0 Members and 1 Guest are viewing this topic.
I think that the reason is that the the processors used in the latest games consoles are designed for parallel processing with around 9 individual processors (with their own cache memory) on the chip. They are also designed to be fast. Given the right application, that requires parallel processing, and the time to write code in the appropriate assembler language, which is no mean feat, then I suspect that such an acceleration factor is likely compared with a normal PC. However, some newer PCs also have multiprocessor chips though the industry is slow to develop software that can make good use of them except in some specific applications. My guess the comparison is not wholly fair and that comparing the best PCs with the best games consoles would still give a clear win for the games consoles but by a lower factor (maybe between 2 and 5); but the games consoles would be a whole lot cheaper.
Do ewe know how access to the console op system is made then ?
Quote from: neilep on 01/09/2010 12:39:22Do ewe know how access to the console op system is made then ?No, but the information has to be available to people developing games for example. These people are often independent from the particular console manufacturer and I expect that a research program, like the one you cite, would not be subject to any particular restrictions as it is good advertising for Sony, Microsoft or whoever.
As reported by Wired Magazine on October 17, 2007, an interesting application of using PlayStation 3 in a cluster configuration was implemented by Astrophysicist Dr. Gaurav Khanna, from the Physics department of University of Massachusetts Dartmouth, who replaced time used on supercomputers with a cluster of eight PlayStation 3s.[51] Subsequently, the next generation of this machine, now called the PlayStation 3 Gravity Grid, uses a network of 16 machines, and exploits the Cell processor for the intended application which is binary black hole coalescence using perturbation theory.[52][53] The Cell processor version used by the PlayStation 3 has a main CPU and 6 floating-point vector processors, giving the Gravity Grid machine a net of 16 general-purpose processors and 96 vector processors. The machine has a one-time cost of over $9,000 to build and is adequate for black-hole simulations which would otherwise cost $6,000 per run on a conventional supercomputer. The black hole calculations are not memory-intensive and are highly localizable, and so are well-suited to this architecture.
QuoteAs reported by Wired Magazine on October 17, 2007, an interesting application of using PlayStation 3 in a cluster configuration was implemented by Astrophysicist Dr. Gaurav Khanna, from the Physics department of University of Massachusetts Dartmouth, who replaced time used on supercomputers with a cluster of eight PlayStation 3s.[51] Subsequently, the next generation of this machine, now called the PlayStation 3 Gravity Grid, uses a network of 16 machines, and exploits the Cell processor for the intended application which is binary black hole coalescence using perturbation theory.[52][53] The Cell processor version used by the PlayStation 3 has a main CPU and 6 floating-point vector processors, giving the Gravity Grid machine a net of 16 general-purpose processors and 96 vector processors. The machine has a one-time cost of over $9,000 to build and is adequate for black-hole simulations which would otherwise cost $6,000 per run on a conventional supercomputer. The black hole calculations are not memory-intensive and are highly localizable, and so are well-suited to this architecture.from halfway down http://en.wikipedia.org/wiki/Cell_(microprocessor) this page on the cell processor developed for the PS3
As graham.d says, the Cell chip used in the PS3 has multiple processing units.In addition to a general purpose processor (like the more common x86 cpus found in typical PCs and servers, and capable running an OS and ordinary applications) the Cell 8i chip also incorporates 8 sub-processors known as SPEs (Synergistic Processing Elements). These SPEs are not general purpose processors (as they omit much of the general purpose logic required to run OSs or applications) but can be thought of as specialised number-crunching processors, tuned to produce the highest performance in just this field and omitting much of the logic hardware needed to run ordinary operating systems and applications. As a result, the eight SPEs in a Cell 8i can turn in a performance of up to 102.4 double-precision GFLOPS.This is about the same double-precision FLOP performance as a six core Intel Core i7 extreme (the top and most expensive model), which has a theoretical peak performance of 107.55 double-precision GFLOPS.The big difference though, is that with the Cell chip you have a single general purpose cpu scheduling work to eight SPEs whereas with the Core i7 the problem has to be distributed to each of the six general purpose cpus before it can actually be executed on an FPU.For numerically intensive problems this amounts to quite a large overhead, not only because of the added communications necessary between the individual cores, but also because the general purpose parts of the typical x86 multi-core processor are actually doing very little work, and only the FPU is being highly utilised.The pluses with using a Cell architecture then, are lower comms overhead (and delays resulting therefrom), lower hardware requirements (because each SPE doesn't need all the h/w logic to run general purpose s/w) and lower energy requirements (because you're not running additional general purpose cores that aren't being used).Recent developments in HPC (High Performance Computing) have lead to graphics cards being used for numerically intensive stuff as these can be viewed as having up to 1000 FPUs and can give up to ~1 TFLOP, but they're currently primarily single-precision devices and their double precision is typically much less than half (and sometimes only ~1/10) of their single precision performance.