Re: [ccp4bb] GPU computing

2010-06-07 Thread Nat Echols
Most of what Kevin says applies equally well to Phenix - refinement is a relatively complex series of calculations, much less inherently parallel than what VMD is using GPUs for. There are many other things we could do to speed up calculations that would probably be more useful and more effective.

Re: [ccp4bb] GPU computing

2010-06-07 Thread Kevin Cowtan
I looked at it and concluded that our FFTs are on the whole too short for it to be worthwhile, and a lot of calculations aren't FFT bound anyway. An awful lot of our stuff is simply not very slow. Also, GPU computing at the moment is crippled for many problems by the bandwidth bottleneck and m

[ccp4bb] GPU computing

2010-06-07 Thread Zhou, Tongqing (NIH/VRC) [E]
Hi All, I am wondering anyone is looking into GPU computing on structural refinement? I see VMD supports GPU computing already, how about other programs such as CCP4 and Phenix Thanks, Tongqing Tongqing Zhou, Ph.D. Staff Scientist Structural Biology Section Vaccine Research Center, NIAID