In terms of performance it leaves every other computer I've seen behind in the dust. The RTX 5090 - the fastest consumer GPU in the world - looks like a toy in comparison.
This marks a very fascinating inflection point in GPGPU compute:
CUDA is not the performance leader anymore. MI300X beats even Nvidia's GH200 94GB. You need a cross-vendor language like
OpenCL to leverage its power. CUDA vendor-lock now only penalizes both developers and users to not be able to use the faster AMD GPUs. Good thing FluidX3D is written in OpenCL and runs natively on all AMD/Intel/Nvidia/Apple GPUs and CPUs out-of-the-box. Find the FluidX3D software & full source code on
GitHub: https://github.com/ProjectPhysX/FluidX3D Full benchmark charts & tables: https://github.com/ProjectPhysX/FluidX3D?tab=readme-ov-file#multi-gpu-benchmarks
Thank you Jon Stevens and Clint Armstrong for letting me test your Hot Aisle Inc. machine! Was up and running literally within 5 minutes, couldn't be easier.