Impeller's Aiks vs Skia: What the Profiler Actually Shows
We ran 40 minutes of GPU traces on Impeller. The rasterizer thread tells a different story than the marketing deck.
Pixel 7 Pro · iPhone 15 Pro · macOS 14 · Flutter 3.22 · RN 0.74 · KMP 0.9
// Skia/Impeller vs JS bridge render calls
// Measured on Pixel 7 / iPhone 14
// State preserved vs full component remount
// Baseline with no assets
// npm has volume advantage; quality varies
// real apps, real stores, real users
Migrated 3 native apps to single Flutter codebase. 40% reduction in UI code, zero jank complaints post-migration.
E-Commerce · 2019 → PresentPixel-perfect parity between Android and iOS. Single widget tree, platform-adaptive navigation, 99.2% crash-free sessions.
Fintech · ProductioniOS, Android, and embedded display from one codebase. Custom render pipeline for in-car HMI at 60fps.
Automotive · EmbeddedBrownfield integration. JS bridge latency visible in animation profiler. Migrating critical paths to native modules.
Social · HybridFlutter leads on rendering consistency and hiring velocity. React Native leads on raw package count. KMP is the right call if your team is already native-Android. The full report breaks this down by team size.
28 pages · 12 benchmark categories · 6 production case studies
// not ready for the full report?
⟳ Subscribe to the Changelog// one email per Flutter stable release · no noise
// arrives when Flutter tags a stable. contains: what changed, what broke, what you should care about.
We ran 40 minutes of GPU traces on Impeller. The rasterizer thread tells a different story than the marketing deck.
Not another todo app comparison. We measured rebuild counts, test coverage ergonomics, and team onboarding time.
Pigeon generates type-safe channel code. KMP uses Kotlin directly. We measured the serialization cost at 10k calls/sec.
Cold build times, cache hit rates, and the hidden cost of code signing automation across all three CI providers.