What each stage asks for
Photo alignment scales with processor cores. The dense cloud and meshing run on the GPU (CUDA in Metashape, RealityCapture and Pix4D) and need system memory to hold the whole project. Texturing uses GPU and disk. And GIS work (QGIS, ArcGIS Pro) prefers a fast core, plenty of memory and a drive that reads layers and rasters quickly.
Processor: cores to align, clock speed for GIS
For projects of up to a few thousand photos, a 16-core Ryzen 9 aligns quickly and is the fastest in GIS, which barely uses several cores. For projects of tens of thousands of photos or LiDAR clouds of billions of points, a Threadripper brings more cores and, above all, the memory channels those volumes demand.
Memory and GPU
As a guide, 64 GB for projects of up to two or three thousand photos, 128 GB for large projects and LiDAR point clouds, and 256 GB for whole cities or blocks of many flights. The GPU, 16 GB to start (RTX 5070 Ti or 5080) and 32 GB (RTX 5090 or RTX PRO 4500) for very dense meshes; photogrammetry software uses CUDA, so NVIDIA.
Drives: where time is lost without noticing
A drone flight is tens of gigabytes of photos and the processed project, several times more. A 2 to 4 TB NVMe for the current project, another for the system and the software, and 8 or 16 TB drives (or a server) for the archive. Copying a 200 GB project from a slow drive takes longer than aligning it.
| Projects | Workstation | Price |
|---|---|---|
| Up to a few thousand photos, daily GIS | Creative Pro: 16-core Ryzen 9, 64 GB, RTX 5070 Ti, 2 TB + 4 TB NVMe | €5,280 |
| Large projects, point clouds, two 4K monitors | Pro: 16-core Ryzen 9, 64 GB ECC, RTX 5080, 2 TB + 4 TB NVMe | €6,770 |
| Tens of thousands of photos, city-scale LiDAR | Advanced Plus: 32-core Threadripper, 128 GB ECC, RTX 5080, 2 TB + 4 TB NVMe | €14,265 |
Notes by program
RealityCapture and Metashape scale well with the GPU and with cores; Metashape also uses several GPUs if present. Pix4D is similar and asks for a lot of memory when meshing. QGIS and ArcGIS Pro rely on a fast core, memory and disk; ArcGIS Pro accelerates some tools with the GPU. CloudCompare and LiDAR viewers load the whole cloud into memory: allow a gigabyte per hundred million points, with headroom.
Related hardware and guides
Frequently asked questions
GeForce or RTX PRO for photogrammetry?
Photogrammetry software uses CUDA and a GeForce performs very well. An RTX PRO brings ECC memory and certified drivers; it pays off if the machine works for hours unattended or if you also use certified CAD.
How much memory for a LiDAR cloud?
Allow a gigabyte per hundred million points loaded, and leave headroom for the program and the layers. For clouds of a billion points, 64 GB is tight; 128 GB is reasonable.
Will a laptop do?
For GIS and small projects, yes. For dense clouds and meshing, a laptop runs out of memory and cooling after two hours. The usual setup is a laptop in the field and the workstation in the office.
Is a Threadripper worth it?
From tens of thousands of photos or city-scale point clouds, yes: more cores to align and four memory channels for the volumes. Below that, a 16-core Ryzen 9 is faster in GIS and costs less.
Shall we size your workstation?
Tell us which software you use and the size of your projects. We will propose the processor, memory, GPU and drives, with a price, within 12 working hours.