Practical articles on CPU, GPU, memory and engineering process optimization.
High-Performance Collections: Multithreading, Burst, Jobs! <3: Most teams do not lose performance in one giant bug. They lose it through small managed allocations that stack into garbage spikes, stutter, and a steady review tax.
Cannot have many real-time lights with Forward Shading in URP... can you? (obviously yes)
Forward+ Awesomeness in Unity URP: Lighting can push your scene off a frame-time cliff before anyone notices why. You add more lights, visuals improve, and suddenly performance becomes a firefighting loop.
No GDC, no Unity, but a 20-minute walk I can do
Run-time Material Fusing: You fix geometry batching, then one extra texture or material variant puts you back into a draw-call bottleneck. That is a classic retention leak: visuals look fine, frame pacing does not.
Well, that did not go quite as planned...
Run-time Static Batching - How?!: Your batching strategy can collapse the moment design asks for procedural rooms, runtime layout changes, or spawned props. When that happens, CPU bottlenecks come back fast and your frame budget starts bleeding.
Curious how I get access to help out in AAA games?
Why devs don't know how static batching truly works in Unity
Static Batching Like You Never Saw: You already toggled static flags and still saw batches explode. That is where a frame-time cliff starts: draw calls stay high, CPU work keeps burning cycles, and optimization becomes firefighting instead of progress.
Let us see why late Z-Kills are so bad for your GPU
Exploring Low-Level GPU Metrics With SMM: When GPU performance collapses, generic profiling views are often too high-level to explain why. You need low-level counters, or you stay stuck in expensive guess-and-check loops.