Wookash Podcast GPU Texture Compression Explained
Apr 4, 2026
Ignacio Castaño, graphics engineer and creator of Spark, specializes in GPU texture compression and tools. He traces his path from early DIY coding to NVIDIA and indie engine work. The conversation covers how block-based formats like ASTC work, making ASTC encoding feasible on GPUs, Spark and Spark.js for real-time compression, performance trade-offs, real-world uses, and the business of shipping graphics tools.
AI Snips
Chapters
Transcript
Episode notes
How Ignacio Learned By Coding On Paper
- Ignacio learned programming self-taught via QBasic and writing code by hand when he lacked a compiler or working PC.
- He typed pages of code in limited one-hour daily sessions, which made him a very fast typist and persistent learner.
Early GPU Optimization That Mattered At Crytek
- At Crytek Ignacio kept geometry resident in GPU memory using vertex buffers, letting the CPU only issue draw calls and enabling massive levels.
- That change contrasted with per-frame submission of polygons and unlocked rendering of gigantic scenes.
Hardware Features Need Both Data And Timing To Win
- Pushing for low-level hardware features can be politically fraught; Ignacio advocated warp intrinsics for block-wide scans and saw the feature added elsewhere later.
- His tests and C-model work validated need, but internal timing and priorities determined rollout.