The PlayStation 2 is the best selling console ever. But what people forget, and what people often give it unfair criticism for, is what it had underneath the hood. The PS2 had, in some ways, some of the most advanced hardware of its generation - but it is often slammed for being “weaker” than the GameCube, Xbox or sometimes even Dreamcast.
GPU: Custom 147MHz “Graphics Synthesizer”
- 294/299MHz MIPS IV/R5900 based CPU core
- Vector Unit 0 (VU0) - 147MHz vector coprocessor tightly tied to main CPU core, often used for polygons, geometry transformations, physics etc (either in parallel or serial)
- Vector Unit 1 (VU1) - 147MHz vector coprocessor which is closer to GPU but still parallel to CPU; often acts as a geometry pre-processor, and during serial VU0 operations the output from VU0 is sent to VU1 to do base work eg. camera, movement
- Two different ways of performing operations with VUs
- Parallel: Results of VU0 sent directly as another display list
- Serial: Results of VU0 sent to VU1
- Variable resolution from 256x224 to 1920x1080, interlaced or progressive
- 4MB embedded memory (32MB of main memory may be used to store textures)
- 48GB/s bandwidth, 2560-bit bus
- 2.352 gigapixel/s (1.2Gpixels/sec with Z-buffer, alpha and texture)
- 32MB RDRAM @ 3.2GB/s (main)
- 2MB audio memory
- 2MB I/O memory (4MB in later slims with PPC Deckard I/O)
- 37.5MHz MIPS I/O processor (PS1 CPU at different, variable clockspeed)
- Image Processing Unit (IPU) for DVD/FMV playback; native hardware MPEG-2 support
- Floating point: 6.2GFLOPS
- Geometry transformation (VU0 + VU1): 150 million vertices/second
- 75,000,000 polygon/s (no texture, flat shaded)
- 37,750,000 polygon/s (1 full texture with diffuse map and Gouraud shading)
- 18,750,000 polygon/s (2 full textures with diffuse map, Gouraud shading, specular light/alpha or others)
So as you can see, the PS2 was no slouch. So what happened?
The biggest issue for the PS2 was that it required a different approach to game development than usual. This made porting difficult, because often code required drastic reworks. Take an example from early in the console’s life of an exclusive game and a ported game; Burnout and Gran Turismo 3.
The difference is very clear; games designed for the PS2 from the ground up fared exponentially better than those which were designed elsewhere, and ported. But something still doesn’t add up; even some exclusive games don’t look as good as their GameCube and XBOX equivalents. So what gives?
The reason for this is because for a long time, the VUs had to be programmed in its own unique assembly. This made them very tough to utilise well. However, eventually tools were released to utilise the VUs using the C language and combined with general updates to development tools and developers sharing tricks, PS2 results improved dramatically from year to year. By 2005, the PS2 was often visually indistinguishable from the GameCube and Xbox.
But this wasn’t always the case. The Xbox’s GPU still supported more shading techniques in hardware and had more total memory, so it was common for Xbox games to still have a bit of an edge. This also meant that some Xbox games did things the PS2 couldn’t (to the same extent), such as pixel shading. The PS2 had to recreate these effects in software.
So what could the PS2 do right, then? The PS2 was capable of better geometry than even the Xbox and Wii, thanks to the VUs acting as geometry processors. The astronomical bandwidth and fill rate also meant the PS2 was excellent at effects and postprocessing; so much so that MGS2/3’s depth of field had to be cut out of the HD remaster because it was problematic for the PS3/360 GPUs, and Zone of the Enders ran at 1 / 3 of the framerate compared to PS2 (despite many compromises, such as quarter resolution effects) until the effects were reworked completely to be more compliant. The PS3 and 360 could not brute force effects like the PlayStation 2.
One game that exploited the bandwidth/fillrate to the limit was Zone of the Enders, as mentioned. The game prided itself on causing storms of effects and still running at 60 frames per second; no game on Xbox, GameCube or Wii tried anything similar because developers who worked with the PS2 knew it wouldn’t stack up.
A game which abused both the bandwidth and geometry capabilities was Metal Gear Solid 3. Easily one of the most advanced games of the 6th generation, MGS3 had incredibly advanced geometry - with every blade of grass in the jungle rendered individually - and a brute-forced, layered depth of field effect that was more accurate than anything until Direct3D 11 introduced proper depth detection.
A final example is Shadow of the Colossus, which I covered extensively in a previous blog post. SOTC had real time bone calculations, simulated HDR and 20,000 polygon monsters with layers of fur and one of the most convincing fur representations of its time. It used the bandwidth to slap layers and layers of textures onto a colossus for the rich fur, and it used the VUs to calculate bone structures. It also used a level of detail (LOD) system to make streaming from disc work efficiently. Another very impressive game.
The PS2 was capable of many brilliant things, and while it did fall short in some areas, it made up in others. So next time someone tells you the PS2 was underpowered… you know what to say.
Thanks for reading this tech talk; if you enjoyed it, leave a like and give me some feedback. You should also check out the previous blog posts if you’re interested. I’d like some suggestions on what to do next, too, but until then thanks for reading.
You need to be logged in to comment