under windows to see if it behaves differently there?

Leave a Reply

Your email address will not be published. Required fields are marked *