UE5 on AWS g4dn instance type (Linux machine)

Hi,

Currently, I’m trying to deploy a prebuild package on an AWS instance with pixelstreaming. I followed the guide on UE5 for Linux. As you can guess, it doesn’t work as expected.

I used both a plain old ubuntu and installed everything myself (drivers,etc) as well as the official nvidia image based on ubuntu. In both cases, I receive the following error message when starting the UE5 application (not pixelstreaming).

[1127/174215.925291:WARNING:gpu_process_host.cc(1274)] The GPU process has crashed 10 time(s)
Resizing viewport due to setres change, 1024 x 768
[1127/174217.908004:ERROR:gpu_process_host.cc(955)] GPU process launch failed: error_code=1002
[1127/174217.908038:WARNING:gpu_process_host.cc(1274)] The GPU process has crashed 11 time(s)
[1127/174218.957955:ERROR:gpu_process_host.cc(955)] GPU process launch failed: error_code=1002
[1127/174218.957995:WARNING:gpu_process_host.cc(1274)] The GPU process has crashed 12 time(s)
[1127/174218.958014:FATAL:gpu_data_manager_impl_private.cc(414)] GPU process isn't usable. Goodbye.
Signal 5 caught.
Malloc Size=262146 LargeMemoryPoolOffset=262162
CommonUnixCrashHandler: Signal=5
Alignment > 16 is not supported
Engine crash handling finished; re-raising signal 5 for the default handler. Good bye.
Trace/breakpoint trap (core dumped)

Can anyone help me out with this issue or point me to someone or something that can help ? I went through the forum, but the most issue or treads are for Windows machines.

Kind regards,
Arif

Hey Arif
I’m also facing a similar issue when I build the package using unreal 5.1
but ue5.0 version runs fine
are u facing similar case
did you get it fixed?
if yes please help me with thatt!

Did you guys figure it out?
@S_h_a_s_h_1 @eredjar

Crash on Linux with preview and latest 5.8 branch

Signal 11 caught.
Malloc Size=262146 LargeMemoryPoolOffset=262162
Alignment > 16 is not supported
CommonUnixCrashHandler: Signal=11
Malloc Size=262146 LargeMemoryPoolOffset=524338
[2026.05.29-09.36.52:300][  2]LogCore: === Critical error: ===
Unhandled Exception: SIGSEGV: invalid attempt to read memory at address 0x00007f9fb8083fd8

[2026.05.29-09.36.52:300][  2]LogCore: Fatal error!
Additional Crash Context (Key=“Breadcrumbs_Parallel_0”, Value="Breadcrumbs ‘Parallel’

Radiosity

LumenSceneLighting

Scene

RenderGraphExecute - /ViewFamilies

SceneRender - ViewFamilies

Frame 2
")[2026.05.29-09.36.52:315][  2]LogRHI: Error: Breadcrumbs ‘Parallel’

Radiosity

LumenSceneLighting

Scene

RenderGraphExecute - /ViewFamilies

SceneRender - ViewFamilies

Frame 2

Additional Crash Context (Key=“Breadcrumbs_Parallel_1”, Value="Breadcrumbs ‘Parallel’

FXAA

PostProcessing

Scene

RenderGraphExecute - /ViewFamilies

SceneRender - ViewFamilies

Frame 2
")[2026.05.29-09.36.52:317][  2]LogRHI: Error: Breadcrumbs ‘Parallel’

FXAA

PostProcessing

Scene

RenderGraphExecute - /ViewFamilies

SceneRender - ViewFamilies

Frame 2

Additional Crash Context (Key=“VKhashes”, Value=“Vertex=D600737B1B121E7E, Pixel=25A40D72244320A0 (tid:559645, pt 0)”)[2026.05.29-09.36.52:318][  2]LogVulkanRHI: CreatePipeline: Vertex=D600737B1B121E7E, Pixel=25A40D72244320A0 (tid:559645, pt 0)

LogInit: Running DelayedAutoRegister Phase IniSystemReady
LogInit: Running DelayedAutoRegister Phase EarliestPossiblePluginsLoaded
LogInit: ExecutableName: CrashReportClientEditor
LogInit: Build: UE5-CL-0
LogInit: Platform=Linux
LogInit: MachineId=a6d5304c4eab4c14a98f8c97a3c9f883-000003e8
LogInit: DeviceId=
LogInit: Engine Version: 5.8.0-0+UE5
LogInit: Compatible Engine Version: 5.8.0-0+UE5
LogInit: Net CL: 0
LogInit: OS: Fedora Linux 43 (Workstation Edition) (7.0.9-105.fc43.x86_64), CPU: Intel(R) Core™ i9-14900KS, GPU: UnknownVendor
LogInit: CPU Model ID: 6b71, microcode revision: 133
LogInit: Compiled (64-bit): May 29 2026 11:00:17
LogInit: Architecture: x64

GPU says unknown vendor but it’s

AMD Radeon™ RX 7900 XT

Mesa version: Mesa 25.3.6