What the LAN and WiFi rows on our spec tables mean, and when paying for a faster one changes anything
Every review on this site prints a LAN row and a WiFi row, and neither one is worth much on its own. A wired port only runs at the speed both ends agree on, so a 5G port on a gigabit switch is a gigabit port. A second Ethernet port is a second network interface, not a faster one. And 6 GHz and multi-link Wi-Fi are properties of the link, so they need a router that speaks them too. The board is one half of every one of these numbers.
The LAN row names the fastest speed the board's wired Ethernet port is rated for. Four values cover almost everything on the market, and each one is a separate IEEE standard rather than a vendor feature:
| Spec row | Standard | Cabling |
|---|---|---|
| 1G | 1000BASE-T, IEEE 802.3ab, approved June 1999 | Category 5 or better, 100 m recommended maximum |
| 2.5G | 2.5GBASE-T, IEEE 802.3bz, approved September 2016 | Cat 5e or better, up to 100 m |
| 5G | 5GBASE-T, IEEE 802.3bz, approved September 2016 | Cat 5e or better, up to 100 m |
| 10G | 10GBASE-T, IEEE 802.3an, approved June 2006 | Cat 6A for the full distance; Cat 5e or 6 may reach 55 m |
The 2.5G and 5G rows come from the same task force and the same 2016 standard. The IEEE P802.3bz group closed "with the approval of IEEE Std 802.3bz-2016 by the IEEE-SA Standards Board", and the point of it was speed on cable that was already in the wall: Wikipedia's summary is that 2.5GBASE-T and 5GBASE-T "can be deployed at a cable length of up to 100 m on Cat 5e or better cables." That is why 2.5G spread so fast across mid-range boards while 10G did not — 10GBASE-T is a decade older as a standard but wants Category 6A to go the distance.
Across the 80 boards we currently review, the LAN row reads 2.5G on 58 of them and 1G on 8. The remaining 14 print either a faster port or a second one. So on this catalogue 2.5G is the ordinary case, not the upgrade, and the interesting question about a 5G or 10G board is almost never about the board.
Ethernet negotiates. Two connected devices "first share their capabilities regarding these parameters and choose the highest-performance transmission mode they both support." A 10G port plugged into a gigabit switch is a gigabit port. Before paying for a faster LAN row, check what it will be talking to: the switch or router, the cable, and the device at the far end all have to be in on it. If your network's fastest other device is a gigabit NAS, a faster board port changes nothing at all.
Seven of our boards print two speeds in one row — 5G+2.5G on four, 10G+2.5G on three. That is two physical Ethernet ports with two different controllers behind them, and the row lists both.
Two ports do not add up. They are two independent network interfaces: two cables, two links, two addresses. The usual reasons to want them are keeping two networks separate, feeding two different machines or switches, or having a spare port when one behaves badly.
Combining them into one faster link is link aggregation, and it is not what a board does out of the box. It "requires assistance from both the Ethernet switch and the host computer's operating system", both ends have to be set up the same way, and a mismatch fails rather than falling back gracefully: "Link aggregation between similarly statically configured switches may work but will fail between a statically configured switch and a device configured for LACP." A managed switch and deliberate configuration are the price of admission. If you do not have those, treat a dual-LAN board as a board with a spare port.
Our spec table prints a speed. It does not print which chip provides it, and that chip is a real variable: the same 2.5G row can sit in front of parts from different vendors with different driver histories.
The clearest documented case is Intel's own. In January 2023 ComputerBase reported that boards on Intel's LGA1700 socket with 700-series chipsets were increasingly hitting network faults traced to the 2.5 Gb/s Intel I226-V controller: "Mainboards mit Intel Sockel 1700 (LGA1700) und einem Chipsatz der 700er-Serie, wie dem Intel Z790, H770 und B760, sind immer öfter von einem Netzwerkfehler betroffen, der durch ein Problem mit dem 2,5-Gb/s-Ethernet-Controller Intel I226-V ausgelöst werden soll." The same report notes the predecessor had it too, and that the workaround for some users was giving up the speed they paid for: the I225-V "muss je nach Anwendungsbereich im erzwungenen 1-Gb/s-Duplex-Modus betrieben werden."
The useful conclusion is narrow, and it is the only one we will draw: the brand on the controller is not a guarantee, so the controller model is worth looking up on the manufacturer's spec page for your exact board, along with its BIOS and driver release notes. We are not ranking vendors here. We could not read Intel's own support article on the issue — the request was refused — so everything above is attributed to ComputerBase's report and dated to it.
The WiFi row on our reviews reads Wi-Fi 6E on 32 boards, Wi-Fi 7 on 31, Wi-Fi 6 on 7, and nothing at all on 10. The difference between those generations is mostly about spectrum and how many parts of it a connection may use at once.
Wi-Fi 6E is not a new generation. The Wi-Fi Alliance describes it as certification that "offers the features and capabilities of Wi-Fi 6, extended to the 6 GHz band" — the same radio technology with a large, quiet band added, and with it channels 160 MHz wide.
Wi-Fi 7 "was introduced in 2024 and enhances Wi-Fi performance in the 2.4 GHz, 5 GHz, and 6 GHz bands." Three of its changes matter on a spec sheet:
All three describe a link between two devices rather than a property of the board on its own, which is the honest caveat. MLO spreads one connection across bands the access point is offering, so it exists only where both ends do it. The 320 MHz channels are channels in the 6 GHz band, so they need an access point using that band. If the router is not being replaced, the WiFi row is a specification for later, not a feature you get on the day.
On Intel platforms the wireless is often CNVi, an interface where "the network adapter's large and usually expensive functional blocks (MAC components, memory, processor and associated logic/firmware) are moved inside the CPU and chipset", leaving "only the signal processor, analog and Radio frequency (RF) functions" on an external M.2 module. That splits the answer to "can I upgrade the Wi-Fi?" in two. A CNVi module "requires chipset and Intel CPU support", so it is not a card you can swap for any M.2 wireless card you like; where CNVi is not in play, "the Wi-Fi + Bluetooth module has to be the traditional M.2 PCIe form factor", which is the more interchangeable case. Check the board's manual before assuming either.
Ten of the 80 boards we review print No WiFi. For a desktop that sits next to the router and takes a cable, that row costs you nothing. If wireless is needed later, there are three routes.
Every review on the site carries the same two rows near the top of its spec table. Four of ours, showing the range:
2.5G, WiFi WiFi 6E. The ordinary configuration on this catalogue.5G, WiFi WiFi 7. Both rows above the norm, and both only pay off with matching gear at the other end.10G+2.5G, WiFi WiFi 7. A dual-LAN board: two ports, not one 12.5G port.2.5G, WiFi No WiFi. Wired only, at the same speed as most of the boards above it.The full set is in our reviews index, and the compare tool puts the LAN and WiFi rows of two boards side by side.
The standards, dates, cabling requirements and Wi-Fi feature descriptions on this page come from the sources listed below, all read directly. The board counts come from the spec tables of the 80 review pages on this site, counted from their LAN and WiFi rows on 9 September 2026. Nothing here is a measurement of our own: no link was set up, no throughput was recorded, and every speed named is the figure the standard defines rather than one anybody observed. We could not read Intel's support article on the I226-V, and we found no source we could read for a comparison between Intel and Realtek 2.5G controllers, so this page does not make one.
Published 28 August 2026. Board counts taken from our own review pages on 2 September 2026.