PageSourceSearch

https://www.bsc.es/supportkc/assets/js/d668726e.ebd1d431.js

js bsc.es collected 2026-09-24 08:43:13 UTC 67,443 bytes, 1 lines download raw bytes

1"use strict";(self.webpackChunksupportkc_new=self.webpackChunksupportkc_new||[]).push([[1693],{3905:(t,a,n)=>{n.d(a,{Zo:()=>u,kt:()=>g});var e=n(67294);function l(t,a,n){return a in t?Object.defineProperty(t,a,{value:n,enumerable:!0,configurable:!0,writable:!0}):t[a]=n,t}function r(t,a){var n=Object.keys(t);if(Object.getOwnPropertySymbols){var e=Object.getOwnPropertySymbols(t);a&&(e=e.filter((function(a){return Object.getOwnPropertyDescriptor(t,a).enumerable}))),n.push.apply(n,e)}return n}function i(t){for(var a=1;a<arguments.length;a++){var n=null!=arguments[a]?arguments[a]:{};a%2?r(Object(n),!0).forEach((function(a){l(t,a,n[a])})):Object.getOwnPropertyDescriptors?Object.defineProperties(t,Object.getOwnPropertyDescriptors(n)):r(Object(n)).forEach((function(a){Object.defineProperty(t,a,Object.getOwnPropertyDescriptor(n,a))}))}return t}function m(t,a){if(null==t)return{};var n,e,l=function(t,a){if(null==t)return{};var n,e,l={},r=Object.keys(t);for(e=0;e<r.length;e++)n=r[e],a.indexOf(n)>=0||(l[n]=t[n]);return l}(t,a);if(Object.getOwnPropertySymbols){var r=Object.getOwnPropertySymbols(t);for(e=0;e<r.length;e++)n=r[e],a.indexOf(n)>=0||Object.prototype.propertyIsEnumerable.call(t,n)&&(l[n]=t[n])}return l}var p=e.createContext({}),d=function(t){var a=e.useContext(p),n=a;return t&&(n="function"==typeof t?t(a):i(i({},a),t)),n},u=function(t){var a=d(t.components);return e.createElement(p.Provider,{value:a},t.children)},k="mdxType",N={inlineCode:"code",wrapper:function(t){var a=t.children;return e.createElement(e.Fragment,{},a)}},o=e.forwardRef((function(t,a){var n=t.components,l=t.mdxType,r=t.originalType,p=t.parentName,u=m(t,["components","mdxType","originalType","parentName"]),k=d(n),o=l,g=k["".concat(p,".").concat(o)]||k[o]||N[o]||r;return n?e.createElement(g,i(i({ref:a},u),{},{components:n})):e.createElement(g,i({ref:a},u))}));function g(t,a){var n=arguments,l=a&&a.mdxType;if("string"==typeof t||l){var r=n.length,i=new Array(r);i[0]=o;var m={};for(var p in a)hasOwnProperty.call(a,p)&&(m[p]=a[p]);m.originalType=t,m[k]="string"==typeof t?t:l,i[1]=m;for(var d=2;d<r;d++)i[d]=n[d];return e.createElement.apply(null,i)}return e.createElement.apply(null,n)}o.displayName="MDXCreateElement"},5159:(t,a,n)=>{n.r(a),n.d(a,{assets:()=>p,contentTitle:()=>i,default:()=>N,frontMatter:()=>r,metadata:()=>m,toc:()=>d});var e=n(87462),l=(n(67294),n(3905));const r={slug:"fhiaims-report",title:"GPU-Acceleration of FHI-Aims Simulations with Various Geometry Inputs on MareNostrum5",authors:["thall"],tags:["report","performance"]},i=void 0,m={permalink:"/supportkc/tech-reports/fhiaims-report",source:"@site/tech-reports/2025-04-16-fhiaims-report/FHIAims.md",title:"GPU-Acceleration of FHI-Aims Simulations with Various Geometry Inputs on MareNostrum5",description:"Background",date:"2025-04-16T00:00:00.000Z",formattedDate:"April 16, 2025",tags:[{label:"report",permalink:"/supportkc/tech-reports/tags/report"},{label:"performance",permalink:"/supportkc/tech-reports/tags/performance"}],readingTime:14.635,hasTruncateMarker:!1,authors:[{name:"Trevor Hall",title:"Second Level HPC Support",url:"https://www.bsc.es/hall-trevor-alan",imageURL:"https://www.bsc.es/sites/default/files/public/personal_photo/thall.jpeg",key:"thall"}],frontMatter:{slug:"fhiaims-report",title:"GPU-Acceleration of FHI-Aims Simulations with Various Geometry Inputs on MareNostrum5",authors:["thall"],tags:["report","performance"]},prevItem:{title:"SIESTA runs on MareNostrum5 GPP and ACC partitions",permalink:"/supportkc/tech-reports/siesta-report"},nextItem:{title:"VASP 6 performance on CTE-POWER",permalink:"/supportkc/tech-reports/vasp6-report"}},p={authorsImageUrls:[void 0]},d=[{value:"Background",id:"background",level:2},{value:"FHI-Aims Compilation Notes",id:"fhi-aims-compilation-notes",level:3},{value:"GPU acceleration (total time-to-solution) in literature:",id:"gpu-acceleration-total-time-to-solution-in-literature",level:3},{value:"Simulation Results",id:"simulation-results",level:2},{value:"Data",id:"data",level:3},{value:"108-atom Si molecule (Si)",id:"108-atom-si-molecule-si",level:4}
1,{value:"CPU-Only",id:"cpu-only",level:5},{value:"GPU-Accelerated",id:"gpu-accelerated",level:5},{value:"Acceleration Factors",id:"acceleration-factors",level:5},{value:"Summary",id:"summary",level:5},{value:"146-atom complex molecule with an initial movement (cplx)",id:"146-atom-complex-molecule-with-an-initial-movement-cplx",level:4},{value:"CPU-Only",id:"cpu-only-1",level:5},{value:"GPU-Accelerated",id:"gpu-accelerated-1",level:5},{value:"Acceleration Factors",id:"acceleration-factors-1",level:5},{value:"Summary",id:"summary-1",level:5},{value:"220-atom complex molecule with velocity (Ac-Lys-Ala19-H)",id:"220-atom-complex-molecule-with-velocity-ac-lys-ala19-h",level:4},{value:"CPU-Only",id:"cpu-only-2",level:5},{value:"GPU-Accelerated",id:"gpu-accelerated-2",level:5},{value:"Acceleration Factors",id:"acceleration-factors-2",level:5},{value:"Summary",id:"summary-2",level:5},{value:"Best-case total time-to-solution for each system",id:"best-case-total-time-to-solution-for-each-system",level:3},{value:"Best-case acceleration (total time-to-solution) for each system",id:"best-case-acceleration-total-time-to-solution-for-each-system",level:3},{value:"Acceleration (total time-to-solution) in literature, for comparison",id:"acceleration-total-time-to-solution-in-literature-for-comparison",level:4},{value:"Conclusions and Observations",id:"conclusions-and-observations",level:2},{value:"Reference Materials",id:"reference-materials",level:2}],u={toc:d},k="wrapper";function N(t){let{components:a,...r}=t;return(0,l.kt)(k,(0,e.Z)({},u,r,{components:a,mdxType:"MDXLayout"}),(0,l.kt)("h2",{id:"background"},"Background"),(0,l.kt)("p",null,"FHI-Aims is a flexible, widely-used electron structure program allowing for simple and complex molecular simulation. FHI-Aims supports GPU-acceleration for many (though not all) operations. The MareNostrum 5 software stack provides implementations of FHI-Aims for both ",(0,l.kt)("a",{parentName:"p",href:"https://www.bsc.es/supportkc/docs/MareNostrum5/overview#marenostrum-5-gpp-general-purpose-partition"},"GPP")," and ",(0,l.kt)("a",{parentName:"p",href:"https://www.bsc.es/supportkc/docs/MareNostrum5/overview#marenostrum-5-acc-accelerated-partition"},"ACC")," nodes, allowing for either CPU-landlocked simulations or acceleration with H100 GPUs, when relevant. Anecdotal observation of various FHI-Aims systems on MN5 displayed inconsistency in the effectiveness of GPU acceleration and the additional performance benefits therein. In comparison with the optimized effects of acceleration in ",(0,l.kt)("a",{parentName:"p",href:"https://arxiv.org/pdf/1912.06636#figure.caption.1"},"literature"),", the acceleration we observed on MN5 was lacking."),(0,l.kt)("h3",{id:"fhi-aims-compilation-notes"},"FHI-Aims Compilation Notes"),(0,l.kt)("p",null,"On MareNostrum5, FHI-Aims is compiled for use with NVIDIA Hopper H100 GPUs (sm_90 architecture) present on the system. FHI-AIMS is build with the NVIDIA-HPC-SDK compiler module, compiled with OpenMPI and Intel MKL Blacs and Scalapack routines. Complete compilation instructions can be found on the BSC GPFS file system at ",(0,l.kt)("inlineCode",{parentName:"p"},"/apps/ACC/FHI-AIMS/SRC/<version>/README.BSC"),"."),(0,l.kt)("h3",{id:"gpu-acceleration-total-time-to-solution-in-literature"},"GPU acceleration (total time-to-solution) in ",(0,l.kt)("a",{parentName:"h3",href:"https://arxiv.org/pdf/1912.06636#figure.caption.1"},"literature"),":"),(0,l.kt)("table",null,(0,l.kt)("thead",{parentName:"table"},(0,l.kt)("tr",{parentName:"thead"},(0,l.kt)("th",{parentName:"tr",align:null},"CPU"),(0,l.kt)("th",{parentName:"tr",align:null},"GPU"),(0,l.kt)("th",{parentName:"tr",align:null},"Acceleration Factor (Compared to CPU-only)"))),(0,l.kt)("tbody",{parentName:"table"},(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16 Cores Ivy Bridge"),(0,l.kt)("td",{parentName:"tr",align:null},"1 Pascal GPGPU"),(0,l.kt)("td",{parentName:"tr",align:null},"3.8x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32 Cores Haswell"),(0,l.kt)("td",{parentName:"tr",align:null},"4 Pascal GPGPU"),(0,l.kt)("td",{parentName:"tr",align:null},"3.7x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null}
1,"20 Cores Skylake"),(0,l.kt)("td",{parentName:"tr",align:null},"4 Volta GPGPU"),(0,l.kt)("td",{parentName:"tr",align:null},"4.0x")))),(0,l.kt)("p",null,"Is is important to note that the acceleration reported here is for a 54-atom Si supercell, diamond structure, without any mention of initial movement, velocity, or other system dynamics."),(0,l.kt)("p",null,"Because reproducible input are not provided in the referenced literature, tests are performed with three input geometries:"),(0,l.kt)("ul",null,(0,l.kt)("li",{parentName:"ul"},"108-atom Si molecule (Si)"),(0,l.kt)("li",{parentName:"ul"},"146-atom complex molecule with an initial movement (cplx)"),(0,l.kt)("li",{parentName:"ul"},"220-atom complex molecule with velocity (Ac-Lys-Ala19-H)")),(0,l.kt)("p",null,"All simulations are performed with the 'pbe' exchange-correlation approach. Further specifics of each simulation can be found in the attached reference materials. Additionally, simulations are run across three different FHI-Aims versions - 221103, 231212, and 240507. Lastly, simulations are run to imitate the hardware configuration from the referenced literature - 16 cores + 1 GPU, 32 Cores + 4 GPUs, 20 Cores + 4 GPUs with MN5's Intel Xeon Platinum 8460Y+ and NVIDIA Hopper H100 architecture. Additionally, we have performed 80 cores + 4 GPU experiments to demonstrate MN5's full-node performance. All accelerated runs were performed with NVIDIA Multi-Process Service (MPS), which we observed further improved performance for accelerated runs."),(0,l.kt)("h2",{id:"simulation-results"},"Simulation Results"),(0,l.kt)("p",null,"The results of the simulations are provided in total time-to-solution, total time for calculation of Density and Force Components (DFC), and total time for integrations (of which the latter two are expected to benefit the most from GPU acceleration). We have also provided a measure of the relative performance of GPU-accelerated vs. CPU runs, reported as the acceleration factor. All measurements are in seconds. The results are as follows:"),(0,l.kt)("h3",{id:"data"},"Data"),(0,l.kt)("h4",{id:"108-atom-si-molecule-si"},"108-atom Si molecule (Si)"),(0,l.kt)("h5",{id:"cpu-only"},"CPU-Only"),(0,l.kt)("table",null,(0,l.kt)("thead",{parentName:"table"},(0,l.kt)("tr",{parentName:"thead"},(0,l.kt)("th",{parentName:"tr",align:null},"Ranks"),(0,l.kt)("th",{parentName:"tr",align:null},"Version"),(0,l.kt)("th",{parentName:"tr",align:null},"Total Runtime"),(0,l.kt)("th",{parentName:"tr",align:null},"DFC Runtime"),(0,l.kt)("th",{parentName:"tr",align:null},"Integration Runtime"))),(0,l.kt)("tbody",{parentName:"table"},(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"265.517"),(0,l.kt)("td",{parentName:"tr",align:null},"79.281"),(0,l.kt)("td",{parentName:"tr",align:null},"17.336")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"183.274"),(0,l.kt)("td",{parentName:"tr",align:null},"72.033"),(0,l.kt)("td",{parentName:"tr",align:null},"17.295")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"183.548"),(0,l.kt)("td",{parentName:"tr",align:null},"71.74"),(0,l.kt)("td",{parentName:"tr",align:null},"17.234")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"214.719"),(0,l.kt)("td",{parentName:"tr",align:null},"63.222"),(0,l.kt)("td",{parentName:"tr",align:null},"13.785")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"149.992"),(0,l.kt)("td",{parentName:"tr",align:null},"58.326"),(0,l.kt)("td",{parentName:"tr",align:null},"13.66")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"151.617"),(0,l.kt)("td",{parentName:"tr",align:null},"58.538"),(0,l.kt)("td",{parentName:"tr",align:null},"13.784")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"152.798"),(0,l.kt)("td",{parentName:"tr",align:null},"45.779"),(0,l.kt)("td",{parentName:"tr",align:null},"9.099")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"108.913"),(0,l.kt)("td",{parentName:"tr",align:null},"41.82"),(0,l.kt)("td",{parentName:"tr",align:null},"9.422")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"111.099"),(0,l.kt)("td",{parentName:"tr",align:null},"41.653"),(0,l.kt)("td",{parentName:"tr",align:null},"9.326")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"105.503"),(0,l.kt)("td",{parentName:"tr",align:null},"30.02"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"5.869")))),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"83.245"),(0,l.kt)("td",{parentName:"tr",align:null},"28.05"),(0,l.kt)("td",{parentName:"tr",align:null},"6.026")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"82.64"))),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"27.958"))),(0,l.kt)("td",{parentName:"tr",align:null},"5.998")))),(0,l.kt)("h5",{id:"gpu-accelerated"},"GPU-Accelerated"),(0,l.kt)("table",null,(0,l.kt)("thead",{parentName:"table"},(0,l.kt)("tr",{parentName:"thead"},(0,l.kt)("th",{parentName:"tr",align:null},"Ranks"),(0,l.kt)("th",{parentName:"tr",align:null},"GPU"),(0,l.kt)("th",{parentName:"tr",align:null},"Version"),(0,l.kt)("th",{parentName:"tr",align:null},"Total Runtime"),(0,l.kt)("th",{parentName:"tr",align:null},"DFC Runtime"),(0,l.kt)("th",{parentName:"tr",align:null},"Integration Runtime"))),(0,l.kt)("tbody",{parentName:"table"},(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"220.871"),(0,l.kt)("td",{parentName:"tr",align:null},"37.415"),(0,l.kt)("td",{parentName:"tr",align:null},"11.338")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"143.676"),(0,l.kt)("td",{parentName:"tr",align:null},"36.975"),(0,l.kt)("td",{parentName:"tr",align:null},"11.06")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"144.001"),(0,l.kt)("td",{parentName:"tr",align:null},"37.047"),(0,l.kt)("td",{parentName:"tr",align:null},"11.36")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"173.028"),(0,l.kt)("td",{parentName:"tr",align:null},"24.039"),(0,l.kt)("td",{parentName:"tr",align:null},"8.738")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"111.212"),(0,l.kt)("td",{parentName:"tr",align:null},"22.869"),(0,l.kt)("td",{parentName:"tr",align:null},"8.639")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"112.494"),(0,l.kt)("td",{parentName:"tr",align:null},"23.033"),(0,l.kt)("td",{parentName:"tr",align:null},"8.691")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"158.18"),(0,l.kt)("td",{parentName:"tr",align:null},"42.733"),(0,l.kt)("td",{parentName:"tr",align:null},"7.031")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"119.848"),(0,l.kt)("td",{parentName:"tr",align:null},"42.967"),(0,l.kt)("td",{parentName:"tr",align:null},"7.061")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"118.324"),(0,l.kt)("td",{parentName:"tr",align:null},"42.79"),(0,l.kt)("td",{parentName:"tr",align:null},"7.158")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"129.134"),(0,l.kt)("td",{parentName:"tr",align:null},"19.263"),(0,l.kt)("td",{parentName:"tr",align:null},"6.111")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"90.826"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"18.941"))),(0,l.kt)("td",{parentName:"tr",align:null},"6.046")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"91.146"),(0,l.kt)("td",{parentName:"tr",align:null},"18.548"),(0,l.kt)("td",{parentName:"tr",align:null},"5.871")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"119.877"),(0,l.kt)("td",{parentName:"tr",align:null},"19.626"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"}
1,"4.549")))),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"102.367"),(0,l.kt)("td",{parentName:"tr",align:null},"20.208"),(0,l.kt)("td",{parentName:"tr",align:null},"4.738")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"100.706"))),(0,l.kt)("td",{parentName:"tr",align:null},"19.586"),(0,l.kt)("td",{parentName:"tr",align:null},"4.616")))),(0,l.kt)("h5",{id:"acceleration-factors"},"Acceleration Factors"),(0,l.kt)("table",null,(0,l.kt)("thead",{parentName:"table"},(0,l.kt)("tr",{parentName:"thead"},(0,l.kt)("th",{parentName:"tr",align:null},"Ranks"),(0,l.kt)("th",{parentName:"tr",align:null},"GPU"),(0,l.kt)("th",{parentName:"tr",align:null},"Version"),(0,l.kt)("th",{parentName:"tr",align:null},"Total Acceleration Factor"),(0,l.kt)("th",{parentName:"tr",align:null},"DFC Acceleration Factor"),(0,l.kt)("th",{parentName:"tr",align:null},"Integration Acceleration Factor"))),(0,l.kt)("tbody",{parentName:"table"},(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"1.202x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.119x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.529x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},"16")),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},"1")),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"1.276x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.948x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.564x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"1.275x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.936x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.517x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"1.241x"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"2.630x"))),(0,l.kt)("td",{parentName:"tr",align:null},"1.578x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"1.349x"))),(0,l.kt)("td",{parentName:"tr",align:null},"2.550x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.581x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"1.348x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.541x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.586x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"0.972x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.081x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.333x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"0.905x"),(0,l.kt)("td",{parentName:"tr",align:null},"0.980x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.387x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"0.935x"),(0,l.kt)("td",{parentName:"tr",align:null},"0.997x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.329x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"1.183x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.377x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.489x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"1.199x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.208x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.558x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"1.219x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.246x"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"1.588x")))),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"0.880x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.530x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.290x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"0.813x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.388x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.272x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"0.821x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.427x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.299x")))),(0,l.kt)("h5",{id:"summary"},"Summary"),(0,l.kt)("p",null,"GPU acceleration, by and large, had a negative effect on performance with the homogeneous Si system. Though we see acceleration in the expected components in almost all cases, the boost is not enough to make up for the costs and additional overhead associated with GPU inclusion. Even in the best-case, GPU acceleration offers relatively minor performance boosts compared to CPU-only. MareNostrum5 full-node performance is also worsened with the inclusion of GPU acceleration, so users with similar systems may be advised to rely on CPU-only architectures, taking note of observed logarithmic scaling (additional CPU-only scaling data is available at the ",(0,l.kt)("a",{parentName:"p",href:"https://fhi-aims-club.gitlab.io/tutorials/scaling-in-fhi-aims/"},"FHI-Aims GitLab"),")."),(0,l.kt)("h4",{id:"146-atom-complex-molecule-with-an-initial-movement-cplx"},"146-atom complex molecule with an initial movement (cplx)"),(0,l.kt)("h5",{id:"cpu-only-1"},"CPU-Only"),(0,l.kt)("table",null,(0,l.kt)("thead",{parentName:"table"},(0,l.kt)("tr",{parentName:"thead"},(0,l.kt)("th",{parentName:"tr",align:null},"Ranks"),(0,l.kt)("th",{parentName:"tr",align:null},"Version"),(0,l.kt)("th",{parentName:"tr",align:null},"Total Runtime"),(0,l.kt)("th",{parentName:"tr",align:null},"DFC Runtime"),(0,l.kt)("th",{parentName:"tr",align:null},"Integration Runtime"))),(0,l.kt)("tbody",{parentName:"table"},(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"1348.24"),(0,l.kt)("td",{parentName:"tr",align:null},"1051.561"),(0,l.kt)("td",{parentName:"tr",align:null},"93.703")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"1043.163"),(0,l.kt)("td",{parentName:"tr",align:null},"827.421"),(0,l.kt)("td",{parentName:"tr",align:null},"96.033")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"1051.617"),(0,l.kt)("td",{parentName:"tr",align:null},"835.97"),(0,l.kt)("td",{parentName:"tr",align:null},"95.721")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null}
1,"1133.52"),(0,l.kt)("td",{parentName:"tr",align:null},"896.293"),(0,l.kt)("td",{parentName:"tr",align:null},"78.616")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"889.877"),(0,l.kt)("td",{parentName:"tr",align:null},"709.668"),(0,l.kt)("td",{parentName:"tr",align:null},"80.186")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"892.903"),(0,l.kt)("td",{parentName:"tr",align:null},"713.524"),(0,l.kt)("td",{parentName:"tr",align:null},"80.78")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"866.599"),(0,l.kt)("td",{parentName:"tr",align:null},"704.786"),(0,l.kt)("td",{parentName:"tr",align:null},"57.198")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"868.453"),(0,l.kt)("td",{parentName:"tr",align:null},"706.075"),(0,l.kt)("td",{parentName:"tr",align:null},"56.966")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"661.991"),(0,l.kt)("td",{parentName:"tr",align:null},"537.485"),(0,l.kt)("td",{parentName:"tr",align:null},"58.443")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"595.636"),(0,l.kt)("td",{parentName:"tr",align:null},"496.553"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"35.059")))),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"424.845"))),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"345.198"))),(0,l.kt)("td",{parentName:"tr",align:null},"36.012")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"427.097"),(0,l.kt)("td",{parentName:"tr",align:null},"346.336"),(0,l.kt)("td",{parentName:"tr",align:null},"36.102")))),(0,l.kt)("h5",{id:"gpu-accelerated-1"},"GPU-Accelerated"),(0,l.kt)("table",null,(0,l.kt)("thead",{parentName:"table"},(0,l.kt)("tr",{parentName:"thead"},(0,l.kt)("th",{parentName:"tr",align:null},"Ranks"),(0,l.kt)("th",{parentName:"tr",align:null},"GPU"),(0,l.kt)("th",{parentName:"tr",align:null},"Version"),(0,l.kt)("th",{parentName:"tr",align:null},"Total Runtime"),(0,l.kt)("th",{parentName:"tr",align:null},"DFC Runtime"),(0,l.kt)("th",{parentName:"tr",align:null},"Integration Runtime"))),(0,l.kt)("tbody",{parentName:"table"},(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"511.647"),(0,l.kt)("td",{parentName:"tr",align:null},"263.821"),(0,l.kt)("td",{parentName:"tr",align:null},"44.689")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"335.247"),(0,l.kt)("td",{parentName:"tr",align:null},"167.417"),(0,l.kt)("td",{parentName:"tr",align:null},"46.909")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"336.044"),(0,l.kt)("td",{parentName:"tr",align:null},"167.337"),(0,l.kt)("td",{parentName:"tr",align:null},"47.193")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"429.532"),(0,l.kt)("td",{parentName:"tr",align:null},"235.713"),(0,l.kt)("td",{parentName:"tr",align:null},"34.427")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"234.617"),(0,l.kt)("td",{parentName:"tr",align:null},"97.518"),(0,l.kt)("td",{parentName:"tr",align:null},"36.808")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"232.654"),(0,l.kt)("td",{parentName:"tr",align:null},"96.949"),(0,l.kt)("td",{parentName:"tr",align:null},"36.749")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"380.658"),(0,l.kt)("td",{parentName:"tr",align:null}
1,"238.438"),(0,l.kt)("td",{parentName:"tr",align:null},"32.532")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"265.56"),(0,l.kt)("td",{parentName:"tr",align:null},"160.3"),(0,l.kt)("td",{parentName:"tr",align:null},"33.154")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"265.197"),(0,l.kt)("td",{parentName:"tr",align:null},"160.774"),(0,l.kt)("td",{parentName:"tr",align:null},"33.195")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"340.578"),(0,l.kt)("td",{parentName:"tr",align:null},"207.553"),(0,l.kt)("td",{parentName:"tr",align:null},"24.85")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"342.007"),(0,l.kt)("td",{parentName:"tr",align:null},"208.484"),(0,l.kt)("td",{parentName:"tr",align:null},"25.084")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"169.011"),(0,l.kt)("td",{parentName:"tr",align:null},"74.207"),(0,l.kt)("td",{parentName:"tr",align:null},"25.957")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"307.123"),(0,l.kt)("td",{parentName:"tr",align:null},"210.477"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"17.369")))),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"143.252"),(0,l.kt)("td",{parentName:"tr",align:null},"68.362"),(0,l.kt)("td",{parentName:"tr",align:null},"17.739")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"142.894"))),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"67.958"))),(0,l.kt)("td",{parentName:"tr",align:null},"17.768")))),(0,l.kt)("h5",{id:"acceleration-factors-1"},"Acceleration Factors"),(0,l.kt)("table",null,(0,l.kt)("thead",{parentName:"table"},(0,l.kt)("tr",{parentName:"thead"},(0,l.kt)("th",{parentName:"tr",align:null},"Ranks"),(0,l.kt)("th",{parentName:"tr",align:null},"GPU"),(0,l.kt)("th",{parentName:"tr",align:null},"Version"),(0,l.kt)("th",{parentName:"tr",align:null},"Total Acceleration Factor"),(0,l.kt)("th",{parentName:"tr",align:null},"DFC Acceleration Factor"),(0,l.kt)("th",{parentName:"tr",align:null},"Integration Acceleration Factor"))),(0,l.kt)("tbody",{parentName:"table"},(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"2.635x"),(0,l.kt)("td",{parentName:"tr",align:null},"3.986x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.097x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"3.112x"),(0,l.kt)("td",{parentName:"tr",align:null},"4.942x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.047x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"3.129x"),(0,l.kt)("td",{parentName:"tr",align:null},"4.996x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.028x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"2.639x"),(0,l.kt)("td",{parentName:"tr",align:null},"3.802x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.284x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"3.793x"),(0,l.kt)("td",{parentName:"tr",align:null},"7.304x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.178x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"3.838x"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"7.360x"))),(0,l.kt)("td",{parentName:"tr",align:null},"2.197x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"2.292x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.970x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.763x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"2.467x"),(0,l.kt)("td",{parentName:"tr",align:null},"3.322x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.736x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"2.475x"),(0,l.kt)("td",{parentName:"tr",align:null},"3.311x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.737x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"2.544x"),(0,l.kt)("td",{parentName:"tr",align:null},"3.396x"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"2.302x")))),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"2.539x"),(0,l.kt)("td",{parentName:"tr",align:null},"3.387x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.271x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"3.917x"))),(0,l.kt)("td",{parentName:"tr",align:null},"7.243x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.252x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"1.939x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.359x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.018x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"2.966x"),(0,l.kt)("td",{parentName:"tr",align:null},"5.050x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.030x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"2.989x"),(0,l.kt)("td",{parentName:"tr",align:null},"5.096x"),(0,l.kt)("td",{parentName:"tr",align:null},"2.032x")))),(0,l.kt)("h5",{id:"summary-1"},"Summary"),(0,l.kt)("p",null,"GPU-acceleration greatly improves performance for this system. Even in the worst-case, time-to-solution is nearly halved when accelerated. In the best-case, we see nearly 4x acceleration, including up to 7.36x acceleration for Density and Force Calculations (DFC). While acceleration is largest for low-CPU and high-GPU configurations, total time-to-solution is shortest with MN5 ACC full-node performance."),(0,l.kt)("h4",{id:"220-atom-complex-molecule-with-velocity-ac-lys-ala19-h"},"220-atom complex molecule with velocity (Ac-Lys-Ala19-H)"),(0,l.kt)("h5",{id:"cpu-only-2"},"CPU-Only"),(0,l.kt)("table",null,(0,l.kt)("thead",{parentName:"table"},(0,l.kt)("tr",{parentName:"thead"},(0,l.kt)("th",{parentName:"tr",align:null},"Ranks"),(0,l.kt)("th",{parentName:"tr",align:null},"Version"),(0,l.kt)("th",{parentName:"tr",align:null},"Total Runtime"),(0,l.kt)("th",{parentName:"tr",align:null},"DFC Runtime"),(0,l.kt)("th",{parentName:"tr",align:null},"Integration Runtime"))),(0,l.kt)("tbody",{parentName:"table"},(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"600.035"),(0,l.kt)("td",{parentName:"tr",align:null},"167.459"),(0,l.kt)("td",{parentName:"tr",align:null},"83.945")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"547.362"),(0,l.kt)("td",{parentName:"tr",align:null},"153.745"),(0,l.kt)("td",{parentName:"tr",align:null},"85.623")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"547.752"),(0,l.kt)("td",{parentName:"tr",align:null},"153.068"),(0,l.kt)("td",{parentName:"tr",align:null},"85.313")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"575.054"),(0,l.kt)("td",{parentName:"tr",align:null},"176.214"),(0,l.kt)("td",{parentName:"tr",align:null},"78.922")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"453.298"),(0,l.kt)("td",{parentName:"tr",align:null},"124.93"),(0,l.kt)("td",{parentName:"tr",align:null},"69.605")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"548.303"),(0,l.kt)("td",{parentName:"tr",align:null},"174.918"),(0,l.kt)("td",{parentName:"tr",align:null},"79.841")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"329.695"),(0,l.kt)("td",{parentName:"tr",align:null},"91.179"),(0,l.kt)("td",{parentName:"tr",align:null},"44.389")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"331.465"),(0,l.kt)("td",{parentName:"tr",align:null},"93.628"),(0,l.kt)("td",{parentName:"tr",align:null},"44.525")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"303.54"),(0,l.kt)("td",{parentName:"tr",align:null},"86.581"),(0,l.kt)("td",{parentName:"tr",align:null},"44.731")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"211.647"),(0,l.kt)("td",{parentName:"tr",align:null},"60.002"),(0,l.kt)("td",{parentName:"tr",align:null},"25.464")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"194.586"))),(0,l.kt)("td",{parentName:"tr",align:null},"58.594"),(0,l.kt)("td",{parentName:"tr",align:null},"25.337")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"204.803"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"56.865"))),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"25.131")))))),(0,l.kt)("h5",{id:"gpu-accelerated-2"},"GPU-Accelerated"),(0,l.kt)("table",null,(0,l.kt)("thead",{parentName:"table"},(0,l.kt)("tr",{parentName:"thead"},(0,l.kt)("th",{parentName:"tr",align:null},"Ranks"),(0,l.kt)("th",{parentName:"tr",align:null},"GPU"),(0,l.kt)("th",{parentName:"tr",align:null},"Version"),(0,l.kt)("th",{parentName:"tr",align:null},"Total Runtime"),(0,l.kt)("th",{parentName:"tr",align:null},"DFC Runtime"),(0,l.kt)("th",{parentName:"tr",align:null},"Integration Runtime"))),(0,l.kt)("tbody",{parentName:"table"},(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"598.696"),(0,l.kt)("td",{parentName:"tr",align:null},"166.277"),(0,l.kt)("td",{parentName:"tr",align:null},"66.671")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"556.409"),(0,l.kt)("td",{parentName:"tr",align:null},"167.079"),(0,l.kt)("td",{parentName:"tr",align:null},"67.77")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"559.434"),(0,l.kt)("td",{parentName:"tr",align:null},"167.492"),(0,l.kt)("td",{parentName:"tr",align:null},"67.693")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"472.29"),(0,l.kt)("td",{parentName:"tr",align:null},"116.643"),(0,l.kt)("td",{parentName:"tr",align:null},"53.591")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"439.148"),(0,l.kt)("td",{parentName:"tr",align:null},"115.902"),(0,l.kt)("td",{parentName:"tr",align:null},"54.806")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{pa
1rentName:"tr",align:null},"436.769"),(0,l.kt)("td",{parentName:"tr",align:null},"114.924"),(0,l.kt)("td",{parentName:"tr",align:null},"54.686")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"441.868"),(0,l.kt)("td",{parentName:"tr",align:null},"161.538"),(0,l.kt)("td",{parentName:"tr",align:null},"39.596")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"422.369"),(0,l.kt)("td",{parentName:"tr",align:null},"161.531"),(0,l.kt)("td",{parentName:"tr",align:null},"40.37")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"419.244"),(0,l.kt)("td",{parentName:"tr",align:null},"161.287"),(0,l.kt)("td",{parentName:"tr",align:null},"39.5")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"359.271"),(0,l.kt)("td",{parentName:"tr",align:null},"89.995"),(0,l.kt)("td",{parentName:"tr",align:null},"36.278")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"351.396"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"87.782"))),(0,l.kt)("td",{parentName:"tr",align:null},"35.957")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"333.514"))),(0,l.kt)("td",{parentName:"tr",align:null},"91.505"),(0,l.kt)("td",{parentName:"tr",align:null},"36.639")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"400.09"),(0,l.kt)("td",{parentName:"tr",align:null},"98.895"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"30.439")))),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"408.39"),(0,l.kt)("td",{parentName:"tr",align:null},"102.911"),(0,l.kt)("td",{parentName:"tr",align:null},"31.185")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"400.112"),(0,l.kt)("td",{parentName:"tr",align:null},"99.655"),(0,l.kt)("td",{parentName:"tr",align:null},"30.568")))),(0,l.kt)("h5",{id:"acceleration-factors-2"},"Acceleration Factors"),(0,l.kt)("table",null,(0,l.kt)("thead",{parentName:"table"},(0,l.kt)("tr",{parentName:"thead"},(0,l.kt)("th",{parentName:"tr",align:null},"Ranks"),(0,l.kt)("th",{parentName:"tr",align:null},"GPU"),(0,l.kt)("th",{parentName:"tr",align:null},"Version"),(0,l.kt)("th",{parentName:"tr",align:null},"Total Acceleration Factor"),(0,l.kt)("th",{parentName:"tr",align:null},"DFC Acceleration Factor"),(0,l.kt)("th",{parentName:"tr",align:null},"Integration Acceleration Factor"))),(0,l.kt)("tbody",{parentName:"table"},(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"1.002x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.007x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.259x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"0.984x"),(0,l.kt)("td",{parentName:"tr",align:null},"0.920x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.263x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"0.979x"),(0,l.kt)("td",{parentName:"tr",align:null},"0.914x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.260x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"1.218x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.511x"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"1.473x")))),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"1.032x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.078x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.270x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"1.255x"))),(0,l.kt)("td",{parentName:"tr",align:null},(0,l.kt)("strong",{parentName:"td"},(0,l.kt)("em",{parentName:"strong"},"1.522x"))),(0,l.kt)("td",{parentName:"tr",align:null},"1.460x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"0.744x"),(0,l.kt)("td",{parentName:"tr",align:null},"0.567x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.117x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"0.720x"),(0,l.kt)("td",{parentName:"tr",align:null},"0.537x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.108x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"0.726x"),(0,l.kt)("td",{parentName:"tr",align:null},"0.539x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.132x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"0.918x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.013x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.224x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"0.943x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.067x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.238x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"0.910x"),(0,l.kt)("td",{parentName:"tr",align:null},"0.946x"),(0,l.kt)("td",{parentName:"tr",align:null},"1.221x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"221103"),(0,l.kt)("td",{parentName:"tr",align:null},"0.529x"),(0,l.kt)("td",{parentName:"tr",align:null},"0.607x"),(0,l.kt)("td",{parentName:"tr",align:null},"0.837x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"0.476x"),(0,l.kt)("td",{parentName:"tr",align:null},"0.569x"),(0,l.kt)("td",{parentName:"tr",align:null},"0.812x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"0.512x"),(0,l.kt)("td",{parentName:"tr",align:null},"0.571x"),(0,l.kt)("td",{parentName:"tr",align:null},"0.822x")))),(0,l.kt)("h5",{id:"summary-2"},"Summary"),(0,l.kt)("p",null,"GPU-Acceleration again has a generally negative effect on performance, most so in high-CPU-count configurations. In best-case for CPU-only runs, total time-to-solution is roughly half of it's GPU-accelerated counterpart. Even in the best-case GPU-only runs, acceleration is relatively minor (~25%). CPU-only full-node performance again displays the best performance, and it would not be advisable to proceed with GPU-acceleration with this system."),(0,l.kt)("h3",{id:"best-case-total-time-to-solution-for-each-system"},"Best-case total time-to-solution for each system"),(0,l.kt)("table",null,(0,l.kt)("thead",{parentName:"table"},(0,l.kt)("tr",{parentName:"thead"},(0,l.kt)("th",{parentName:"tr",align:null},"System"),(0,l.kt)("th",{parentName:"tr",align:null},"Ranks"),(0,l.kt)("th",{parentName:"tr",align:null},"GPU"),(0,l.kt)("th",{parentName:"tr",align:null},"Version"),(0,l.kt)("th",{parentName:"tr",align:null},"Total time-to-solution (s)"))),(0,l.kt)("tbody",{parentName:"table"},(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"Si"),(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"CPU"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"82.64")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"cplx"),(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"142.894")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"Ac-Lys-Ala19-H"),(0,l.kt)("td",{parentName:"tr",align:null},"80"),(0,l.kt)("td",{parentName:"tr",align:null},"CPU"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"194.586")))),(0,l.kt)("h3",{id:"best-case-acceleration-total-time-to-solution-for-each-system"},"Best-case acceleration (total time-to-solution) for each system"),(0,l.kt)("table",null,(0,l.kt)("thead",{parentName:"table"},(0,l.kt)("tr",{parentName:"thead"},(0,l.kt)("th",{parentName:"tr",align:null},"System"),(0,l.kt)("th",{parentName:"tr",align:null},"Ranks"),(0,l.kt)("th",{parentName:"tr",align:null},"GPU"),(0,l.kt)("th",{parentName:"tr",align:null},"Version"),(0,l.kt)("th",{parentName:"tr",align:null},"Acceleration Factor (Compared to CPU-only)"))),(0,l.kt)("tbody",{parentName:"table"},(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"Si"),(0,l.kt)("td",{parentName:"tr",align:null},"16"),(0,l.kt)("td",{parentName:"tr",align:null},"1"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"1.276x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"cplx"),(0,l.kt)("td",{parentName:"tr",align:null},"32"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"231212"),(0,l.kt)("td",{parentName:"tr",align:null},"3.917x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"Ac-Lys-Ala19-H"),(0,l.kt)("td",{parentName:"tr",align:null},"20"),(0,l.kt)("td",{parentName:"tr",align:null},"4"),(0,l.kt)("td",{parentName:"tr",align:null},"240507"),(0,l.kt)("td",{parentName:"tr",align:null},"1.255x")))),(0,l.kt)("h4",{id:"acceleration-total-time-to-solution-in-literature-for-comparison"},"Acceleration (total time-to-solution) in literature, for comparison"),(0,l.kt)("table",null,(0,l.kt)("thead",{parentName:"table"},(0,l.kt)("tr",{parentName:"thead"},(0,l.kt)("th",{parentName:"tr",align:null},"CPU"),(0,l.kt)("th",{parentName:"tr",align:null},"GPU"),(0,l.kt)("th",{parentName:"tr",align:null},"Acceleration Factor (Compared to CPU-only)"))),(0,l.kt)("tbody",{parentName:"table"},(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"16 Cores Ivy Bridge"),(0,l.kt)("td",{parentName:"tr",align:null},"1 Pascal GPGPU"),(0,l.kt)("td",{parentName:"tr",align:null},"3.8x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null},"32 Cores Haswell"),(0,l.kt)("td",{parentName:"tr",align:null},"4 Pascal GPGPU"),(0,l.kt)("td",{parentName:"tr",align:null},"3.7x")),(0,l.kt)("tr",{parentName:"tbody"},(0,l.kt)("td",{parentName:"tr",align:null}
1,"20 Cores Skylake"),(0,l.kt)("td",{parentName:"tr",align:null},"4 Volta GPGPU"),(0,l.kt)("td",{parentName:"tr",align:null},"4.0x")))),(0,l.kt)("h2",{id:"conclusions-and-observations"},"Conclusions and Observations"),(0,l.kt)("p",null,"As demonstrated, the acceleration factor is heavily dependent on the input system. For some input systems, total time-to-solution, Density and Force Calculations, and Integrations are not always accelerated using GPUs and CPU-only runs may be faster. However, in all systems, there are configurations where time-to-solution is accelerated with GPU inclusion relative to CPU-only runs with the same core count. In particular, system based on a 146-atom complex molecule with an initial movement (cplx) saw acceleration which rivals the referenced literature, up to 3.92x acceleration, including acceleration factors for Density and Force Calculations and Integrations of 7.24x and 2.25x, respectively. More information and testing is needed to determine which system components most directly effect the factor of acceleration. However, we can note, in all cases, the configuration with the fastest overall time-to-solution was full-node (80 cores). The challenge for FHI-AIMS users on MareNostrum5 will be to evaluate whether their input system benefits from GPU acceleration (as in the case of the ",(0,l.kt)("inlineCode",{parentName:"p"},"cplx")," system) or if GPU acceleration does not effect or even hampers performance."),(0,l.kt)("h2",{id:"reference-materials"},"Reference Materials"),(0,l.kt)("ul",null,(0,l.kt)("li",{parentName:"ul"},(0,l.kt)("a",{target:"_blank",href:n(3514).Z},"108-atom neutral Si system (Si)")),(0,l.kt)("li",{parentName:"ul"},(0,l.kt)("a",{target:"_blank",href:n(51315).Z},"146-atom neutral complex system with an initial movement (cplx)")),(0,l.kt)("li",{parentName:"ul"},(0,l.kt)("a",{target:"_blank",href:n(79443).Z},"220-atom charged complex system with velocity (Ac-Lys-Ala19-H)"))))}N.isMDXComponent=!0},79443:(t,a,n)=>{n.d(a,{Z:()=>e});const e=n.p+"assets/files/Ac-Lys-Ala19-H.tar-9aa835dedc5404b1990653dbd73728d8.gz"},51315:(t,a,n)=>{n.d(a,{Z:()=>e});const e=n.p+"assets/files/cplx.tar-7be1d953c275b3d83dba28d42f23c478.gz"},3514:(t,a,n)=>{n.d(a,{Z:()=>e});const e=n.p+"assets/files/si.tar-ef2401a2f89233846aed258064be9cbe.gz"}}]);

Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.