Introduction
Play Station 3 is a powerful video game console that provides a platform to the virtual world of games produced by Sony computer entertainment. It provides the ultimate gaming experience with its robust multimedia capabilities. With excellent visual & sound effects and having thousands of games, Playstation-3 is a clear leader in-game console market. It’s the internal hardware architecture of Playstation-3 along with its software that makes it a high-performance output machine.
Architectural development
Like computers, the gaming console is also a processor which has memory and architecture which processes the digital data and produces the audio & video outputs through the interfaced peripherals. The hardware architecture of Playstation-3 utilizes a 64 bit 3.2 GHz cell processor developed by IBM having 8 synergistic processing elements (SPE), 256 MB of XDR DRAM, NVIDIA RSX graphics chip with 256 MB GDDR3 RAM for video memory which can also access the cell’s main memory i.e. the XDR DRAM and can generate 1080p output signals also known as full HD. (Gschwind, 2005)
The cell processor equipped with 512 KB of L2 cache acts as a “managing processor” It manages the computational workload and assigns it to the 7 SPE processors using1 PowerPC processing element (PPE). The main function of a cell processor is to efficiently use the capabilities of the available SPE whereas a reserve SPE is available for the case of failure of any of the 7 functioning SPE. SPE are 128 bit single instruction multiple data (SIMD) vector processors each of them clocked at 3.2 GHz. The internal communication is carried away by Element interconnect Bus (EIB) which can transfer data at 204 GB/s. Every SPE uses its own memory flow controller & Direct memory access controller. (Stokes, 2000) (Langdon, 2008)
Another controller is Flex I/O controller which communicates with the NVIDIA RSX graphic chip & the southbridge components such as hard drive, Ethernet, USB, memory card reader, WiFi, Optical Media like Blue-ray. The Bandwidth of Flex I/O is
60 GB/s and the majority of this bandwidth are allocated for the communication with the RSX graphic chip. (Blachford, 2005)
PowerPC Processing Element (PPE)
The PPE contains a 64 bit, dual-thread PowerPC Architecture RISC core. It has 32 KB L1 I-cache & 32 KB L2 D-cache along with 512 KB L2 unified cache. The PPE provides an excellent platform to control the functioning of SPEs. As PPE uses PowerPC instruction set and therefore the clock speed comparisons for this architecture lose significance. (Koranne, P. 20)
Synergistic Processing Element (SPE)
SPE are completely independent processors clocked at 3.2 GHz. It’s the core processor of the system and this vector processor is also known as single instruction multiple data processors (SIMD) which means that at one clock cycle multiple operations can be performed by a single instruction. Also the SPE access 128 registers with 128 bits each with a dedicated memory controller which makes the operations on 32 bit single precision floating point data types possible with high speed & performance.
The SPE along with PPE are in order processors which makes the role of the compiler very important. (Chen, Raghvan, Dale & Iwata, 2005)
The Architecture Behavior
The Element Interconnect Bus (EIB) is connected to PPE, 8 SPE, 1 Flex I/O controller & 1 XDR memory controller. There is an EIB protocol which is followed by the system for the communication. The EIB is connected by 4 16 byte data rings with 64 bit tags. The EIB bus is capable of transferring 96 bytes per cycle and is capable of handling over 100 outstanding requests. These components acts as a slave to the EIB master which controlled by an arbiter. If one component needs to send data then the component sends a request for the ring of the bus then once acknowledged the ring is been allocated to that specific slave for instance 3rd SPE. Being a vector processor there could be multiple request & multiple allocation of rings in the bus, so data can be sent & received by multiple components simultaneously, also called multithreading. (Stokes, “The Cell Architecture”)
This system helps managing the memory recourses within the system. Apart from the memory controller all other components can request read & write operation to the EIB, here the I/O related request is given the highest priority. Each and every element of this cell has its own memory controller. The PPE instructions are carried out through L2 Cache memory but the SPE is controlled by software to access the RAM. This can be considered as completely separate application of code & memory. This feature of keeping SPE controlled by software speeds up the cell and adds security to the system. (Chen, Raghvan, Dale & Iwata, 2005)
NVIDIA RSX Graphics chip
As the graphics are the most important part of the game console there has to be a dedicated microprocessor to do the graphic processing. That is why this is known as Graphical Processing Unit (GPU). NVIDIA RSX is a 550 MHz GPU having specifically allotted 256 MB GDDR3 video memory which can also access the main memory i.e. XDR memory. The special feature is that it has a dedicated vertex shader pipeline & dedicated pixel shader pipeline as compared to unified shader pipeline in Xbox 360.
The GPU is also connected to 256 MB XDR memory through Flex I/O interface. As mentioned earlier the SPEs are made independent as they have separate memory controller, this can be a cause of concern if the functioning of SPE overlaps with the functioning of the vertex processors on the RSX. This is simplified in the architecture as it is up to the programmer to decide whether to use GPU or the SPE to accomplish the tasks. (Dokken, 2006) (Minh, 2008)
Software Development
The software gives the flexibility to the programmer to use the cell processor according to their need. The cell is not designed for a specific computer or a specific game therefore the programmer can modify the memory distribution as per the need.
This is an example of parallel programming and usually it is considered very complex, but the operating system in playstation-3 distribute the memory according to the tasks & availability of resources. Here the programmer has to design a parallel distribution system initially and the task management is taken care by the operating system itself.
Other Peripherals
Hard disc drive: Playstation-3 has a hard drive of 20 GB & its latest versions are coming with 60GB hard drive which means a lot of storage capacity.
Optical media drive: Playstation-3 has Blue ray drive which is a clear front runner when compared to DVD. The capacity, high definition movie content, copyright protections are some of the features where blue ray beats DVD. The only disadvantage that the blue ray drive has, when compared to DVD is the data read speed which results in increase in load time.
Controllers: Playstation-3 has motion sensors to provide better control. It also has the pressure sensitivity and supports 7 Bluetooth devices. That means along with increasing number of players the headset, keyboard & mouse can also be connected simultaneously. (Stokes, 2005).
Performance analysis
The biggest advantage of the Playstation-3 architecture is that the RSX has dedicated memory of 22.4 GB/s and the cell has dedicated 25.6 GB/s main memory and additionally if the programmer wants to assign extra memory to the RSX then the RSX can fetch the memory from cell’s memory through Flex I/O controller on the EIB.
Where as in Xbox 360 the CPU access through the GPU’s memory controller and shares the same 128 bit bus to GDDR3.
The number of transistors in NVIDIA RSX graphical chip alone is more than the combined transistors of cell processor unit and graphical processing unit in any other game console. (Shimpi, 2010)
Conclusion
In terms of games processing power the Playstation-3 definitely holds an upper hand although Xbox have higher bandwidth but the number does not reflect on performance as the bandwidth is been shared between GPU & CPU where as the Playstation-3 has dedicated bandwidth for the cell processor & GPU. Also the SIMD feature is a remarkable one in Playstation-3. Having said that no technology is good enough to last forever, there are certain aspects where other console takes the lead but considering the overall performance “The Plastation-3” is an extremely powerful machine which takes you, closest to the reality.
List of References
Blachford, N, 2005, Cell Architecture Explained.
Chen, T, Raghavan, R, Dale, J & Iwata, E, 2005, Cell Broadband Engine Architecture & its first implementation. Web.
Dokken, Torr, 2006, The Graphics Processing Unit (GPU) as a high performance computational recourse for simulation and geometry processing. Web.
Gschwind, Micheal, 2005, Chip Multiprocessing & Cell Broadband Engine. Web.
Koranne, Sandeep, Practical computing on the cell broadband engine, 1st edition, 2009, Springer, New York.
Minh, Tri Do Din, 2008, GPUs – Graphics Processing Units. Web.
Shimpi, Anandlal, 2010, Sandy Bridge integrated graphics performance. Web.
Stokes, John, 2000, SIMD Architectures. Web.
Stokes, John, 2005, introducing the IBM/Sony/Toshiba Cell Processor – Part II: The Cell Architecture.
W Langdon & W Banzhaf. 2008, A SIMD interpreter for genetic programming on GPU graphics cards. Web.