We have seen all sorts of outrageous (and conflicting) claims. It will be interesting to see what can really be done.

However, there are two major problems that prevents this being a panacea: first, not all grid projects can benefit from massive parallelism. They are already cut down into the smallest units that can run in parallel, and there is no benefit in splitting it up further because of memory requirements. Some of our current projects may fail this test, others may pass. Image filtering can run in parallel with shared memory, for example.

The second problem is variation in GPUs and the amount of effort it requires just to adapt one project to run on one GPU. Currently, there's not a snowball's chance in hell that WCG can do this for any of our current projects. But this may change: efforts are underway to abstract much of the architectural differences, and provide a uniform platform. Also, future projects may be delivered to WCG with science applications already designed to take advantage of this kind of processing.

Time will tell, but WCG are unlikely to jump the gun on this one.

Yes any technology advance means a lot of work... My grandfather always told me that nothing big comes lonely to us without big effort from us...

I think a 20x to 40x computing power is something serious enough to invest time into...




Nvidia has already annouced that this layer would be certainly available for their G80 chips ;-)



I think you are mistaken here in your analysis... you should not take this advance in technology information (and reality) made by standford folding project so in the style "entering by one ear and leaving by the other ear, and let see what happens"....

If WCG takes too much gap with other project and do not invest time in that new possibility... it can mean that a lot of people will quit this project because they donate their time but want their time being used at best...
So if with folding at home you know you can process 20x faster because of GPU processing technology... then you know you can process more complex proteins for the same time, then the quality of the project is better...

Just think about it

PS: i know for the moment there are not a lot of crunchers who have X1900 series... but things come fast today... so if you say it takes some time perhaps you should already invest time and work to it...

some foreign language sites discussed and figured it to take about 3 years before it will be more common / can go mainstream.... the x1900 cards run at 90C, suck 125 wh, put fans in hi gear and have to be throttled down from 3d to 2d to get it down into the 70's C's....and the CPU still loaded at 50%. Is it head first or feet first for a clients computer?

BTW, some of these cards seem to outdo the price of a PC & wonder what it does to the warranty if run continuously at 90C.

I think Sekerob is right. That is a a lot of heat/energy to dissipate. That can not be good for the rest of the components much less the chip itself, unless you have some fancy cooling system. His point about the energy usage (125W) is also a good one. Most of us do not exist on the bleeding edge of technology, so the added expense of both the card and its energy use are significant, besides the fact that most older systems are not compatible with the new highpower cards. At any rate, the proof of concept I have yet to see in any real world testing. Let us see some published results, weigh the costs and time of the additional programming against the perceived benefits of some potential(probably small) number of cards. Would the amount of work required justify the return ?? This may be a viable alternative at some time in the future, but probably not until there is some significant market penetration of the hardware (if the claims prove to be true). Beware of the hype.

Cheers

Joe

Read the information more carefully: certain operations can run 20x faster. However, that's not everything, and some things will run slower, bringing the average speed down to something much less impressive. I think they are actually claiming 4x, which is still useful.

But 125°? People are already complaining about overheating with normal CPUs. That sort of temperature effectively turns your computer into an oven.