A new GPU-resident simulation mode is now available in NAMD. In addition to performing force calculations on GPU devices, the GPU-resident mode also performs numerical integration and rigid bond constraints on GPU, with the simulation data residing on the devices between time steps. By eliminating work performed on the CPU and the overhead from per-step transfer of data between CPU host and GPU device, this new mode achieves much higher utilization of GPU devices.

On a single high-end GPU, GPU-resident mode will provide a 2x or greater speedup over NAMD's traditional GPU-offload approach. GPU-resident mode is also capable of scaling across multiple tightly coupled GPUs on a single node, utilizing NVLink technology for NVIDIA devices or Infinity Fabric for AMD devices. Scaling for GPU-resident mode can provide even more efficiency than is possible for GPU-offload mode.