Skip to content

Why Is a Multi-Sample Session Limited to ~15 Datasets in Chrome?

When setting up a Multi-Sample session, Google Chrome and Chromium-based browsers limit the number of active sample viewports to around 15 datasets per session, whereas Mozilla Firefox allows loading more datasets.

This difference stems from browser-level WebGL graphics architecture:

  • Chrome & Chromium Browsers (Chrome, Edge, Brave, Opera) impose a strict hard limit of 16 active WebGL contexts per domain/origin. Given the UMAP takes one WebGL context, and each sample slot initializes an independent WebGL context, a multi-sample session only allows 15 datasets.

  • Mozilla Firefox manages WebGL contexts dynamically with higher allocation thresholds (up to 32 or more contexts per origin) and recycles off-screen GPU memory more aggressively, allowing larger grid sessions.

Session Size Recommendation

We do not encourage running sessions with a very large number of datasets (e.g., >10–15 slides).

Even if your browser supports allocating more WebGL contexts, rendering dozens of gigapixel datasets simultaneously causes high GPU VRAM consumption, tile streaming delays, and browser interface lag.

Large Cohort Exploration

Consider the following best practices:

  1. Sub-Cohort Sessions: Split datasets into logical contrast groups (e.g., 6 Control vs. 6 Treated slides) rather than loading 30+ slides in a single grid.
  2. Focus Mode or Dual-Focus Mode: Switch from Gallery Mode to Focus Mode (1 slide) or Dual-Focus Mode (2 slides) to suspend off-screen viewports and significantly lower active WebGL context overhead.
  3. Single-Sample Sessions for Individual Slides: For deep-dive single-slide analysis, launch the dataset in a Single-Sample session to access 100% of GPU resources.