Distributed Computing on Next Gen Consoles
anonymous lion writes "Wired has a story on the need for Xbox 360 and PlayStation 3 to support distributed computing with a non-gaming purpose. The article goes on to discuss SETI@home, distributed.net, and Folding@Home." From the article: "The next generation of console gaming is going to see a huge increase in machine performance and overall computing power. Already planned for both the Xbox 360 and the PlayStation 3 are multiple 3.2-GHz PowerPC processors capable of handling advanced gaming and graphics simulations, along with out-of-the-box internet capabilities such as Xbox Live Silver. With all that horsepower in a machine that is used for only a fraction of a day, we should offer gamers a chance to put these unused resources to good use."
At least with current platforms architectures. The author seems to do plenty of research on current distributed computing projects, but does none on how the consoles perform.
I know that SETI@home has been ported and tested at least on the XBox, and it performs miserably. These console gaming systems are designed to play games, not do radio signal analysis or other scientific calculation. For example, there's little need for fast memory writing when you're mostly reading textures from RAM, but there's an extreme need when you do millions of in-place Fourier transforms. Unless Microsoft and Sony change their architectures for some inexplicable reason, I can't imagine future architectures would perform much better.
This article smacks of ignorance on the part of the author, who clearly did no research into the actual performance of consoles in regard to standard scientific computing.
There are many houses that do not have access to natural gas, and so heat with electricity.
What planet do you live on?
The unofficial
really? and what components does your computer have outside of disk drives doesn't your television have? people turn on/off their TV tons of times a day, it's the same electronics component. Please!
Shed the myth! Hard disks for the most part are now better designed than back in the days, systems boot very fast, there is no need to keep your computer on if you will not be using it for a long time.
------ Curiosity killed the cat. {satisfaction brought it back | it didn't die ignorant | lack of it is killing mankind
Perhaps IBM doesn't just want to sell chips to these people. Perhaps it has a reason for selling the PC division to Lenovo. Perhaps it sees an opportunity to create a business architecture in which the virtual business world runs on the server farm, while the graphics and sound capability of the very cheap clients delivers a superior user experience that makes users happy not to have a "PC" on their desk. Meanwhile the data mining and compute-intensive activities are farmed out to those clients while they aren't being used. Fault tolerant. Cheap to extend. And round objects to Microsoft.
Panurge has posted for the last time. Thanks for the positive moderations.
I've never been a big fan of game consoles for that reason. I modchipped a few X-Box's for friends and played with XBMC a bit, but it was very much a toy in my eyes too. It also seemed like Microsoft was fighting our attempts to turn it into a PC at every turn. This next generation is going to be different from the looks of things though. I found this quote particularly interesting in that article I linked to:
That very much sounds like multitasking to me, and multitasking at a new level similar to Xen. Very interesting times are ahead.
This is exactly my point. Individually the processor may perform well, but when it's placed in the actual system, perform will undoubtedly drop. Right now, I'm doing performance tests on FFTs performed on GPUs (graphics cards). Theoretically, these should perform at the same "incredible" speed as the Cell processor (10 Gflops or better), but in reality bandwidth and cache constricts performance to half a Gflop.
You'll also note they say nothing about the system they tested on. Was it a PS3? I bet you it wasn't. Let's wait and see the actual performance of the PS3 before we get excited from in house tests done on an unspecified system by a company that is eager to boost impressions of its new chip.
A hint for you:
A 650W PSU doesnt draw 650W if its only under 100W load.
So my 350W enermaxx is perfectly happy drawing 50W when the pc is idle. Its efficiency may be lower, but thats not THAT huge of a difference.
AND PLEASE, learn your units. Saying "drawns more voltage then needed" really makes you look stupid.
If you put a hd, a 50W cpu, 512MB high speed ram and a GPU in a console, it doesnt magically NOT use that much less energy than in a PC.
And using DC on a console defeats to total purpose: using idle cycles, mostly on little used computers.
If you turn the computer on to run the DC client, you are doing something wrong (and if you BUY stuff to run DC clients, please die)
HI O WISE PRINCE. WHT TOOK U SO DAM LONG?
The Cell processor (PS3) is made for those applications. At the Power.org convention in Barcelona, IBM presented a programming example of large FFTs on Cell. It turned out, that large FFT calculations are about 100 times faster than on a Xeon 3.2 GHz processor.
Keep in mind, that this presentation was held in front of super computer professionals and its not that easy to trick them.
MS doesn't have PowerPC CPUs, those cores are in-order PPC-based chips for embedded systems (same goes for the Cell's all-purpose core). They'd perform awfully in PCs. Chris Hecker claimed the chip is 3-10 times slower than a comparably clocked PPC in the Burning Down The House session at GDC. A P4 at a similar clock speed would kick the shit along with the intestines out of that "PPE".
It's actually not known what kind of chip Nintendo will use, not the clock speed, not the features, nothing except the codename "Broadway".
Justice is the sheep getting arrested while an impartial judge declares the vote void.