Editing Computational fluid dynamics (section)

=== CPU versus GPU ===
Traditionally, CFD simulations are performed on CPUs.<ref>{{cite arXiv |last1=Lao |first1=Shandong |last2=Holt |first2=Aaron |last3=Vaidhynathan |first3=Deepthi |last4=Sitaraman |first4=Hariswaran |last5=Hrenya |first5=Christine M. |last6=Hauser |first6=Thomas |title=Performance comparison of CFD-DEM solver MFiX-Exa, on GPUs and CPUs |date=2021 |class=cs.DC |eprint=2108.08821 }}</ref>

In a more recent trend, simulations are also performed on GPUs. These typically contain slower but more processors. For CFD algorithms that feature good parallelism performance (i.e. good speed-up by adding more cores) this can greatly reduce simulation times. Fluid-implicit particle<ref>{{cite journal |last1=Wu |first1=Kui |last2=Truong |first2=Nghia |last3=Yuksel |first3=Cem |last4=Hoetzlein |first4=Rama |title=Fast Fluid Simulations with Sparse Volumes on the GPU |journal=Computer Graphics Forum |date=May 2018 |volume=37 |issue=2 |pages=157–167 |doi=10.1111/cgf.13350 |s2cid=43945038 }}</ref> and lattice-Boltzmann methods<ref>{{Cite web|url=http://www.nvidia.com/content/intersect-360-HPC-application-support.pdf|title=Intersect 360 HPC application Support}}</ref> are typical examples of codes that scale well on GPUs.