Large-scale memory requirements
Large-scale memory requirements
As proteins get bigger, what are the dependencies on the memory? Quadratic? Linear? Something between or worse? Can I balance RAM and swap area well or will the system "thrash"?
Re: Large-scale memory requirements
For large proteins the memory requirements are linear with the number of spins and the number of cores in the parallel calculation. The amount actually needed depends a lot on what is being computed. HSQC of a 30 kDa system would fit into 64 GB of RAM, but NOESY (where the full relaxation superoperator needs to be computed) might hit a terabyte, but the active "window" would still be about 50 GB, so a system with a solid state disk would swap the rest out without difficulty. The best way to find out is to get an Amazon EC2 cloud instance (they have 256GB systems with solid state disks at a few bucks an hour) and give it a try.