I agree with Movieman.
Some articles come immediately into my mind when I read about GPU folding/crunching.
http://www.theinquirer.net/inquirer/news/751/1045751/nvidia-55nm-parts-bad (scroll down for more links)
They are just not desinged for long-term full load usage.

I don't know how valid it is, but I can hear the users already screaming "YOU DESTROYED MY LAPTOP".

Whenever GPU usage comes to WCG I definitely want an option to opt-in/-out.

Cheers.

All makers have bad parts[link], I am curious as to failures related to use 24/7. I have posted a thread two different places to try and accumulate user information.

I am sure cuda work would be "opt-in" for WCG.

There are three projects using cuda to crunch[FAH, SETI, an GPUgrid] with another being close to starting[Einstein].

I don't see these projects creating work that would inherently damage equipment, that would be bad for the projects image[SETI is the grandaddy of Boinc].

This is new and there will be those against it just to be contrary without any factual basis.

@ lavaflow; I don't think didactylos coding abilities are in question here, but I would think that SETI and the other projects using cuda have just as capable coders.

mike

Hello Bill,
That's good news but then I'll have to spend lots of money on vid cards instead of new systems! LOLOL..
Out looking for a dual socket 1366 board now but they are scarce since they don't release for another month or two..
Think a 50,000 point a day machine doing 16 WU at a time..
(Giggles like a schoolgirl)

Thanks Bill! That is good news and a reasonable plan.

The following are my experiences and opinions so I hope neither side gets their knickers in a knot. I agree with both sides; it is not stable and it is the wave of the future and worth pursuing.

I have been crunching at GPUGrid since Sept ’08 and have crunched almost a million credits of work there. It has been both a frustrating and learning experience. The technology is not stable. The current cycle is; day 1 the project changes the science, day 2 nVidia changes the driver, day 3 BOINC changes the client and on day 4 ... goto day 1. It is still far from being “plug and crunch”. However, when all the stars align and all the technologies are working, you can crunch an incredible amount of work.

Those that understand the technologies involved and like to tweak their machines will have no problems. My concern over an early roll out here is for the casual cruncher. If one does not understand how their PC works, how BOINC works, how vid cards and drivers work and have no experience clocking, the GPU experience could be a disaster for them.

The vid cards make a big difference too. At GPUGrid the return times are short so a 8800GT is about the lowest end card that can return a wu in time. I can only speak for the 8 and 9 series cards but they consume large amounts of power, produce huge amounts of heat, are harder to cool than a pc and have a shorter time to failure than cpu’s. Factory OC’ed cards (such as GTC, GTX, GTX+) can be problematic too. I can’t afford ... or power ... the high-end cards so I can’t speak to them.

I do not follow the Folding and SETI threads so I would be interested in hearing in how they are doing. I do plan to GPU at Einstein when the roll out their GPU app.

If you like playing with technology, have patience, want to crunch a lot of work and don’t mind turning your PC into a 500W space heater, then GPU is for YOU.

For what it is worth ... that’s my experience so far.

Cheers!

Great feedback ! I agree with you. There are a lot of frustation behind the scenes, but there are compensations too.

Wonder read from bbover this news. I will wait and believe that WCG are in the right way to evolve on that subject.