Re: GPU computing for PLM

Phenom
Phenom

Here's a SOL103 example of a 3M dof model ran with and without GPGPU on a laptop (6-core Core i7, 64GB RAM, GeForce RTX2060, Windows 10 Pro). Although not a gigantic step forward, it's better than expected!

 

Engine Block.png3M dof Engine Model

 

Standard solve

Real: 851.578 seconds ( 0:14:11.578)
User: 774.890 seconds ( 0:12:54.890)
Sys: 73.156 seconds ( 0:01:13.156)

gpgpu=nvidia

Real: 717.659 seconds ( 0:11:57.659) User: 452.515 seconds ( 0:07:32.515) Sys: 203.500 seconds ( 0:03:23.500)

 

And here is a set of results on a desktop with an AMD Radeon VII:

 

Standard solve

Real:     1204.756 seconds ( 0:20:04.756)
User:     1127.312 seconds ( 0:18:47.312)
Sys:        80.062 seconds ( 0:01:20.062)

gpgpu=AMD

Real:      906.790 seconds ( 0:15:06.790)
User:      487.984 seconds ( 0:08:07.984)
Sys:       198.093 seconds ( 0:03:18.093)