The origin of the hardware
The PlayStation hardware project began as a collaboration between Nintendo and Sony. In 1988, Sony was developing a CD-ROM add-on for the Super Famicom — known internally as the Play Station — that would allow the SNES to play CD-ROM software. The project fell apart in 1991 after Nintendo signed a competing agreement with Philips that gave Philips more favourable licensing terms than the Sony agreement had provided. The details of the contract dispute have been reported in multiple versions; what is certain is that Sony's Ken Kutaragi — a Sony engineer who had been the primary advocate for the SNES sound chip collaboration that gave the SNES its SPC700 audio processor — redirected Sony's gaming hardware investment from the SNES add-on toward an independent console.
The PlayStation's hardware was designed by a team that included Kutaragi and was influenced by the understanding that 3D polygon graphics were becoming the primary competition axis in the console market. The Sega Saturn, announced at roughly the same time and launching in Japan in November 1994 (the PlayStation launched in Japan in December 1994), was designed with a different priority: it used dual SH-2 processors optimised for the 2D sprite operations that had defined 16-bit gaming, with 3D capability added as a secondary design goal. The PlayStation's GPU — the Graphics Processing Unit — was designed from the beginning around 3D polygon rendering, with 2D sprite support treated as secondary. This architectural choice determined which games would be easier to develop for each platform and which would look better.
The GTE and polygon processing
The PlayStation's Geometry Transformation Engine (GTE) was a coprocessor embedded in the R3000A CPU, running at the same 33.87 MHz as the main processor. The GTE was dedicated to the mathematical operations required to transform 3D geometry: matrix multiplication, perspective division, and coordinate transformation — the calculations needed to project a 3D vertex onto a 2D screen position. These operations, performed for every vertex of every polygon in every frame, were the computational bottleneck for 3D rendering on CPUs without dedicated hardware. The GTE performed these operations in fixed-point arithmetic at speeds the main CPU couldn't achieve, allowing the PlayStation to render approximately 360,000 flat-shaded triangles per second or 180,000 texture-mapped triangles per second under optimal conditions.
The GPU — Sony's proprietary chip — received transformed vertex data from the GTE and handled rasterisation: filling the triangles with colour or texture data, performing hidden surface removal to determine which triangles were visible, and writing the results to the framebuffer. The GPU did not perform Z-buffering — the depth comparison operation that determines which of two overlapping polygons is in front — using per-pixel depth values. Instead, it used an ordering table: the CPU sorted polygons from back to front and submitted them to the GPU in that order, relying on the painter's algorithm to approximate correct depth ordering. This meant that triangles were always drawn in the order the CPU specified; if the ordering table was calculated incorrectly, polygon sorting errors were visible as polygons briefly appearing in front of objects they should be behind. This was a known and accepted limitation of the architecture that developers worked around through careful level design.
The MDEC and texture limitations
The PlayStation's MDEC (Motion DECoder) chip handled hardware-accelerated MPEG video decompression, enabling the full-motion video sequences that CD-ROM storage made practical and that publishers used for cutscenes, pre-rendered backgrounds, and attract sequences. The MDEC was one of the PS1's less discussed but commercially significant components: the ability to play video at CD quality on a game console was a capability that marketing could demonstrate clearly, and pre-rendered backgrounds — full-resolution photographs or rendered environments compressed to video — allowed visual fidelity for static backgrounds that the GPU couldn't achieve for real-time 3D environments. Final Fantasy VII's pre-rendered backgrounds are the most recognisable example: each location was rendered at high resolution and stored as compressed video, with the real-time 3D character models composited on top.
The PlayStation's texture limitations were significant. Textures were stored in a 1 megabyte VRAM texture cache and could be a maximum of 256×256 pixels — small by modern standards, and small enough that textures that looked reasonable at close range became visibly pixelated as the camera moved closer. The GPU's texture filtering was limited to nearest-neighbour sampling, which meant textures didn't smooth as they scaled — the pixelation that is now read as PS1 aesthetic was a hardware constraint rather than a visual choice. Texture coordinates also used fixed-point arithmetic, which produced the characteristic "swimming" or warping that PS1 textures exhibited: flat-shaded polygons on screen as flat planes, but texture-mapped polygons warped as the perspective correction was approximated rather than calculated precisely. The warping was visible in most PS1 games on textured surfaces at oblique angles and is now one of the format's most distinctive visual characteristics.
Why it beat the Saturn
The comparison between PlayStation and Saturn 3D performance was not straightforward because the platforms' architectures prioritised different operations. The Saturn's dual SH-2 CPUs were excellent at the 2D sprite operations that dominated 16-bit game design — sprite scaling, rotation, and the large numbers of independent sprites required for 2D shooters and fighting games. Its 3D performance was lower than the PlayStation's in most comparisons because its quad-based polygon architecture (dividing surfaces into quadrilaterals rather than triangles) and the programming complexity of coordinating two CPUs worked against developers trying to produce competitive 3D results quickly.
The practical consequence was visible in cross-platform ports. Games developed primarily for the PlayStation and ported to Saturn — Tomb Raider, Resident Evil, Castlevania: Symphony of the Night — typically ran at lower frame rates, with fewer on-screen polygons, or with other quality reductions on Saturn. Games developed specifically for Saturn could look exceptional on that hardware — Panzer Dragoon Zwei (1996), Guardian Heroes (1996) — but the development effort required to exploit the Saturn's architecture was greater, and most third-party developers chose to optimise for PlayStation first. The PlayStation's architecture was better suited to the software that defined the era, and the software that defined the era was easier to produce for the PlayStation. The circular reinforcement between hardware architecture and software availability determined market outcome more reliably than hardware specification comparisons predicted.