Mobile game optimisation

A scene can run comfortably on your development handset and struggle on a midrange device. A busy main thread, a streaming stall or expensive GPU work can each set the limit. Another device may start well, then lose frame rate as it warms. Memory and bandwidth constraints change the choices available to your team.

We help your team work through those constraints on Android devices, iPhones and iPads. Start with the device range, representative scenes and visual target. We identify the limiting work, agree the trade-offs, and implement the changes in your engine branch.

We investigate gameplay, animation, physics and render submission alongside allocations and cache behaviour. Data layout and algorithm changes can reduce the work first; SIMD through Arm NEON, vectorised kernels and ISPC where the target toolchain supports it can improve suitable data-parallel paths. We measure the result on the target devices.

Multithreading has to account for the device’s performance and efficiency cores, the operating system’s scheduler and its power budget. We tune job sizes, dependencies, lock contention and waiting behaviour, and keep background work from competing with the frame. CPU time, responsiveness and sustained energy use all matter.

We trace asset reads, decompression, deserialisation and uploads as a connected pipeline. Asynchronous disk I/O, request batching, asset packaging, caching and bounded background jobs help control startup time and streaming hitches without letting memory use grow unchecked. The implementation uses the storage facilities available on Android and iOS.

Mobile network conditions change during a session. We investigate payloads, serialisation, update frequency, caching and download scheduling against bandwidth, latency and connection loss. Retries, reconnection and resumable downloads need to fit the game’s state and power constraints as well as its data-delivery budget.

We agree a session length and test conditions that expose sustained behaviour, recording temperature and operating clocks alongside the workload. That gives us a useful baseline for the devices your players will use.

Captures connect rendering markers to bandwidth, local memory, shader stalls and CPU/GPU scheduling. On Android, we use Snapdragon Profiler, Sokatoa and Arm Performance Studio on their respective platforms. For iOS, Xcode’s Metal frame capture and Instruments help us inspect GPU work, CPU activity and memory use on the target iPhone or iPad.

The fix may involve attachment lifetime, material evaluation, pass structure, transfers or asset preparation. We compare the result across the target devices and check frame time, memory, power and image quality together, so a gain on one GPU does not hide a regression on another.

Your engine and rendering leads speak directly to the engineers making the changes. We work in your codebase and follow your build and review process. Priorities stay tied to the game’s production schedule and the visual decisions your team owns.

Work with Qualcomm, Samsung and Arm has given us practical experience with mobile workloads, ray tracing and neural-graphics integration. We use that knowledge to ask better questions about your captures and make implementation decisions your team can review.