Skip to content

Limits are part of the result

  • This is a publisher-operated laboratory, not an independent certification. TitanCSA develops some of the tested drivers.
  • Measurements use one Windows 11 Hyper-V guest and one host. Host workloads were not frozen. Results do not generalize to every PC, Windows build, controller or driver version.
  • Secure Boot is enabled, but guest HVCI is not enabled. Passing these tests does not establish HVCI qualification, HLK compliance, absence of memory leaks or production stability.
  • No physical keyboard, mouse, USB gamepad, USB analyzer or external timing apparatus was used. No physical input latency or Bluetooth transport latency is measured.
  • XInput exposes state. The closed-loop benchmark waits for observation; unobserved intermediate states in a free-running test would not automatically prove report loss.
  • USB emulation, VHF, joystick creation, device hiding and input filtering are different architectures. A global best-driver ranking would be misleading.
  • Windows libvirtualhid driver-backed profiles require a license according to upstream documentation. No license was purchased or bypassed. Source review is not a runtime performance result.
  • oblitum/Interception documents separate commercial-use terms. No commercial license was identified for this publisher's promotional comparison, so its driver is not installed or scored. Its public source and ABI are audited.
  • Microsoft VHF/UdeCx samples were identified as reference architectures; no newly compiled unsigned sample driver was treated as a signed, working competitor.
  • DS4 and Xbox One/GIP have selected virtual delivery results, not exhaustive protocol coverage. GIP-filter and ViGEm-filter attempts with extra uncorrelated captured reports remain excluded from the latency table.
  • Multi-hour soak, crash-injection, physical reconnect, DualSense, physical-device protocol coverage and total driver CPU attribution require additional campaigns. A five-minute monitored HID trial and one separately instrumented WPR/DPC trial are available; neither establishes long-term stability.
  • WDF-001 reproduced a Vdevice queue-purge guest bugcheck in DS4 and GIP controller/filter campaigns. It remains unresolved. Successful completed runs do not override that finding.

Failed runs are listed in the processed results and retained with their diagnostics. No simulation enters the measured charts. No superiority claim is made from a feature checklist or from an unverified filter path.