Additional information

RealityScan Workstations – Hardware Requirements & Recommendations

Epic’s RealityScan (formerly RealityCapture) is a demanding 3D photogrammetry application that requires balanced, high-performance hardware. Since Epic has not yet published official requirements, these recommendations reflect hardware capabilities and widely adopted best practices for large-scale photogrammetry workloads. The configurations below are selected as the best options for professionals seeking speed, stability, and scalability.

About RealityScan & Hardware Demands

RealityScan converts thousands of overlapping photographs into detailed 3D models, placing sustained load on every subsystem:

• Image Alignment – CPU-intensive; benefits from high core counts.
• Dense Cloud & Mesh Generation – GPU-accelerated with CUDA; scales with VRAM.
• Texture Processing – Demands high GPU memory and fast storage.
• Large Datasets – Out-of-core processing continuously reads/writes to disk; NVMe speeds directly impact total time.

Performance depends on the balance between CPU, GPU, memory, and storage—bottlenecks in RAM or disk speed can negate gains elsewhere.

Hardware Recommendations

Processor (CPU) – Most Critical for RealityScan

The CPU has the largest impact on overall throughput. High core counts accelerate large reconstructions, while high-end mainstream processors excel on smaller projects.

• AMD Threadripper™ 9980X (64 cores) – Best choice for large-scale photogrammetry with extremely high image counts.
• AMD Ryzen™ 9 9950X (16 cores) – Excellent option for small to medium projects with strong performance at lower cost.
• Intel Core™ Ultra 9 285K (24 cores) – Strong alternative for users preferring Intel’s hybrid architecture.

Graphics Card (GPU) – CUDA Required

RealityScan requires an NVIDIA GPU for CUDA acceleration; without it, mesh/reconstruction functionality is limited. GPU choice influences speed, VRAM headroom, and performance in companion apps (Unreal Engine, Blender, Autodesk).

• GeForce RTX™ 5080 16GB – Recommended single-card solution with excellent price/performance.
• GeForce RTX™ 5090 32GB – Premium option with maximum VRAM for massive image sets and complex texturing.
• Dual GPU Configurations – Can add ~5–13% additional performance; most users are well served by one or two GPUs.

Memory (RAM) – Scale With Project Size

Out-of-core design keeps requirements modest, but more RAM reduces bottlenecks and speeds processing:

• 32GB DDR5 – Recommended minimum for small projects.
• 64GB–128GB DDR5 – Best range for medium to large datasets.
• 128GB+ ECC/REG – Ideal for enterprise workloads and multi-day jobs.

Guidelines by project scale: 2,000 images → 16GB min; 4,000 → 32GB; 8,000 → 64GB; 16,000 → 128GB.

Storage (Drives) – NVMe for Best Performance

RealityScan frequently reads/writes during processing, making SSD speed critical. Recommended layout:

• OS & Apps: 1TB NVMe SSD
• Cache: 1–2TB NVMe SSD dedicated to temporary data
• Projects: 2–4TB NVMe or SATA SSD for active datasets
• Archive/Backup: Large HDDs or NAS with redundancy (e.g., RAID)

NVMe drives offer substantially higher throughput than SATA SSDs, directly benefiting out-of-core workflows.

Why Choose a VRLA Tech RealityScan Workstation

A VRLA Tech RealityScan Workstation is engineered specifically for photogrammetry—built for sustained loads, stability, and future-proof scalability:

• Faster alignment, reconstruction, and texturing on well-balanced, high-core CPUs.
• Greater stability on long, resource-heavy projects with ECC and workstation-grade components.
• Scalable RAM and storage to match growing dataset sizes.
• Seamless compatibility with Unreal Engine, Blender, Maya, 3ds Max, and related pipelines.
• Backed by VRLA Tech’s 3-year warranty and expert support.