It’s a Solid! How Dong Knows Tech Does Networking Speed Testing for Its In-Depth Hardware Reviews

If you ever wonder how I perform networking speed testing on devices—including Wi-Fi broadcasters (routers, mesh systems, and access points) and switches—for reviews on this website, you’re reading the right post.

I’ll reveal here my testing methods and the equipment I used. If you pay close attention, you might be able to build similar testing “labs” yourself. But it would be best if you took my word for it.

I don’t suggest you create your testing system, but if you do, keep in mind that you’ll always get (slightly) different results, even if you use the same equipment and data sets. That’s because, among other things, Wi-Fi behaves differently from one location to another. That’s just the nature of this popular wireless technology.

This post is mainly for your information if you want to know how I came up with the performance numbers and the rating for each review. Spoiler: It’s easier, though more time-consuming than you think.

Dong’s note: I first published this piece on November 9, 2022, and last updated it on January 12, 2024, to add up-to-date information on new hardware and test procedures.

Networking speed testing for Dong Knows Tech’s reviews: It’s a lot of work

It usually takes me at least a week to finish evaluating a Wi-Fi device and about three days for a network switch—I only review Multi-Gig switches.

During this time, I generally use the hardware for my personal needs over an extended period. The testing mentioned here is mostly about throughput speeds, which only play a part in a hardware review.

In any case, I use the same testing method mentioned in the speed testing post for all reviewed hardware.

Specifically:

• I use a “server” unit, a Windows Pro computer, that connects to the test hardware via a wired connection, which is always Gigabit or faster (Multi-Gig), depending on the hardware.
• For the sustained throughput rates, I perform the test using Windows’ File Explorer (Windows Explorer) from the server to a connected client. It’s a simple drag-and-drop copy test via the standard Server Message Block (SMB, a.k.a. Samba) protocol. I time the process to figure out the performance speed, measured in megabits per second (Mbps).

Below are the general specs of my equipment, which I upgrade pretty frequently—I will update this post when the changes are significant enough.

Server specs for networking speed tests

My server is a custom-built computer with the following specs:

• CPU: Intel Core i5-12600K.
• Motherboard: ASUS ROG Maximus Z790 Formula. This motherboard has a built-in Intel BE200 Wi-Fi adapter and a 5Gbps network port.
• RAM: 32GB DDR5.
• Storage: 2TB Crucial T700.
• LAN connection: ASUS XG-C100C 10GBASE-T network adapter card, in addition to the motherboard’s built-in 5GbE port.

This server hosts specific test files, which I copy to clients via the test device (router, switch, or access point) using Wi-Fi or wired connections to determine the speeds.

Again, the server always connects to the test hardware via a wired connection, which varies from Gigabit to 10Gbps, depending on the device. I don’t test any device that has sub-Gigabit ports.

A router with a Multi-Gig LAN port benefits greatly from this wired-to-wireless testing method regarding sustained Wi-Fi speeds compared to one of the same Wi-Fi specs with only Gigabit LAN ports. The latter’s Gigabit port limits its wireless performance.

Clients’ specs for networking speed tests

The test client is a computer that connects to the test hardware—a Wi-Fi router, an access point, or a switch—using the fastest connection method possible, wireless or wired.

I generally use five clients for official throughput testing. They all have a relatively high-end CPU, 16GB or more RAM, and a top-tier NVMe SSD as the primary storage. They all run Windows 11 Pro (fresh installation with no bloatware) but can be quickly imaged to Windows 10 Pro if necessary.

• Client #1 (for wired and Wi-Fi speed tests): This is a desktop computer with similar specs to the server above, plus an Asus PCE-AC88 4×4 Wi-Fi 5 adapter—the fastest Wi-Fi 5 client on the market. I use this mainly for a close-range (10 feet or shorter) throughput test of a router’s Wi-Fi 5 band. This client also has a 10GbE BASE-T card used for testing wired connection speeds.
• Client #2 (Wi-Fi only): A Dell XPS laptop with an Intel BE200 Wi-Fi 7 adapter.
• Client #4 (Wi-Fi only): An HP AMD-based laptop with an Intel AX210 Wi-Fi 6E adapter.
• Client #4 (Wi-Fi only): A Dell Inspiron laptop with an Intel AX200 Wi-Fi 6 adapter.
• Client #5: An Intel-based Macbook Pro with built-in 3×3 Wi-Fi 5 client. This computer runs both macOS and Windows Pro via BootCamp.

For additional real-world Wi-Fi experience, I also use over a dozen other laptops, desktops, tablets, USB Wi-Fi adapters, IoT devices, and smartphones with different Wi-Fi standards, from Wi-Fi 4 to Wi-Fi 7.

Networking speed testing on Wi-Fi broadcaster and switches: Data and configurations

The following are the data and how I conduct tests on standalone networking devices, including Wi-Fi broadcasters (routers, access points, extenders) and switches.

Depending on a particular device, there might be more testing. For example, a Wi-Fi 7 router will also be tested as though it were a Wi-Fi 6/6E or Wi-Fi 5 router. But all of them have to go through these:

• I upgrade the hardware to the latest firmware available. When applicable, the last mobile app or desktop software is used.
• For speed testing, the test device has all of its features—QoS, Parental Control, etc.—turned off. These features can hinder the throughput speeds.
• I use single large files for test data, which generally take less time to copy than multiple small files. Depending on the tests and how fast a router is, I use a 2GB, 6GB, 10GB, 20GB, or 60GB test file. Generally, I use the smaller test file when a bigger one would make the test take too long. By the way, these test files have no extension—they are just solid chunks of data.
• Wi-Fi broadcasters are first tested with settings that favor speed (and not compatibility, which is generally the default) using each available Wi-Fi channel separately. After that, I also try them using the default (Auto) settings.
• When possible, each frequency band of a Wi-Fi broadcaster is tested separately. Wi-Fi 7’s MLO feature is also evaluated using supported clients.
• All switches are tested in unmanaged mode (applicable only to managed switches.)
• For official performance scores, I test the hardware with just one connection at a time.
  • For Wi-Fi broadcasters: I place this client at two specific locations that are (a) less than 10 feet (3 m) and (b) 40 feet (12.2 m) away from the router—within the line of sight ^(*). I do multiple tests at various times during the day and on different days of the week and use the highest consistent numbers as the final scores.
  • For switches: The client is connected to the switch via a 7-foot CAT6a or CAT5e cable. The best throughput rates are recorded. I might also do additional tests using Gigabit client or CAT5e cables.
  • If a router has two or more Multi-Gig LAN ports, these ports are also tested, just like a Multi-Gig switch.

^(*) This distance is used for consistency, and it’s also where most broadcasters show their best performance over range yet with noticeable degradation.

Besides the performance, I use the hardware for an extended period in real-world anecdotal testing, from a few days to even a few weeks.

During this time, Wi-Fi broadcasters are used with clients of different standards and tiers, placed at various locations and distances, at times behind one or more walls. While this experience is nuanced and can’t be put in numbers, it helps determine a broadcaster’s real-world performance.

In addition to the local data transfer test, I also test the speed via my 10Gbps Fiber-optic Internet connection and report it when applicable.

This test is mainly to confirm the router’s WAN port can deliver the expected speed to the rest of the local network. I don’t use the Internet to test Wi-Fi (or local network speed) and vice versa—it’s never accurate that way.

I report network performance in megabits per second (Mbps).

Wi-Fi mesh speed test: Hardware placement

Generally, I test the multi-unit Wi-Fi system in the wireless setup. That’s because, in the wired backhaul configuration, the extender unit (satellite) performs similarly to the router unit—you can use the router’s throughput numbers as those of the satellite.

For standard testing, I place the satellite 40 feet away from the router unit. Then, the test client is set at 10 feet and 40 feet away from the satellite. All are within lines of sight, with no objects or walls between them.

Also, I use the star topology for the testing, meaning the satellite units (if more than one) are placed around the router unit. In other words, a 2-pack mesh will generally deliver the same test results as a 3-pack one.

When the mesh hardware is Tri-band—where it has an additional 5GHz band—I will test it so that one of the 5GHz bands works as the backhaul and the other works as the front haul. The idea is to get the best possible performance.

In exceptional cases, I also do additional testing on the hardware via wired backhauling.

Similar to the case of standalone broadcasters, I use the mesh in different real-world scenarios, with anecdotal tests, to understand how it works and use that experience for the review.

Wi-Fi 7 mesh wireless backhaul speed test

In early 2024, for any Wi-Fi 7 canned mesh system, I added a test for the wireless backhaul link if the hardware features Multi-Gig ports. Specifically:

• I connect the router unit of the mesh system via a Multi-Gig connection to the test server mentioned above.
• I placed a satellite 40 feet away from the primary router, and the two were connected wirelessly within a line of sight.
• I connect a Thunderbolt-enabled test client to the satellite unit’s Multi-Gig LAN port using an OWC Thunderbolt 3 10G Ethernet Adapter.
• The test client and the server are only two devices connected to the mesh system. During the tests, no other clients, wired or wireless, are linked to the system.

With this setup, I perform the copy test as mentioned above to determine the bandwidth of the system’s wireless backhaul link. This link determines the total bandwidth of the satellite unit shared among its connected clients.

Important note on speed testing

Again, I measure the official test score using a single high-speed Wi-Fi client using a single band at a time. This client pulls data from the test server connected to the router (or access point) via a network cable.

Three things to note:

1. Using a single client is the only way to determine all broadcasters’ speeds consistently. If I used multiple clients, there would be too many variables due to different speed tiers, standards, and the fact that Wi-Fi bandwidth is shared, making it impossible to come up with relevant throughput numbers to say if this router is faster than the other, etc.
2. A router (or AP) with a Multi-Gig LAN port has a significant advantage in test scores over those with only Gigabit ports. That’s because, without this super-fast port, the hardware’s performance is limited at Gigabit.
3. All hardware is officially tested once at the time of the review. Over time, the changes in firmware or software drivers might affect their real-world performance, but the comparison charts generally only show the performance of official tests.

Something to keep in mind: On the one hand, the Wi-Fi scores reported in my reviews are likely those of the best-case scenario—they show the broadcaster’s Wi-Fi band’s total real-world bandwidth. On the other, with devices that have 10Gbps or faster ports, chances are the performance numbers are affected by other devices involved, such as switches, routers, clients, applications, or even the network cables themselves. Since my fastest wired devices are used for the testing cap at 10Gbps, it’s impossible to gauge 10Gbps correctly due to software and hardware overhead.

Router network-attached storage (NAS) speed test

If a router has a USB or eSATA port that can host a storage device, I test the performance of its NAS feature, too.

In this case, for consistency’s sake, the following is the standard way I go about it:

• I use at least two random portable SSDs out of this best list for the test and pick the one with higher consistent scores as the official numbers. So far, any portable SSD has proved to have way higher speeds than the router’s USB port. In other words, regardless of the drive I used, all routers’ NAS performances remained the same.
• The drive is formatted in NTFS. If the router doesn’t support this file system, I’ll use the applicable one and note that in the review. So far, all routers I’ve reviewed support NTFS.
• I perform the test using Windows File Explorer (a.k.a. Windows Explorer). It’s a simple drag-and-drop copy test via the standard Server Message Block (SMB, a.k.a. Samba) protocol with a 20GB single file as the data.
• The test computer has the old and insecure SMBv1 disabled by default, meaning the router must support SMBv2 for the test to work. (If SMBv1 must be enabled, I’ll note that in the review.)
• I do the test using a wired Gigabit connection. If the router has a Multi-Gig port, I’ll also use that port.
• I perform each test (write and read) at least three times and pick the highest consistent numbers (within 5 percent) as the router’s scores.
• I report NAS performance in megabytes per second (MB/s). (You can find the scores of all the routers I’ve tested in this post on the best NAS-enabled routers.)

Note: The test for the router’s NAS feature is just for reference in terms of raw speeds. I use real NAS servers (and I’d always recommend that you do, too.) That said, I generally only use the router’s USB port for a short time and do not try all available applications or functions, nor do I spend time on in-depth analysis to check for conflicts or issues.

Power consumption measurement

Since mid-2021, I’ve started adding real-world power consumption on reviewed hardware that uses a power adapter, which doesn’t apply to PoE or USB bus-powered devices.

I measure this by connecting the hardware via a simple over-the-shell power meter and reporting the result after 24 hours.

It’s worth noting that the measurement is intended to give a rounded ballpark number of the device’s energy use in watt-hours (Wh). It’s not scientifically accurate.

A device consumes more or less energy depending on its loads. In my case, the reviewed hardware is generally being used more intensively than average during the testing phases.

Networking speed testing: You can’t put your finger on it.

After years of working with hundreds of networking devices, I have to admit that the process can get repetitive and tedious. It’s also not 100% accurate. Also, no one can try every scenario, considering a router can have many features and settings.

It’s impossible to say how fast a device truly is because so many factors and elements can affect a test’s outcome. That’s not to mention the fact a router can change dramatically via firmware updates. For this reason, I try to keep my testing as consistent as possible. Ultimately, I aim to show how a particular device does against others.

In other words, my testing doesn’t mean to represent your experience of a device in terms of throughputs. Instead, it describes how better or worse a choice it is compared to others at the time of the review.

Again, speed testing is just one of many things I do to evaluate networking devices. I always take the time to have real-life experiences with them before publishing my reviews. You can count on that.