PTX JIT compilation failed on 5070Ti with GROMACS 2025

User discussions
1

GROMACS version: 2025.2
GROMACS modification: No

When trying to launch a simulation on a node with RTX 5070Ti GPU, I get an ‘invalid PTX’ CUDA error:

 1 GPU selected for this run.
Mapping of GPU IDs to the 2 GPU tasks in the 1 rank on this node:
  PP:0,PME:0
PP tasks will do (non-perturbed) short-ranged interactions on the GPU
PP task will update and constrain coordinates on the CPU
PME tasks will do all aspects on the GPU
Using 1 MPI thread
Using 16 OpenMP threads 

starting mdrun 'Protein'
10000 steps,     40.0 ps.

-------------------------------------------------------
Program:     gmx mdrun, version 2025.2-dev-20250512-0eaa03516a
Source file: src/gromacs/gpu_utils/cudautils.cuh (line 281)
Function:    launchGpuKernel<gmx::NBAtomDataGpu, gmx::NBParamGpu, gmx::GpuPairlist, int>(void (*)(gmx::NBAtomDataGpu, gmx::NBParamGpu, gmx::GpuPairlist, int), const KernelLaunchConfig&, const DeviceStream&, CommandEvent*, const char*, const std::array<void*, 4>&)::<lambda()>

Assertion failed:
Condition: stat == cudaSuccess
GPU kernel (k_pruneonly) failed to launch: CUDA error #218
(cudaErrorInvalidPtx): a PTX JIT compilation failed.

I used CUDA 12.2 for building GROMACS.

Carsten

2

I used Cuda 12.9 and Gromacs 2025.2 and got the same error. Try adding an explicit gpu architecture when cmake:

-DGMX_CUDA_TARGET_SM='90;100;120' -DGMX_CUDA_TARGET_COMPUTE='90;100;120'

This solved my problem.

Related links:

3

Were you able to fix it? I’m having the same problem with the 5060.

If you solved it, could you explain what you did? I’m very new to this.

4

My problem vanished after upgrading from CUDA 12.2 to 12.8 and from GROMACS 2025.2 to 2025.3. I believe the relevant step was the CUDA upgrade to 12.8 which then means that on the 5070Ti you don’t rely on generating code from PTX any more.