HPC modeling
Posted: Fri Sep 11, 2009 10:05 pm
I ran across this by accident today: arxiv.org/pdf/0908.4362
"The performance results show that CUDA is an attractive parallel computing environment for MHD simulations." This is similar to the theory we need to model for Polywell, and can be taken as proof of principle.
I also priced a 4 Terraflop unit with 48 GB of ram from Amax. It came in at $12k. I am not finding a whole lot of software support yet for the kind of work I'd like to do with PDE's, so you can't just buy stuff off the shelf yet. But I don't think it'll be too much longer.
Kind of stunning to me. 10 years ago that kind of processing power ran better than $20 million. And it took and army of programmers to make it go.
I don't thing processing power is going to be a problem any more!
"The performance results show that CUDA is an attractive parallel computing environment for MHD simulations." This is similar to the theory we need to model for Polywell, and can be taken as proof of principle.
I also priced a 4 Terraflop unit with 48 GB of ram from Amax. It came in at $12k. I am not finding a whole lot of software support yet for the kind of work I'd like to do with PDE's, so you can't just buy stuff off the shelf yet. But I don't think it'll be too much longer.
Kind of stunning to me. 10 years ago that kind of processing power ran better than $20 million. And it took and army of programmers to make it go.
I don't thing processing power is going to be a problem any more!