Gamepad MIDI Latency Benchmark — Real Numbers vs Hardware Controllers

Every time someone hears "gamepad as a MIDI controller" their first question is "yeah but isn't the latency terrible?" Spoiler: it isn't. We ran a 10,000-sample benchmark of a DualSense, an Xbox Series X controller, and a Switch Pro pad against three pieces of dedicated MIDI hardware. The numbers are public, the methodology is reproducible, and the answer is: USB gamepads are within 1–2 ms of a $700 Push 3. Behold the data.

What we measured

End-to-end MIDI latency has three components:

• Input-to-bridge: from the moment the button is pressed to the moment the bridge process receives a HID event.
• Bridge-to-DAW: from HID event to the MIDI message landing in the DAW's input queue.
• DAW-to-audio: from MIDI in to sound out the audio interface. Dominated by audio buffer size, not the controller.

For a fair comparison against hardware MIDI controllers we only measure the first two components — call it "HID-to-MIDI" or "controller-to-DAW". The audio buffer is constant across all tests.

How we measured it

The bridge has a built-in benchmark mode that timestamps every HID event with clock_gettime(CLOCK_MONOTONIC_RAW) on Linux/macOS or QueryPerformanceCounter on Windows. A small Reaper script timestamps every incoming MIDI message with the same clock. Subtracting gives the HID-to-MIDI delta with sub-microsecond resolution.

# Run the benchmark for 10,000 samples
$ ucmidi benchmark --device dualsense --transport usb --samples 10000

[1/3] Capturing HID events...
[2/3] Capturing MIDI events from loopback...
[3/3] Computing deltas...

Median:  4.21 ms
p50:     4.20 ms
p95:     4.78 ms
p99:     5.12 ms
p99.9:   6.04 ms
Max:     7.18 ms
Stddev:  0.41 ms
Samples: 10000 / 10000

Report saved to: ./benchmark-2026-05-27-1421.json

The numbers

Tested on an M2 MacBook Pro running macOS 14.6, audio buffer 64 samples at 48 kHz. Same room, same USB cable where possible, same ten-second warm-up.

Controller	Transport	Median	p99	Cost (AUD)
Ableton Push 3	USB-C	3.1 ms	3.9 ms	$1,499
Xbox Series X controller	USB-C	3.8 ms	4.6 ms	$89
Akai MPK Mini Mk3	USB	4.1 ms	5.0 ms	$129
DualSense	USB-C	4.2 ms	5.1 ms	$109
Novation Launchpad X	USB	4.4 ms	5.5 ms	$269
Switch Pro Controller	USB	4.6 ms	5.4 ms	$99
DualSense Edge	USB-C	3.9 ms	4.7 ms	$329
DualSense	Bluetooth	12.7 ms	18.3 ms	—
Xbox Series X	Bluetooth	14.1 ms	21.0 ms	—

The Push 3 wins, as you'd hope for $1,499. But the Xbox Series controller comes in under the Launchpad X and within 0.7 ms of the Push. The DualSense is essentially tied with an Akai MPK Mini. For 99% of producers this is well under the threshold of perception (~10 ms is where most humans start noticing).

Get Universal Controller MIDI Pro — $49 →

Why the bridge isn't adding latency

The bridge runs in a single high-priority thread that does nothing but read HID events and write MIDI bytes. On macOS that thread gets thread_policy_set with THREAD_TIME_CONSTRAINT_POLICY. On Linux it's SCHED_FIFO at priority 80. On Windows it's MMCSS "Pro Audio" class. The total in-process time from HID event to MIDI write is ~0.3 ms — the rest is OS scheduling and the virtual port.

// The hot loop, simplified
while (running) {
  HidEvent ev = hid_read_blocking(device);
  const auto t0 = clock_now();
  MidiMessage msg = map_event(ev);
  port.write(msg.bytes, msg.length);
  if (benchmark_mode) log_delta(ev.timestamp, t0);
}

Bluetooth is the trade-off

Bluetooth is convenient. Bluetooth is also 8–14 ms slower than USB on every gamepad we tested. The DualSense uses Bluetooth Classic with a 7.5 ms connection interval; the Xbox Series uses BLE HID with a 15 ms interval (configurable down to 7.5 on some host stacks). Neither is suitable for kick-drum-precise live triggering. For producing pads, sweeping filters, or mapping macros to a couch session, Bluetooth is fine.

Limitations of the benchmark

• USB hubs add jitter. A direct laptop port is 0.2 ms tighter than going through a hub.
• USB 3.x ports occasionally renegotiate. If you see a 12 ms outlier, blame USB power negotiation, not the bridge.
• The benchmark measures one event at a time. Pressing four buttons simultaneously increases p99 by ~0.4 ms due to HID coalescing.
• Windows numbers are 1–2 ms higher across the board due to the WDM driver layer. Windows 11 24H2 with MIDI Services closes most of that gap.

TL;DR — if anyone tells you gamepads are too laggy for serious MIDI work, send them the JSON. Run the benchmark yourself with Universal Controller MIDI and post your numbers.