Issue using CUDA 11.0 on NVHPC 23.3

I am trying to build an app that uses OpenACC. We have installed NVHPC SDK 23.3. The installed driver is:

NVIDIA-SMI 510.39.01 Driver Version: 510.39.01 CUDA Version: 11.6

So I try to build using CUDA 11.0 which comes in the SDK. I set -acc -gpu=cuda11.0 in the compiler options, but I get this error.

nvlink fatal   : Input file '/path/to/nvc-23.3/Linux_ppc64le/23.3/comm_libs/11.0/nvshmem/lib/libnvshmem_device.a:alltoall.o' newer than toolkit (118 vs 110)

We have used the modulefile that comes with the SDK, so the environment should be set up properly.

Why is this error happening?

We have been using NVHPC 21.9 and haven’t been able to use a newer version due to different issues/bugs in the toolkits.