DD-WRT K2.6 is Still Faster than K3.x 8

UPDATE: I repeated my testing in January 2017, and published the results here.

The older K2.6 (Linux Kernel 2.6) versions of the DD-WRT open source firmware are still way faster for both wired and wireless networks than the newer K3.x versions of DD-WRT… at least on a Linksys E4200.

How can I be so certain? Because I did a whole bunch of back to back testing today at my parents’ house.

Why was I doing firmware and network speed testing at my parents’ house? That’s a good question!

It started out as a standard “tech support visit” to their house, because their “Chromecast wouldn’t connect.” I arrived on the scene to investigate.

My parents’ house is not far from an apartment complex, meaning there’s a lot of competition for WiFi channel-space in the area. My MacBook Air picked up no fewer than 21 different wireless  networks from their living room:

WiFi channel scan at my parents' house

WiFi channel scan at my parents’ house

Because there’s a lot less competition in the 5 GHz space near their house, I suggested to my dad that some of his newer 5 GHz enabled WiFi devices might work better if he had a 5 GHz capable router, such as my beloved Linksys E4200. Though admittedly a bit long in the tooth, the Linksys E-series routers are still very capable… especially when combined with open source firmware like DD-WRT. So I temporarily swapped out his 2.4 Ghz D-Link DIR-655 with one of my DD-WRT flashed Linksys E4200 test units to see if its dual-band antennas worked better. My dad’s desktop computer is hard-wired to the router, so he wanted to make sure his wired speeds weren’t affected. He immediately visited Speedtest.net and got 75.18 Mb/s down and 12.15 Mb/s up. He complained that these results were “way slower than it used to be.”

“Really?” I asked. “What do you normally get?”

“At least 125 down, usually,” he replied, and immediately asked me to plug his DIR-655 back in.

I fired up the Linksys’ DD-WRT admin interface, and looked at the build version. It was a beta build that was less than a month old. Then I looked at the Kernel version of the build, and noted that it was a K3.x. I remembered DD-WRT contributor Kong talking about how the K3.x builds might be slower on a lot of hardware, because it uses so many resources… but I didn’t think it would drop download speeds from the 120s into the 70s. So I decided to download a few different DD-WRT builds for the E4200, in addition to the latest Linksys stock firmware for the device and run some back-to-back tests.

Testing Method

My parents’ house isn’t exactly a lab environment, but I did my best to test things as scientifically and fairly as I could.

  • Before and after each firmware flash, I did a 30-30-30 reset, and kept all the firmware’s default values.
  • I didn’t set any WiFi passwords for this test, and used “Auto” for the WiFi channel and channel width selections (I did restore my parents’ WPA2 password when the tests were done, and link to those results below, too).
  • Because my parents use Comcast cable internet, I used the “Seattle, WA – Comcast” server on Speedtest.net for every test.
  • I ran three wired and wireless tests for each firmware, and provided the results were all close (and they all were), I selected the download speed that was in the middle (neither fastest or slowest) as the result to link to from this article.
  • I used my dad’s Dell desktop, wired directly to the router via a Gigabit network card, for the wired tests.
  • I used my iPhone 6 for the WiFi tests, and placed it on the desk in the same location each time, about 3 feet from the router.
  • I used the latest DD-WRT beta build as of the test date, which is 21396, released March 22, 2016.

Test Results

You can click on any of the results to see the actual results for that test hosted on Speedtest.net. Results are posted in the order the tests were taken.

D-Link DIR-655 with stock firmware (v1.37)

Linksys E4200 with DD-WRT 21396 K3.x “mega” build

Linksys E4200 with DD-WRT 21396 K2.6 “mini” build

Linksys E4200 with DD-WRT 21396 K2.6 “mega” build

Linksys E4200 with stock firmware (v1.0.06 – Build 3)

Winners and Losers

It’s clear from the results on the Linksys E4200 that both the wired and wireless speeds for the DD-WRT K3.x build are massively slower than the DD-WRT K2.6 “mini” and “mega” builds, as well slower than the the stock firmware. The “mini” showed the best speeds, but only slightly (though I re-flashed and tested the “mini” build again after the E4200 stock firmware test, and still got slightly faster results for both wired and wireless vs. the “mega” build).

My guess about the 5 GHz antenna was right, and the E4200’s wireless speeds both the K2.6 builds and the stock firmware were consistently above 70 Mb/s, which is far better than the D-Link’s 47.88 Mb/s. But with the K3.x build, either the 5 GHz antenna didn’t connect properly or the router’s resources were bogged down, because it was at least 7 Mb/s slower than the D-link, which only has a 2.4 GHz antenna.

So it seems clear that, at least with a Linksys E4200, that the DD-WRT K2.6 build yields wired and wireless speeds comparable to the stock Linksys firmware, and far greater than the K3.x build… with the “mini” K2.6 build consistently yielding the best results. This supports the argument that the router runs faster when fewer resources (including RAM) are used.

It’s possible that on newer hardware with faster processing power and more RAM that the K3.x DD-WRT builds work just as well (or better) than their K2.6 counterparts.

But for now, I’m sticking with the K2.6 builds for maximum speed… and that’s what I recommend on my DD-WRT E4200 Max Speed Tweaks blog post.