v148 · WebGPU

Compat vs Core

The recommended pattern: try core first, then compatibility, then fall back gracefully. This page probes both adapter types on your device and shows which path would be taken by a real-world WebGPU app.

Feature detection: checking…

This device — live probe results

WebGPU available—
Core adapter—
Compat adapter (featureLevel: "compatibility")—
Recommended path—
maxTextureDimension2D (adapter in use)—
1

Try core adapter

requestAdapter() or requestAdapter({ featureLevel: "core" })

checking…
2

If null, try compatibility adapter

requestAdapter({ featureLevel: "compatibility" }) — runs on OpenGL ES 3.1 / D3D11

checking…
3

If both null, use Canvas 2D or WebGL

WebGPU is unavailable on this device entirely — degrade to WebGL2 or canvas

checking… 
// Robust adapter detection: core → compat → fail
async function getBestAdapter() {
  if (!navigator.gpu) return { adapter: null, level: "none" };

  // 1. Try core (modern GPU)
  const core = await navigator.gpu.requestAdapter();
  if (core) return { adapter: core, level: "core" };

  // 2. Try compatibility (older GPU, Chrome 148+ origin trial)
  try {
    const compat = await navigator.gpu.requestAdapter({
      featureLevel: "compatibility"
    });
    if (compat) return { adapter: compat, level: "compatibility" };
  } catch {}

  // 3. No WebGPU at all
  return { adapter: null, level: "none" };
}

const { adapter, level } = await getBestAdapter();
if (level === "compat") {
  // Avoid: fine derivatives, storage buffers in vertex stage,
  //        cube arrays, per-attachment blend, linear interpolation
}

see also

implementation reference

Need the exact API surface, compatibility boundaries, errors, lifecycle, and source links? Read the matching gendn reference ↗