Windows Server 2022 RDP session crash after initial login (Event ID 36: 0x80004003, 0x80070102 )

Question

Windows Server 2022 RDP session crash after initial login (Event ID 36: 0x80004003, 0x80070102 )

승우 정 0

Hello,

I am experiencing an RDP session crash on a bare-metal Windows Server 2022.

1. Environment

OS: Windows Server 2022 Standard (Physical Bare-metal)

RDP Port: TCP 5729

Network: Closed internal network

2. Symptoms

TCP 5729 port check (Test-NetConnection) succeeds with TcpTestSucceeded : True.

Initial account login and user profile load succeed (Event ID 21 and 22 are logged normally).

However, right after desktop entry or upon reconnecting, the RDP client disconnects with error 0x10b.

Subsequent reconnection attempts hang with a timeout.

3. Event Logs (TerminalServices-LocalSessionManager)

User's image

Event ID 36 (Error): An error occurred when transitioning from CsrConnected in response to EvCsrInitialized. (ErrorCode 0x80004003) Note: 0x80004003 is E_POINTER (Invalid pointer).

Event ID 36 (Error): An error occurred when transitioning from CsrConnected in response to EvCsrInitialized. (ErrorCode 0x80070102) Note: 0x80070102 is ERROR_TIMEOUT (Wait operation timed out).

Event ID 40 (Information): Session 2 has been disconnected, reason code 2147952454 (WSAECONNRESET).
Continuous Loop: Following the crash, Event ID 36 and Event ID 40 are continuously repeating in a loop upon every reconnection attempt.

4. Observations

In Device Manager, the discrete GPU has a yellow exclamation mark (Code 43 / driver error).

Even after setting the discrete GPU to "Disable device" in Device Manager, the RDP session still crashes with the same 0x80004003 / 0x80070102 errors.

OS reinstallation is not possible due to licensing and lack of prior backups.

5. Questions

Does this sequence of errors during EvCsrInitialized (0x80004003 / 0x80070102) indicate that the RDP graphics pipeline is crashing while attempting to allocate display buffers from the faulty discrete GPU?
Even after disabling the GPU in Device Manager and setting "Use WDDM graphics display driver for Remote Desktop Connections" to Enabled, the crash persists. Should this WDDM policy be set to Disabled instead to force software fallback?
To completely stop RDP from binding to this faulty GPU, is it necessary to disable the PCIe slot in BIOS or physically remove the GPU card?

Thank you for your help.

0 comments

1 answer

Your answer

Answer 1

Hola 승우 정,

Based on the symptoms and the Code 43 on the discrete GPU, I'd treat the graphics stack as the likely cause and move directly to forcing RDP away from the hardware GPU.

The key policy is: Computer Configuration > Administrative Templates > Windows Components > Remote Desktop Services > Remote Desktop Session Host > Remote Session Environment > Use hardware graphics adapters for all Remote Desktop Services sessions

Set that policy to Disabled, then reboot the server. When disabled, Remote Desktop Services uses the Microsoft Basic Render Driver instead of the physical GPU for remote sessions. Microsoft documents that behavior here.

I would also set Use the WDDM graphics display driver for Remote Desktop Connections to Disabled and reboot again. That forces RDP off the WDDM remote display path and onto the older XDDM path. Microsoft still documents this behavior for the policy.

This combination is the most practical workaround when a faulty or incompatible display driver disrupts RDP session initialization. There are also multiple Microsoft Q&A cases in which disabling the WDDM RDP display driver resolved RDP freezes/crashes associated with NVIDIA and other display driver issues.

The fact that disabling the physical GPU in Device Manager didn't fix the problem doesn't completely rule out the graphics stack. RDP can still initialize through WDDM components and the remote display driver, even when the discrete adapter itself is disabled. That is why changing the RDP rendering policies is more important here than simply turning off the device.

I would not remove the GPU physically or turn off the PCIe slot yet. First force RDP to use the Microsoft software rendering path and legacy remote display driver. If the server becomes stable after that, you've confirmed that the graphics subsystem is the source of the issue without making any hardware changes.

If that resolves the issue, I'd then replace or reinstall the GPU driver before deciding whether the card itself is faulty. Code 43 can be caused by hardware failure, but it can also be raised when the display driver cannot initialize the device correctly.

Thanks,

James