Is it possible to use two or more GPUs with the UAV Toolbox simultaneously?

In a current research project, we are using UAV Toolbox to assist UAV design. We are using results from partners which run OpenFOAM for CFD and give us the aircraft aerodynamic coefficients. We are simulating the UAV using UAV Toolbox and hardware in the loop to control the aircraft. However, our current PC is struggling. We want to specify a workstation with one or two GPUs to simulate using the UAV Toolbox. We have access to NVidia 4060 with low price and, instead of going to 5080-like GPUs, we are analysing the possibility to use two or more 4060 in parallel.

Answer 1

[Umar]

Hi @Jones,

Thanks for your update.

Noted that you're using UAV Toolbox for UAV simulation with HIL and integrating aerodynamic coefficients from CFD results (OpenFOAM). Given that your current PC is struggling, and you're considering specifying a new workstation with one or two GPUs:

1. GPU Utilization: UAV Toolbox and Simulink primarily rely on CPU for dynamics and control; GPU acceleration helps only in specific areas (e.g., sensor simulation, vision, or visualization). Not all parts of the simulation pipeline will benefit from multiple GPUs.

2. Multiple RTX 4060s: Using two 4060s can be cost-effective for parallel GPU workloads, but gains are limited unless tasks can be distributed cleanly. Also note that 4060s have limited FP64 performance and no NVLink support, so cross-GPU communication can be a bottleneck.

3. Recommendation: If budget allows, consider a single higher-end GPU (e.g., RTX 5000–6000 class) for better compute density and memory bandwidth. Otherwise, two 4060s may suffice if workloads (e.g., image processing or vision pipelines) are split efficiently across them. Also ensure a strong CPU, fast RAM, and NVMe SSD to support HIL performance.

Let me know if you’d like help drafting specs.