GROMACS version: I think 2023 or latest
GROMACS modification: No
I am using GROMACS on Windows through WSL2. Currently, relying on CPU (i5 12th Gen or i7 8th Gen) power for MD simulations. I read the GROMACS guidelines, and it says I can integrate Intel GPU support through the OneAPI system.
I am a medical researcher, not well-versed in software developments (compiling, integrating, etc). Still, I tried to ChatGPT this issue (since I could not find any video on YouTube or on GROMACS forum. From my experience, I am able to install the OneAPI base-tool kit successfully, I set the envrironment, but when I run the sycl-ls command, it does not detect the GPU…
Initially, after installing the OneAPI base-toolkit, I ran the following commands (as instructed by ChatGPT) to confirm (I don’t know what, lol):
icx --version
Intel(R) oneAPI DPC++/C++ Compiler 2025.2.1 (2025.2.0.20250806) Target: x86_64-unknown-linux-gnu Thread model: posix InstalledDir: /opt/intel/oneapi/compiler/2025.2/bin/compiler Configuration file: /opt/intel/oneapi/compiler/2025.2/bin/compiler/../icx.cfg
dpcpp --version
icpx: warning: use of ‘dpcpp’ is deprecated and will be removed in a future release.
Use ‘icpx -fsycl’ [-Wdeprecated] Intel(R) oneAPI DPC++/C++ Compiler 2025.2.1 (2025.2.0.20250806) Target: x86_64-unknown-linux-gnu
Thread model: posix InstalledDir: /opt/intel/oneapi/compiler/2025.2/bin/compiler
Configuration file: /opt/intel/oneapi/compiler/2025.2/bin/compiler/../icpx.cfg
sycl-ls
[opencl:cpu][opencl:0] Intel(R) OpenCL, Intel(R) Core™ i7-8700 CPU @ 3.20GHz OpenCL 3.0 (Build 0) [2025.20.8.0.06_160000]
According to ChatGPT, the installation was successful, but as you could see, it did not, detect Intel GPU, only CPU is listed (I don’t know if this is relevant, but this PC also had a Radeon 520 GPU, initially, I spent so much time configuring this GPU but failed, I think it is not supported by GROMACS, anyway).
ChatGPT said, there could be two reasons:
#ChatGPT
Right now, your SYCL device list only includes OpenCL CPU, which means:
The Intel GPU runtime (
level-zeroorintel-opencl-icd) isn’t visible or installed in WSLOr, your Windows GPU driver does not yet expose the GPU to WSL
Step 1. Check your Windows side
Run this in PowerShell (Windows, not WSL):
Get-WmiObject Win32_VideoController | Select-Object Name,DriverVersion
I checked this in PowerShell, it was Ok.
#ChatGPT
Step 2. Install GPU runtime inside WSL
In WSL (Ubuntu), run:
sudo apt update sudo apt install intel-level-zero-gpu intel-opencl-icd clinfo
I ran this in WSL.
sudo apt install intel-level-zero-gpu intel-opencl-icd clinfoReading package lists…
Done Building dependency tree…
Done Reading state information…
Done E: Unable to locate package intel-level-zero-gpu
GPT said: “Means your Ubuntu (inside WSL) doesn’t yet have Intel’s GPU runtime repository configured.” It suggested the following solution:
#ChatGPT
Step 1. Add Intel GPU runtime repository
Run these commands inside Ubuntu (WSL):
wget -qO - https://repositories.intel.com/graphics/intel-graphics.key | sudo gpg --dearmor -o /usr/share/keyrings/intel-graphics.gpg echo 'deb [arch=amd64 signed-by=/usr/share/keyrings/intel-graphics.gpg] https://repositories.intel.com/graphics/ubuntu jammy arc' | sudo tee /etc/apt/sources.list.d/intel.gpu.jammy.list sudo apt updateStep 2. Install Intel GPU runtimes
sudo apt install intel-opencl-icd intel-level-zero-gpu level-zero-dev clinfo
Step 3. Verify GPU detection
sycl-ls
Now, I got this error:
sycl-ls
Segmentation fault (core dumped)
ChatGPT said, there could be two reasons for this error:
“Segmentation fault (core dumped)” from
sycl-lsis actually a known symptom when:
The Intel GPU runtime (Level Zero + OpenCL) isn’t fully installed or mismatched with your Ubuntu version, or
You’re on WSL 2 without GPU passthrough enabled yet (so there’s no accessible
/dev/dxgdevice).
I did the 2nd step first, and got this output:
ls /dev/dxg
/dev/dxg
GPT said,
Perfect — that means your Intel GPU is visible to WSL2 via
/dev/dxg, so GPU passthrough is active!
We just need to install the proper Level Zero + OpenCL runtimes so the Intel oneAPI stack can actually use the GPU.
Then, it provided an updated code (don’t know what’s the difference):
sudo rm /etc/apt/sources.list.d/intel-graphics.list
echo “deb [arch=amd64 signed-by=/usr/share/keyrings/intel-graphics.gpg] ``https://repositories.intel.com/graphics/ubuntu`` jammy main” |
sudo tee /etc/apt/sources.list.d/intel-graphics.list
sudo apt update
But, even after running these modifications, I got the same outout
sycl-ls
Segmentation fault (core dumped)
Now, wondering, what’s the solution, GPT seems to be confused now, or out of options. Want to ask a simple basic question as well, is it even possible to integrate Intel GPUs in this way? Thank you.