Typhoon is mostly GPU-based: point clouds are fully resident in GPU memory and most processing occurs there. A capable graphics card with sufficient VRAM matters far more than a high-end CPU or large system RAM.

Platform

Operating System: Windows 10 (64-bit) or later.
Architecture: x64. (ARM64 builds are not currently supported.)
Runtime: .NET 8 Desktop Runtime (installer will add if missing).

Hardware Focus

Component	Minimum	Recommended	Notes
CPU	Any modern 4-core	6–8 cores	Used mainly for I/O & coordination; not performance-critical.
System RAM	4–8 GB	8–16 GB	Data resides on GPU; RAM pressure is typically low but needed mainly for I/O.
GPU	DX11 compatible card w/ 2+ GB VRAM	DX11 compatible card w/ 8+ GB VRAM	VRAM size gates maximum cloud size you can load.
Storage	1 GB free	SSD/NVMe	Faster load from disk to GPU.
Display	1080p	1440p or 4K	Higher resolution aids inspection detail.

DirectX 11 & Drivers

Requires a DirectX 11 capable GPU and current drivers.
Update drivers if you encounter visual artifacts or failed loads. If the issue persists after updating, contact support.

Supported Data Formats (Indicative)

Common point cloud and model formats: LAS, LAZ, PLY, PCD, E57, splat packages (.splat), and IFC (.ifc).

Dataset Guidance (VRAM-Based)

Each point consumes approximately 24 bytes of GPU memory (including position and required attributes). This means roughly:

1 GB VRAM free -> ~53 million points
2 GB VRAM free -> ~107 million points

You can load datasets as long as you have enough free VRAM. Interaction remains smooth; larger sets simply take a couple of seconds to reach full visual detail (depends on points rendered per frame).

Approx Free VRAM	Approx Point Capacity*	Notes
1 GB	~53 M	Leave headroom for frame buffers; avoid loading right at limit.
2 GB	~107 M	Reference figure provided; still smooth interaction.
4 GB	~214 M	Multiple medium clouds or one large cloud.
8 GB	~427 M	Room for very large cloud plus auxiliary data.
12 GB	~640 M	Multiple large clouds concurrently.

*Approximate. Actual usable capacity is slightly lower due to overhead (render targets, staging buffers, color/intensity/label channels). If you approach the limit, consider splitting or reducing the dataset.

Factors such as optional attributes, temporary processing buffers, and driver allocation strategies can shift effective capacity by several percent.

Network Requirements

An active network connection is required for:

Authorization / licensing checks
Telemetry (if enabled)
Automatic updates

Offline operation is limited: clouds already available locally can be opened, but functionality tied to authorization or update services may be restricted until connectivity returns.

Security & Permissions

Standard user privileges are sufficient.
Administrator rights only needed for installation or writing into protected folders.

Troubleshooting

Symptom	Likely Cause	Action
App fails on launch	Missing .NET 8 runtime	Re-run installer; verify runtime installed.
Blank viewport	Outdated DX11 driver	Update GPU drivers via vendor tools.
Early render feels sparse	Full-detail pass still converging	Wait a couple of seconds; increases automatically.
Load fails for large cloud	Insufficient VRAM	Close other GPU-heavy apps; reduce dataset size.

For extremely large or dense scans, ensure adequate free VRAM before loading to avoid partial or failed imports.