GPU support: disabled! Any simple solution?

User discussions
1

GROMACS version: 2025.3
GROMACS modification: Yes/No
Here post your question

Hi all,

I have recently installed Gromacs following the Quick and dirty installation on gromacs.org, but gmx -version shows “GPU support: disabled”. Do I have to uninstall the existing Gromacs2025.3 to realize GPU support with a complete installation?

Is there any simple method to get GPU support?

Thanks in advance.

2

Do I have to uninstall gromacs? Many thanks.

3

See my answer to your other thread here.

4

Thanks a lot. When downloading CUDA toolkit from CUDA Toolkit 13.0 Update 2 Downloads | NVIDIA Developer by choosing linux — x86_64 — Ubuntu — 24.04 — deb(local) or runfile(local), I dont known whether deb(local) or runfile(local) is appropriate, because someone says this one is better and someone says the other. My workstation has AMD Ryzen 9950X CPU and RTX 5090 GPU. Or maybe neither one is just fine?

5

They should be equivalent, just install a CUDA toolkit version to link to GROMACS. For GROMACS 2025.3 you can safely use the 13.0 that you are downloading! If you want older versions of GROMACS (< 2024.6) then you will want to switch to an older toolkit version < 13.0.

6

Thanks a lot. Your answer touched another concern of mine: I downloaded and installed gromacs 2025.3.tar.gz, but gmx –version command gives the following:

GROMACS version: 2023.3-Ubuntu_2023.3_1ubuntu3
Precision: mixed
Memory model: 64 bit
MPI library: thread_mpi
OpenMP support: enabled (GMX_OPENMP_MAX_THREADS = 128)
GPU support: disabled
SIMD instructions: SSE4.1
CPU FFT library: fftw-3.3.10-sse2-avx
GPU FFT library: none
Multi-GPU FFT: none
RDTSCP usage: enabled
TNG support: enabled
Hwloc support: hwloc-2.8.0
Tracing support: disabled
C compiler: /usr/bin/cc GNU 13.2.0
C compiler flags: -fexcess-precision=fast -funroll-all-loops -msse4.1 -Wno-missing-field-initializers -O3 -DNDEBUG
C++ compiler: /usr/bin/c++ GNU 13.2.0
C++ compiler flags: -fexcess-precision=fast -funroll-all-loops -msse4.1 -Wno-missing-field-initializers -Wno-cast-function-type-strict SHELL:-fopenmp -O3 -DNDEBUG
BLAS library: External - detected on the system
LAPACK library: External - detected on the system

Is this indicating that my gromacs version is actually 2023.3 rather than 2025.3?

7

Yes this doesn’t look like a normal installation to me, the version line should look something like GROMACS version: 2025.3. How did you install it? Are you in a HPC environment or on your machine?

8

I installed gromacs on my own machine, following the guidance in this video https://www.youtube.com/watch?v=rSDu2bCMqAQ. The steps in video was basically the same as that in the Quick and dirty installation in gromacs.org, and I get the same result every step as in the video. Do I need to uninstall the existing gromacs to rebuild?

9

I think I have figured out what’s going on. Two versions exist, and the gromacs 2025.3 is not in the default path. Thanks. Next I will try the NVIDIA driver and the CUDA toolkit.

10

Yes, most likely. Check the directory where you installed the 2025.3 version and be sure to source that and you should be fine!

11

I think I have installed the gromacs 2025.3 with GPU support:

GROMACS version:     2025.3
Precision:           mixed
Memory model:        64 bit
MPI library:         thread_mpi
OpenMP support:      enabled (GMX_OPENMP_MAX_THREADS = 128)
GPU support:         CUDA
NBNxM GPU setup:     super-cluster 2x2x2 / cluster 8 (cluster-pair splitting on)
SIMD instructions:   AVX_512
CPU FFT library:     fftw-3.3.10-sse2-avx-avx2-avx2_128-avx512
GPU FFT library:     cuFFT
Multi-GPU FFT:       none
RDTSCP usage:        enabled
TNG support:         enabled
Hwloc support:       disabled
Tracing support:     disabled
C compiler:          /usr/bin/gcc-12 GNU 12.4.0
C compiler flags:    -fexcess-precision=fast -funroll-all-loops -march=skylake-avx512 -Wno-missing-field-initializers -O3 -DNDEBUG
C++ compiler:        /usr/bin/g++-12 GNU 12.4.0
C++ compiler flags:  -fexcess-precision=fast -funroll-all-loops -march=skylake-avx512 -Wno-missing-field-initializers -Wno-cast-function-type-strict SHELL:-fopenmp -O3 -DNDEBUG
BLAS library:        External - detected on the system
LAPACK library:      External - detected on the system
CUDA compiler:       /usr/local/cuda/bin/nvcc nvcc: NVIDIA (R) Cuda compiler driver;Copyright (c) 2005-2025 NVIDIA Corporation;Built on Wed_Aug_20_01:58:59_PM_PDT_2025;Cuda compilation tools, release 13.0, V13.0.88;Build cuda_13.0.r13.0/compiler.36424714_0
CUDA compiler flags: -O3 -DNDEBUG
CUDA driver:         13.0
CUDA runtime:        13.0

The Hwloc and Tracing support show “disabled“. Does this matter?

12

Hello again. I have successfully enabled GPU acceleration for gromacs, but without Hwloc nor Tracing support. Does this matter? Thanks.

13

Not as far as I know. If you pass the make check tests you are good to go!

14

No, you should be fine! Like @obzehn says, just check if you pass the tests. Hwloc is not really that important if you’re just running on your local workstation or laptop since it helps with hardware detection on compute clusters, and tracing support is useful if you want to run in-depth performance analysis.

15

Thanks a lot. I passed the make check tests.

16

Thanks for the explanation. I passed the make check tests.