How Does the Nintendo GameCube Work? Hardware Architecture Explained

The GameCube (2001) is a compact, fixed-function console built around a custom IBM/Nintendo/ATI architecture codenamed "Dolphin."

CPU (Gekko): PowerPC 750CXe derivative @ 486 MHz. 32 KB L1 I/D cache, 256 KB L2. SIMD "paired singles" for geometry
GPU (Flipper): ATI/Nintendo @ 162 MHz. 3 MB embedded 1T-SRAM (framebuffer, textures, Z-buffer). Fixed-function T&L, TEV (Texture Environment) pipeline (8 stages, 16 indirect)
Main RAM: 24 MB 1T-SRAM (low latency, high bandwidth ~10.4 GB/s) + 16 MB DRAM (audio, OS, disc cache)
Media: 8 cm Mini-DVD (1.46 GB). CAV/CLV drive ~3.5 MB/s. No DVD-Video playback (no license)
Audio: Macronix DSP (64 voices, ADPCM, 48 kHz). Dolby Pro Logic II
I/O: 4 Controller Ports (serial), 2 Memory Card Slots, Serial Port 1 (modem/broadband adapter), Serial Port 2 (Game Boy Player), Digital AV (component), Analog AV

Power On → IPL (Initial Program Load) from masked ROM in Flipper
IPL validates disc (wobble groove, BCA mark) → loads Apploader from disc
Apploader initializes hardware, loads main.dol (game executable) into RAM
Game OS (thin library, no multitasking) takes control. Direct hardware register access
Frame Loop: CPU builds display lists → GPU renders via TEV → Output to DAC

Controller: Ergonomic "handle" layout, analog trigger range (0-255), octagonal gate, built-in rumble
Game Boy Player: Hardware GBA (CPU + LCD driver) in a cartridge; outputs via GameCube video
Efficiency: 1T-SRAM + TEV pipeline allowed late-gen games (RE4, Twilight Princess) to rival early Xbox 360/PS3 visuals at 480p

GameCube