Keyboard Matrix Scanning, N-Key Rollover (NKRO) and Ghosting Elimination: Competitive Benchmark Analysis and Hardware Interactions

[Jayden Morris]

I’ve been bench-testing a few keyboards for tournament play and keep hitting a wall with the matrix scanning question. My current setup uses a diode-matrix controller advertised with NKRO, but I’m seeing input drops during fast chorded presses in competitive FPS titles. Before I commit to buying gear for my roster, I need real-world data: what scanning frequency are most of you actually getting from your controllers, and does the advertised NKRO hold up under rapid multi-key activation, or do you hit ghosting at a certain key count? Have any of you measured frame-time deltas between 1000Hz polling matrices versus 8000Hz, and was the difference measurable in actual gameplay or just paper specs?

Also curious how you’re handling debouncing—software debounce delays versus hardware solutions. I’ve read that aggressive debounce settings can add 5-10ms of latency, but insufficient debouncing causes phantom registers. How do you balance that tradeoff, and what latency numbers have you actually measured on your rig? If anyone’s run controlled A/B tests comparing ghosting behavior across different matrix architectures, I’d love to see your methodology and results so I can standardize what we bring to competition.

[Max Rivera]

Great topic to break down, because a lot of people conflate ghosting and NKRO when they’re actually different phenomena. The root of ghosting is purely a matrix-scanning artifact: cheap keyboards route keys in a grid of rows and columns without a diode on each switch, so when you press three keys forming an L-shape, the current back-feeds through the matrix and the controller reads a phantom key that nobody touched. Adding a diode in series with each switch forces current to flow in one direction, which is what eliminates true ghosting entirely. That’s why the difference between a “6KRO without ghosting” keyboard and a “6KRO with ghosting” board comes down to whether they bothered to diode each key, not the rollover count itself.

NKRO is a separate spec that governs how many simultaneous presses the firmware will report before it starts dropping input — a properly dioded 6KRO board can still drop keys beyond six, while a full NKRO board (often USB HID UHPRO, up to 32 keys) will report every keyhold regardless. For competitive FPS specifically, the practical takeaway is that NKRO rarely matters for single-hand combos since you’re not pressing more than ~4-5 keys at once on WASD, but it absolutely protects you in games with heavy macro bindings or fighting/sim titles. One thing worth checking before you pay the NK

"Ryzen 7 7800X3D | 32GB DDR5 6000MHz CL30 | Arctic Liquid Freezer III 360"

[Author]

Posts