Because of how Implicit vs. Explicit models work, GPUs will only really offer GPU speed increases to Abaqus/Standard models where large amounts of storage are required and "solutions" are required. On the other hand, Abaqus Explicit models do not need to perform iterations to solutions, and do not need to use as much storage, and therefore run mostly on the CPU and will not (for the most part) benefit from inclusion of GPUs.

So firstly, I would make sure the model you are running is an Abaqus/Standard analysis.

Secondly, using the gpus is more easily done in the command line with "gpus = 1". So if you're doing it from Abaqus CAE, it might be easier to run your jobs from the command line to ensure it is actually using the GPU.

Lastly, it's important to acknowledge that for most FEA problems, GPUs only offer a marginal improvement in speed. There is a document listed on Nvidia's website about using GPUs for Abaqus (https://www.nvidia.com/es-la/data-center/gpu-accelerated-applications/abaqus/)

and the graph at the bottom shows improvements but not "dramatic" by any means. So I would not expect it to make your model run significantly faster, and as a result, there is only so much load that it will make on your GPU, since a lot of the speed limitations lie beyond the GPU.