GameCube technical specifications

Nintendo originally offered a digital video output on early GameCube models. However, it was determined that less than one percent of users utilized the feature. The company eventually removed the option starting with model number DOL-101 of May 2004. The console's technical specifications are as follows.

IBM PowerPC "Gekko" processor

IBM PowerPC "Gekko" processor (180 nm) shaven down to show the silicon die

ATi "Flipper" processor

ATi "Flipper" (180 nm) shaven down to show the silicon die

One of the two 1T-SRAM RAM modules shaven down to expose the die

GameCube Game Disc

	Details	Source(s)
CPU	32-bit 486 MHz IBM "Gekko" PowerPC CPU (based on the 750CXe and 750FX) 180 nm IBM six-layer, copper-wire process, 43 mm2 die with 4.9 watts dissipation 1.8 V for logic and I/O 27×27 mm PBGA package with 256 contacts 18.6 million transistors; of which 6.35 million transistors are for logic Superscalar 3 issue (2 plus branch folding) Out-of-order execution Two 4-stage integer units: IU1 and IU2, 32-bit Front-side bus: 162 MHz 64-bit enhanced 60x bus northbridge to Flipper 1.3 GB/s peak bandwidth (32-bit address, 64-bit data bus) On-chip caches: 32 KB 8-way set-associative L1 instruction cache 32 KB 8-way set-associative L1 data cache Half of L1 data cache (16 KB) can be locked for fast DMA-from-cache in parallel with instruction execution 256 KB 2-way set-associative L2 cache 64-bit Floating Point Unit (FPU) coprocessor: PowerPC 750 + roughly 50 new SIMD instructions, geared toward 3D graphics 7-stage pipeline 32x 64-bit FPR registers Usable as 1x 64-bit (double-precision) or 2× 32-bit (paired singles) SIMD per clock cycle. 1.9 GFLOPS (single precision 32-bit floating point) IEEE compliant Data Compression 2:1 and 4:1 compression for graphics data yields 5.2 GB/s peak effective bus bandwidth Load Q instruction: converts 8-bit or 16-bit, signed or unsigned integers to SP floating point Store Q instruction: converts SP floating point to 8-bit or 16-bit, signed or unsigned integer Branch Prediction Unit (BPU) Load-Store Unit (LSU) System Register Unit (SRU) Memory Management Unit (MMU) Branch Target Instruction Cache (BTIC) CPU performance: 1125 DMIPS (Dhrystone 2.1)
GPU	162 MHz ArtX-designed ATI "Flipper" ASIC (9.4 GFLOPS) 180 nm NEC eDRAM manufacturing process, 51 million transistors (approximately half dedicated to 1T-SRAM), 106 mm2 die Contains GPU, audio DSP, I/O controller and northbridge 3 MB of on-chip 1T-SRAM (2 MB Z-buffer/framebuffer + 1 MB texture cache) with ~18 GB/s total bandwidth Embedded 24-bit Z-buffer/framebuffer RAM: 2 MB (4× 512 KB) Bus width: 384-bit (4 buses, each 96-bit wide) Bandwidth: 7.8 GB/s Sustainable latency: Under 5 ns Embedded texture cache: 1 MB (32× 256 Kb) Bus width: 512-bit (32 buses, each 16-bit wide) Bandwidth: 10.4 GB/s Sustainable latency: Under 5 ns 24 MB (2× 12 MB) 1T-SRAM main memory @ 324 MHz, 64-bit bus, 2.6 GB/s bandwidth 1 vertex pipeline, 4 pixel pipelines with 1 texture mapping unit (TMU) each and 4 render output units (ROPs) Simultaneous textures per pass: 4 Color depth: 24-bit RGB, 32-bit RGBA System floating-point arithmetic capability: 11 GFLOPS (peak) (MPU, Geometry Engine, Hardware Lighting Total) Fillrate: 648 megapixels/sec, with Z-buffering, alpha blending, fogging, texture mapping, trilinear filtering, mipmapping and S3 Texture Compression Raw polygon performance: 66 million polygons/sec 40 million polygons/sec, with fogging, Z-buffering, alpha blending and Gouraud shading 33 million polygons/sec, with fogging, Z-buffering, alpha blending and texture mapping 25 million polygons/sec, with fogging, Z-buffering, alpha blending, texture mapping and lighting 6-12 million polygons/sec, assuming actual game conditions, with complex models, fully textured, fully lit, etc. 16-stage TEV fixed-function texture combiner unit (4 inputs, 1 output)[1] Image processing functions: Volumetric fog, heat haze, motion blur, bloom, subpixel anti-aliasing, per-vertex lighting, 8 hardware lights, alpha blending, hardware transform and lighting (T&L), virtual texture design, multi-texturing, emboss bump mapping, Dot3 bump mapping (normal mapping), lightmapping, shadow mapping, shadow volumes, planar projection shadows, environment mapping, mipmapping, LOD, depth of field, perspective-correct texture mapping, bilinear filtering, trilinear filtering, anisotropic filtering, real-time hardware texture decompression (S3TC) (6:1 ratio), 8 simultaneous texture layers, 256 levels of transparency, clipping, hidden surface removal/culling, Zfreeze, Zcomploc/early-Z reject, bounding box, destination alpha test, alpha test, depth test, render to texture, TEV compare, color combiners, alpha combiners, texture combiners, transparency effects, framebuffer effects, post-processing effects, Gouraud shading, cel shading, dithering, can emulate 1-bit stencil buffer through a Zfreeze function Other: Real-time decompression of display list, hardware motion compensation capability, HW 3-line deflickering filter
System Memory	43 MB total non-unified RAM 24 MB (2× 12 MB) MoSys 1T-SRAM @ 324 MHz (codenamed "Splash") as main system RAM 3 MB embedded 1T-SRAM cache within "Flipper" GPU (2 MB framebuffer/Z-buffer, 1 MB texture cache) 16 MB DRAM used as I/O buffer for audio and DVD drive Memory bus width: 64-bit main system RAM, 896-bit internal GPU memory, 8-bit ARAM Memory bandwidth: 1.3 GB/s Gekko to Northbridge, 2.6 GB/s Flipper to main system RAM, 10.4 GB/s texture cache, 7.8 GB/s framebuffer/Z-buffer, 81 MB/s auxiliary RAM Latency: Under 10 ns main memory, 5 ns texture cache, 5 ns framebuffer memory
Audio	Audio processor integrated into Flipper: custom 81 MHz Macronix 16-bit DSP Sampling frequency: 48 kHz 64 simultaneous channels, ADPCM encoding Instruction memory: 8 KB RAM, 8 KB ROM Data memory: 8 KB RAM, 4 KB ROM External auxiliary RAM: 16 MB DRAM @ 81 MHz Auxiliary RAM bus: 8-bit Auxiliary RAM bus bandwidth: 81 MB/s CPU can read/write blocks from RAM to ARAM through DMA; ARAM can be used for miscellaneous low-bandwidth purposes[2] Stereo output (may contain 5.1-channel surround via Dolby Pro Logic II)
Video Modes	640×480 interlaced (480i) @ 60 Hz 640×480 progressive scan (480p) @ 60 Hz (mostly NTSC games only) 720×576 interlaced (576i) @ 50 Hz (PAL games only)
Connectivity	4 controller ports, 2 memory card slots 2 high-speed serial ports Serial Port 1 is reserved for a broadband adapter or modem adapter Serial Port 2 is unused (DOL-001 models only) 1 high-speed parallel port up to 81 MB/s (reserved for the Game Boy Player) Analog AV Out NTSC models: S-Video, composite PAL models: RGBS (SCART), composite Stereophonic analog audio output Digital AV Out (DOL-001 model only) Interlaced or progressive scan YCBCR synthesized to YPBPR using in-cable custom Macronix DAC[3] RCA (NTSC-U), D-Terminal (NTSC-J) Stereophonic I²S digital audio (not used by any cable)
Storage	8 cm optical GameCube Game Disc Approx. 1.5 GB capacity 16 Mbit/s–25 Mbit/s transfer rate operating in CAV mode 128 ms average access time Memory card Capacities: 512 KB (59 blocks), 2 MB (251 blocks), 8 MB (1,019 blocks, incompatible with some games) 8 KB sectors
Other	Power supply DC 12 volts 3.25 A Dimensions: 4.3 in (110 mm) (H) × 5.9 in (150 mm) (W) × 6.3 in (160 mm) (D)