: Parallel performance across large numbers of processors has been significantly tuned. This revision allows for more efficient scaling on clusters and multi-CPU workstations, reducing the computational bottleneck often found in large-scale DFT and post-Hartree-Fock jobs. Dynamic Task Allocation
: Raman Optical Activity (ROA) invariants for vibrational modes are now printed only when normal mode derivatives are specifically requested, rather than by default, reducing unnecessary output clutter. gaussian 16 revision c.01
Gaussian has been slowly integrating GPU support, and Revision C.01 expanded this significantly. : Parallel performance across large numbers of processors
| Cores | Speedup (Rev B.01) | Speedup (Rev C.01) | |-------|--------------------|--------------------| | 1 | 1.00 | 1.00 | | 8 | 5.90 | 6.40 | | 16 | 9.20 | 12.10 | | 32 | 12.50 | 20.30 | rather than by default