Dev Log: January- Physics EngineR&D

Apex Engine Dev Log

January 1, 2026

January 1 is not about launches. It is a line in the sand.

2025 was a year of foundational work. Much of it invisible by design. Architecture validation, subsystem testing, refactors, and technical decisions that only matter if you plan to scale a real-time collaborative engine beyond a single-user prototype.

One of the more time-consuming areas this past year was physics evaluation and validation.

Physics Evaluation: PhysX vs Havok

Apex Engine has very specific requirements around determinism, collaboration, and long-term platform flexibility. Physics is not an isolated system in that context. It directly impacts synchronization, world state consistency, live editing, and multi-user authority.

Throughout 2025, I spent time validating both NVIDIA PhysX and Havok against those constraints, including cross-platform behavior, renderer independence, and integration cost at engine level. This marked our third dedicated R&D testing run, initiated in response to a significant NVIDIA PhysX update pushed earlier in the year. Re-running validation after major upstream changes is deliberate, not redundant, and is necessary to ensure prior assumptions still hold.

A recurring point of confusion in industry discussions is the relationship between physics engines and graphics APIs. This testing phase was as much about separating myths from reality as it was about performance.

Below is the consolidated reference table used internally during evaluation.

NVIDIA PhysX Platform and API Compatibility (Updated)

Key Clarifications

PhysX vs Graphics APIs

PhysX does not depend on any graphics API. Vulkan, DirectX, Metal, and OpenGL are consumers of physics output, not requirements. Physics runs independently and feeds transforms to the renderer.

Vulkan Support

CPU PhysX is fully compatible with Vulkan-based renderers. Vulkan is irrelevant to PhysX internals. PhysX does not use Vulkan compute shaders.

The HLSL Myth

PhysX does not require HLSL. GPU PhysX uses CUDA C++. HLSL may appear only in optional visualization or sample layers, not in the physics simulation itself.

Practical Takeaways for Engine Design

Havok and NVIDIA PhysX are both mature, capable physics solutions, each with distinct strengths. After extensive evaluation, we have chosen NVIDIA PhysX as the primary physics foundation for Apex Engine.

That decision was not based on branding or familiarity. It was driven by a close alignment with our clients’ needs, particularly around deterministic simulation, cross-platform support, Vulkan-first rendering, and long-term maintainability in a real-time collaborative environment.

Havok continues to be a strong solution in specific contexts, and its design tradeoffs remain relevant across the industry. For Apex Engine, however, PhysX provides the right balance of control, flexibility, and platform independence required to support live collaboration, shared world state, and consistent behavior across development, testing, and production environments.

• Cross-platform engine
• Vulkan-first renderer
• Apple compatibility
• Avoid vendor lock-in
• Deterministic simulation

Looking Forward

January marks a transition from internal validation to outward clarity.

The focus moving into early 2026 is not rushing features. It is making systems understandable, dependable, and composable. Real-time collaboration is not a feature layered on top of Apex Engine. It is a constraint that informs every system decision, including physics, networking, tooling, and permissions.

If 2025 was about proving the foundation, 2026 is about making that foundation usable without compromising the principles it was built on.