Most people think good photogrammetry is all about software. Metashape, Pix4D, Reality Capture, and so on. In reality, the real hero is the hardware under the table.
Back in 2022, we had to build a workstation for heavy industrial photogrammetry work, think thousands of drone photos turning into centimeter-level 3D maps for mines, roads, and large sites. The goal was simple: cut processing time as much as possible and keep the system stable for long, non-stop runs.
Hardware has moved forward since then, but this build is still a very good reference for anyone in India setting up a serious photogrammetry PC today.
Photogrammetry in simple words
Photogrammetry is the process of taking many overlapping 2D images (usually from drones/UAVs) and converting them into a 3D model or map.
It is widely used in:
- Land surveying
- Construction and infrastructure
- Mining and agriculture
The workload is not like normal video editing or gaming. You are often processing 5,000+ high-resolution images. That needs:
- A high core-count CPU
- Strong NVIDIA CUDA performance
- A huge amount of RAM
If any one of these is weak, the whole workflow slows down.
Spec Sheet
This was not built for gaming. This was a pure workstation, tuned for photogrammetry.
CPU
AMD Threadripper 3970X (32 Cores / 64 Threads)
Motherboard
Gigabyte TRX40 Aorus PRO Wi-Fi
RAM
256GB (8×32GB Corsair Vengeance 3600MHz)
GPU
2× NVIDIA Quadro RTX A4000 (16GB each, NVLink)
Storage (Speed)
PNY CS3040 4TB Gen4 NVMe SSD
Storage (Bulk)
2× Seagate Ironwolf 18TB NAS HDD (36TB total)
Power Supply
Silverstone DA-1650 GM (1650W)
Why these parts made sense
1. CPU: Many cores for point clouds
For photo alignment and dense point cloud generation, the software uses the CPU heavily.
The Threadripper 3970X, with 32 cores and 64 threads, was chosen because normal consumer CPUs at that time struggled with such large datasets. This chip allowed the workstation to handle huge projects without freezing or slowing down badly, even when multiple things were happening in parallel.
2. Dual GPUs for depth maps and meshing
Once the basic geometry and alignment are done, the workload shifts to the GPU, especially for depth map generation and meshing.
Why dual GPUs: Most photogrammetry tools scale very well with multiple GPUs. Two RTX A4000s can almost cut depth-map time in half compared to a single GPU, depending on the project.
Why Quadro instead of GeForce:
- Enterprise-grade stability
- Meant to run at 100% load for many hours or even days
- Better suited for professional environments where a crash in the middle of an overnight run is not acceptable
For a survey company or engineering firm, this reliability matters more than a few extra FPS in games.
3. RAM: 256GB so the software can breathe
A normal “high-end” PC with 32GB or 64GB RAM is not enough once you start loading thousands of 4K images.
Here, every RAM slot on the motherboard was filled to reach 256GB.
Reason:
- When RAM is low, the system starts swapping data to the SSD
- This creates bottlenecks
- The CPU then keeps waiting for data instead of crunching it
With 256GB, large projects stay in memory for most of the heavy stages, which keeps the CPU properly fed.
4. Storage: One fast drive, one big archive
Fast CPUs and GPUs are useless if the storage is slow.
This build used a two-tier setup:
- 4TB Gen4 NVMe SSD
- 36TB HDD storage (2× 18TB Ironwolf NAS)
This balance keeps ongoing work fast while still giving enough room to store multiple full sites locally.
Then vs Now
In 2022, this configuration was one of the best price-to-performance setups for a serious survey or mapping company.
Today, we have:
- Newer Threadripper 9000-series CPUs
- RTX 50-series and Blackwell-generation GPUs
Even with these upgrades available, the overall logic of the build has not changed:
- High core-count CPU for alignment and dense point clouds
- NVIDIA CUDA GPUs (one or more) for depth maps and meshing
- As much RAM as your budget allows, ideally 128GB+ for heavy jobs
- Fast NVMe for active work, large HDDs for archive
If you are planning a photogrammetry workstation today, whether for mining, roads, solar farms, or infra projects in India, same rule still applies:
