3D project performance monitor for studios on shared hardware

The moment that drives someone here: a designer opens a complex scene, the software starts lagging, and they have no idea whether the bottleneck is the model's polygon count, the texture resolution, the number of lights, or simply that another team member's job is hammering the shared workstation's GPU. They just know everything is slow, and they either wait or crash. The complaints land on hardware: 'takes up a lot of resources and powerful computers are needed,' 'lags when heavy models are imported,' 'high system requirements might pose challenges for users with less powerful hardware.'

The reason this hasn't been solved well for small studios specifically is that performance profiling tools exist for game engines and VFX pipelines — but those are built for teams with dedicated technical directors. A five-person product visualization studio doesn't have a technical director. The person doing the modeling is also the person responsible for keeping the pipeline running, and they have no visibility into why a specific scene is expensive until the machine already can't handle it.

What's missing is something that watches a 3D session in real time and surfaces actionable signals before the crash: which assets in the scene are driving memory use, when a texture is large enough to cause slowdowns on import, what the projected render time is before the job starts. The 'you should take a break while rendering' complaint reflects learned helplessness — designers have accepted that they can't predict or manage this, so they just leave.

This is a business because the need recurs with every project and every new team member who doesn't yet know the unwritten rules of what the studio's hardware can handle. Without visibility, the cost is wasted render attempts, lost work from crashes, and time spent waiting on machines that shouldn't be waiting.

What to build

Build a background monitoring agent for Windows workstations that tracks per-application GPU and CPU consumption during 3D sessions in Dimension, Substance, and SpaceClaim, and pushes real-time alerts to a shared Slack channel when a running job is projected to exceed available memory or when render time is trending past a configurable threshold.

Where to start

Start with Nvidia GPU users specifically, where NVML gives reliable programmatic access to per-process GPU stats, and recruit beta users from small Adobe Substance studios who are already complaining publicly about crashes — their pain is high and their hardware cohort is homogeneous enough to make the monitoring reliable.

The hard part

Getting reliable per-application GPU memory attribution on Windows across different GPU vendors (Nvidia, AMD) requires driver-level instrumentation that varies by hardware, making the monitoring data inconsistent enough that early users may distrust the alerts — eroding the core value before the product earns credibility.

How it makes money

Per-seat monthly subscription billed to the studio, with a free single-seat tier for individual freelancers and paid tiers starting at three seats for small studios.