|
1
|
- Dr. Mike Froggatt
- Sony Computer Entertainment
- Cambridge Studio
|
|
2
|
- Where have we come from?
- History of consoles over last decade
- How did we get where we are today?
- Technological driving factors
- Where are we going in the future?
- Developments in software and hardware
- Challenges to be overcome
|
|
3
|
|
|
4
|
|
|
5
|
|
|
6
|
- Megadrive
- Motorola 68000 (1979)
- 68000 transistors
- PlayStation
- MIPS R3000 (1987)
- Custom 3D geometry coprocessor
- PlayStation 2
- 128 bit Emotion Engine (MIPS III/IV)
- Integrated vector coprocessors
- 10.5 million transistors
|
|
7
|
|
|
8
|
- Super Nintendo
- DMA-driven sprite/background processor
- Hardware background scaling/rotation
- PlayStation
- Fixed function display list processor
- PlayStation 2
- Programmable geometry display list
- Embedded display memory
- Newest PC graphics cards
|
|
9
|
|
|
10
|
- Mask programmed ROM cartridges
- Limits on maximum capacity (64 Mbit)
- Long lead times for production
- Archaeologists will dig them up 1000 years from now!
- CD-ROM
- 650 MB capacity (then multiple discs
)
- Extensive manufacturing infrastructure
- DVD-ROM
- 9 GB capacity (per side)
- Still heading for multi-disc titles
|
|
11
|
- Pure assembly language
- No libraries or operating system
- Very little code re-use
- Programming to the bare metal
- C and assembly language
- System libraries and minimal O/S
- Some code re-use
- C++/C and assembly language
- Extensive system libraries
- Middleware for graphics, physics and networking
- Engine re-use essential
|
|
12
|
- No standard tools
- In-house developed editors
- Commercial packages for art (DPaint)
- Adapt industry-standard tools
- Alias PowerAnimator for modelling
- Custom tools for world/game editing
- Standard tools adapt to games industry
- Extend Maya via plugins
- All 3D editing done with single package
|
|
13
|
- 16 bit generation
- Well behind technology curve
- 32/64 bit generation
- Somewhat behind curve (graphics better)
- 128 bit generation
- On the curve, or much better (graphics)
- The next generation?
|
|
14
|
- Very high computational performance
- PS2 delivers 6.2 GFLOPs (Cray Y-MP!)
- Memory performance also increasing
- But not as much as compute performance
- 30x bandwidth increase from PSone to PS2
- Latency becoming ever more problematic
- Need efficient software techniques
- Work smart, not hard!
|
|
15
|
- Work from a polygonal mesh
- Create new vertices as a weighted sum of existing vertices
- Repeat process recursively to generate smooth surface
- Many different schemes available
- Advantages:
- Similar to editing a polygonal model
- No problems with defining sharp edges
- Avoids problems with texture-mapping
|
|
16
|
|
|
17
|
- Typically 3D games are perfectly in focus at all distances
- Mimicking depth of field of an actual lens can provide increased realism
- Accurate simulation is hard
- Acceptable cheat is to blur and blend image with itself at different
distances from camera
|
|
18
|
|
|
19
|
- Already available for PS2 in US/Japan
- SOCOM has shipped 400K units
- 175K regular online players
- European roll-out in 2003
- Spring in UK
- Elsewhere to follow
- Trials going on at this very moment
|
|
20
|
- Everyone is using middleware
- Scalable for:
- PS2 to PS2 over iLink
- LAN play
- Internet gaming
- Massively multiplayer?
- Why would you want to do that?
|
|
21
|
- Implementation
- But it still isnt fast enough
- Full product download is not practical
- Users will still need to buy discs
- Should we be looking at Korea?
- 20 Mb/s to the home by 2005
|
|
22
|
|
|
23
|
- Faster maths
- Needed for geometry, physics, A.I.
- Faster graphics
- More programmability at pixel and vertex level
- Faster chips
- CMOS-10S (SOI, Cu interconnect, low-k dielectrics)
- Faster memory
- Can we get another 30x?
- Rambus Yellowstone
- Performance target for PS3 is 1000 times PS2!
- Is aggressive parallelism the key?
|
|
24
|
- Experiment in parallel processing
- Sixteen enhanced PS2s connected together
- Peak performance:
- 1.2 billion polygon/s, 38.4 billion pixels/s
- Still not fast enough for some people!
|
|
25
|
- Compilers to target parallel systems
- Coarse or fine-grained parallelism
- Programming languages for graphics
- High-level shading languages on the way
- OpenGL 2, DirectX 9, VectorC
- Tools for art and graphics
- Support the techniques we are using
- Shading languages, subdivision surfaces
- Project management (code/data/staff)
- Everything is getting bigger!
|
|
26
|
|
|
27
|
|
|
28
|
|