# dxdiag hangs indefinitely with /dontskip on driver 616.56 and 616.92 (RTX 5060 Ti) — breaks Call of Duty launch
## Summary
On an RTX 5060 Ti, `dxdiag.exe` invoked with the `/dontskip` flag hangs
indefinitely and never produces its output file. This happens on driver
versions **616.92** and **616.56**, and does **not** happen on **610.88**.
The hang is reproducible outside of any game. It breaks Call of Duty: Modern
Warfare II, whose Activision bootstrapper calls `dxdiag` with exactly those
flags during startup and waits for it to return — so the game sits forever on
the RICOCHET anti-cheat splash screen.
—
## System
| | |
|—|—|
| GPU | NVIDIA GeForce RTX 5060 Ti (8 GB) |
| CPU | AMD Ryzen 7 9700X |
| iGPU | AMD Radeon Graphics (integrated) |
| Motherboard | Gigabyte B650M AORUS ELITE AX ICE, BIOS F42b (2026-04-19) |
| RAM | 32 GB DDR5 (single module, running at 4800 MT/s) |
| PSU | Corsair CX750 |
| OS | Windows 11, build 26200.9457 |
| Secure Boot | Enabled |
—
## Reproduction
Run from an elevated PowerShell prompt:
```powershell
Remove-Item “$env:TEMP\test.txt” -Force -ErrorAction SilentlyContinue
dxdiag /dontskip /whql:off /t “$env:TEMP\test.txt”
Start-Sleep -Seconds 45
Test-Path “$env:TEMP\test.txt”
Get-Process dxdiag -ErrorAction SilentlyContinue | Select-Object Name, CPU
```
### Results by driver version
| Driver | `/dontskip` present | Output file created | dxdiag exits |
|—|—|—|—|
| 616.92 | yes | **no** | **no** |
| 616.56 | yes | **no** | **no** |
| 610.88 | yes | yes | yes |
| 616.92 | no | yes | yes |
| 616.56 | no | yes | yes |
On the affected versions, the process accumulates CPU time and then stalls at
a consistent point — measured at 42.78 s, 43.73 s, 43.32 s and 43.39 s of CPU
across four separate runs. The consistency suggests a deterministic block
rather than a race condition.
Without `/dontskip`, `dxdiag` completes in a few seconds on every driver
version tested.
—
## Symptom in Call of Duty: Modern Warfare II
The Activision bootstrapper launches `dxdiag` during game startup. Captured
command lines:
```
dxdiag.exe /dontskip /whql:off /t
“C:\Users\\AppData\Local\Temp\Activision\bootstrapper\\dxdiag_onboot.txt”
dxdiag.exe /dontskip /whql:off /t
“C:\Users\\AppData\Local\Temp\Activision\Call of Duty MWII\\dxdiag_onboot.txt”
```
`cod22-cod.exe` then sits at a fixed **1012 MB** working set with slowly
increasing CPU and never advances past the RICOCHET splash screen. The
process reports `Responding = True` — it is waiting, not crashed. No
application error is written to the Windows event log.
Reverting to driver 610.88 resolves the game launch immediately.
—
## Possibly relevant
The `dxdiag` GUI, run without arguments on the affected drivers, displays:
> "DxDiag has detected that there may have been a problem accessing Media
> Foundation the last time this program was used. Would you like to bypass
> Media Foundation this time?"
Since `/dontskip` explicitly prevents that bypass, the hang may be in the
Media Foundation enumeration path. On 610.88 the same prompt does not lead to
a hang, so this looks like a driver-side interaction rather than an OS
component fault — see “ruled out” below.
—
## Ruled out
Each of the following was tested and did not resolve the hang on 616.56 or
616.92:
- **Clean driver install** with the “perform clean installation” option
- **DISM /Online /Cleanup-Image /RestoreHealth** — completed successfully
- **sfc /scannow** — found no integrity violations
- **Disabling the integrated AMD Radeon GPU** via `Disable-PnpDevice`
- **Closing all overlay software** — NVIDIA App, NVIDIA Overlay, Discord,
TeamViewer, Gigabyte Control Center
- **Stopping third-party kernel-mode services** — EasyTuneEngineService
(AMDRyzenMasterDriverV31), GigabyteUpdateService
- **Clearing shader caches** — game shadercache, NVIDIA DXCache and GLCache
- **Verifying the RICOCHET kernel driver** — `atvi-randgrid_sr` starts
successfully via `sc.exe start` and reports RUNNING; the driver itself is
not the blocker
- **Closing OneDrive**, which syncs the game’s Documents folder
—
## Notes
CUDA remains functional on 610.88 (`torch.cuda.is_available()` returns
`True`, CUDA 13.0), so the rollback carries no cost for compute workloads on
this machine.
Driver 616.92 was installed automatically by Windows Update, signed by
“Microsoft Windows Hardware Compatibility Publisher”, rather than through the
NVIDIA installer. Whether the WHQL-distributed package differs from the
NVIDIA-distributed one is not something I can verify.
—
## Ask
NVIDIA_bug_report_dxdiag_hang.pdf (7.6 KB)
Confirmation of whether this is reproducible on other RTX 50-series cards,
and whether the Media Foundation enumeration path changed between 610.88 and
616.56.