@wersy wrote:
I can highly recommend a mini PC like this for Fusion.
That recommendation might be correct for many Fusion users. Still, I often find that the more experienced a user is and the more they understand what is going on under the hood of their computer and software, sometimes the less computing power they need, because they can apply better workflows and techniques.
Often the most powerful hardware and software is underutilized because the user uses the wrong techniques and/or are not aware of limitations.
3 years ago, I built myself a computer with an AMD Ryzen5950x and 64Gb of RAM and an NVIDIA 3070Ti FE graphics card. I splurged on a 48" OLED TV I use as a monitor.
That machine has multiple purposes, all of which revolve around 3D Computer Graphics Software such as Blender, SideFX Houdini, Indigo Renderer. LuxCore Render and Fusion.
Most CAD modeling tasks in Fusion use only a single core on that machine. CAD modeling is not why I wanted the 16 cores!
However, there are a couple of modeling-related things in Fusion that are multithreaded. There the 16 cores make a difference.
1. Local STEP model import. I heavily use imported geometry in my professional work. Manufacturing automation systems use thousands of components. When you open such a STEP model in File->Open Fusion and monitor core utilization with the resource monitor or Task Manager, you can see that Fusion uses many cores during the conversion. As it pertains to STEP import and export, Fusions outperforms SolidWorks by magnitudes.
2. T-Spline conversion. When you convert a quad-mesh into a T-Spline and then a T-Spline into surfaces, certain aspects of that are also multithreaded and use all cores.
While not modeling related, rendering in the Render workspace in Fusion is done entirely on the CPU, NOT on the GPU. That also uses all 16 cores!
However, if you are modeling with a lot of geometry, say 10000+ components, the GPU can also make a difference in viewport navigation.
Oddly enough, where the 16 cores of my CPU make most of a difference is when I render specific models in Indigo Renderer. Many models can utilize simple path tracing. Simple Path Tracing is where GPU rendering comes into effect and Indigo and Blender do this beautifully.
Unfortunately, the models I render and the way I develop my LED lights and the optics for them involve tiny, high intensity light sources (LED's) and dielectric materials (Acrylic Glass/Plastic). The key works her is Caustics. That unfortunately does not allow me to utilize the GPU. A Pathtracing render that after hours on a 3070Ti is still almost black is good enough after a few minutes of Bidirectional Pathtracing with MLT (Montecarlo Light Transport). That, unfortunately is not possible on a GPU (yet).
It is also rendering and lighting simulations where I use most of the 64GB of memory on my machine.
I guess the conclusion of the above is that I find it difficult to recommend any hardware without know what exactly the user wants to do with the hardware.