Hi Folks,

My apologies if I've missed the boat but I couldn't see a mention of this.

As reported by the BBC, Folding@Home are now in the Guiness Book of Records with an estimated Petaflop processing power distributed computing project. It looks like a majority of the estimated processing power comes from their PS3 client. Article Here

Anyone know how to recode the Seti client for a PS3??

I reckon we should consider it a challenge now :-)

Happy Crunching!

Keith

---

Hi Folks,

My apologies if I've missed the boat but I couldn't see a mention of this.

As reported by the BBC, Folding@Home are now in the Guiness Book of Records with an estimated Petaflop processing power distributed computing project. It looks like a majority of the estimated processing power comes from their PS3 client. Article Here

Anyone know how to recode the Seti client for a PS3??

I reckon we should consider it a challenge now :-)

Happy Crunching!

Keith

you can run Yellow Dog Linux on a PS3. should work for running seti@home.
but u can only use 6 of the 7 cells when running YDL on PS3 because the PS3 OS is still running under the YDL.

you can run Yellow Dog Linux on a PS3. should work for running seti@home.
but u can only use 6 of the 7 cells when running YDL on PS3 because the PS3 OS is still running under the YDL.

The Cell processor used in the PS3 has one PPE (a PPC G4, I believe) and six SPEs. Linux runs only on the PPE, and if you do a plain PPC/linux version of the application, it too will run only on the PPE. The performance is equivalent to a G4 mac. The application would need to be re-written to use the SPEs, which is where the real performance is.

For example, PS3GRID ran some comparison tests. The PS3 (using the SPEs) is 16x faster than an AMD Opteron 146. Credits are awarded accordingly, and are in the 3000-4000/day range for a single PS3.

More information can be read here:

http://www.ps3grid.net/PS3GRID/technology.php

---

Dublin, California
Team: SETI.USA

you can run Yellow Dog Linux on a PS3. should work for running seti@home.
but u can only use 6 of the 7 cells when running YDL on PS3 because the PS3 OS is still running under the YDL.

The Cell processor used in the PS3 has one PPE (a PPC G4, I believe) and six SPEs. Linux runs only on the PPE, and if you do a plain PPC/linux version of the application, it too will run only on the PPE. The performance is equivalent to a G4 mac. The application would need to be re-written to use the SPEs, which is where the real performance is.

For example, PS3GRID ran some comparison tests. The PS3 (using the SPEs) is 16x faster than an AMD Opteron 146. Credits are awarded accordingly, and are in the 3000-4000/day range for a single PS3.

More information can be read here:

http://www.ps3grid.net/PS3GRID/technology.php

i have the boinc_client_ps3grid.gz file. i just need to find the time too DL (YDL) but thx. i belive i read somewhere that the cell processor has 8 SPE's but the PS3 will only use 6 but im not sure if that is true or not, so don't quote me on that.

the PS3 will only use 6 but im not sure if that is true or not, so don't
quote me on that.

Correct. Cell processors are designed with 8 SPEs.
Those that roll off production flawlessly get used in blade servers such as QS20.
Those that have defects in one of the SPEs get used in the PS3.
It keeps one SPE for itself, leaving 6 available for games/apps.

Processors will 2 or more defective SPEs can still be used, and are much cheaper.
If you have a few cheap 5 or 6 SPE chips, you could connect them together and make a 20 SPE behemoth...

Defects in silicon crystals occur randomly and are a fact of life.
One has to hope that they don't occur where a processor is etched,
so smaller processors => more working processors => cheaper processors
big processors => smaller chance of success => fewer => expensive.

So you can play with these criteria and make big expensive rare processors or small cheap and common processors.

So, if you take the 65nm design of the Core 2 and shrink to 45nm as the Penryn is doing then you will make many more processors that are faster and cheaper, more in demand and sell for more money.

Then once you're getting familiar with 45nm technology you can experiment and increase the size of the die and add more transistors until the yield balances out between performance capability and cost.

As the money rolls in, you do the same for 32nm and then 22nm and then have to think of something else because managing electrons becomes as difficult as herding cats! Ask msattler!

Perhaps the next advance will be to use the new tiny transistors spaced a bit further apart to reduce heat and then build up into 3 dimensional cubes.