Your GPU folding configuration

[drizzt81]

So, did we kiss and make up or am I going to have to kill this thread?

*kiss kiss*

My point is if Stanford people think that 30x the work is being done then some valuation in the point system should reflect that. I'll do the most work I can afford and that will be based on points. It is up to Stanford to balance the points based on the best effect for science.

very rational choice. You'd be my favorite customer if I sold something.

 

[7im]

My point is if Stanford people think that 30x the work is being done then some valuation in the point system should reflect that. I'll do the most work I can afford and that will be based on points. It is up to Stanford to balance the points based on the best effect for science.

The ATI FAQ tries to explain why 30x the science does not equal 30x the points. http://folding.stanford.edu/FAQ-ATI.html Note the history section. It talks about the original Gromacs core being 10x faster than the Tinker core, but because the Gromacs were limited in the simulations they could do compared to the Tinkers, the points were made equal (at least that's how I read it).

The same then lends itself to the GPU client, and how it is 30x faster than the CPU Gromacs core, but again, is not able to do many of the calculations that the CPU client can do. GPUs require special new work units, and isn't able to fold Gromacs, DGromacs, GBGroamacs, Gromacs33, Ambers, QMDs, Tinkers, etc. So then it's value looks to be somewhat less than 30x the CPU client.

The beta GPU client is only a week old. Beta testing is more about fixing the software, not about fixing the points. ATI will soon have better drivers, and Stanford will have improved the client, core, and reduced how many CPU cycles are needed to feed data the GPU client. They will also support more models of GPUs soon, as well as multiple GPUs. All this will make the GPU client more competative. And if the improvements are slow to develop, they can always normalize the points to 5x or more to do the job. Please give Stanford a fair shake, and a chance to improve the GPU client. You would be doing you and your team a dis-service to write off the GPU client too soon.