Discussion Share your High-END GPU Non-GAMING use case and stories - We use GPU for mission critical quality inspection in manufacturing.

traderjay

Senior member
Sep 24, 2015
220
165
116
It has been a crappy year but manufacturing has rebounded and we are leading the way for economic recovery. Without manufacturing and consumers, we can say goodbye to the Western standard of living!

Taking the portable STRIX RTX 3090 + Ryzen 3950X workhorse to its natural habitat with industrial robots and high powered 3D laser scanner. I am privileged to work for a very innovative company called Bluewrist where we specializes in 3D vision and robotic automation. Our customers include to top names in automotive and consumer goods manufacturers and it really gives me a good insight on how products are built and dedication that goes into them.

In this picture, we are testing an automated system for inspecting the door panels for several DOZENS of defects. A high powered 3D laser scanner with 0.01 mm accuracy scans the entire door panel and the 3D image in point cloud format is streamed into our software for analysis. Those images are sample scans of the door panel, made up of millions of points. It is extremely computational intensive and our talented computer vision team developed the software in-house and it utilizes shape matching, AI and deep learning.

Gamers compete on frame rate, in our line of work, our performance is measured in milliseconds to seconds! You can see a demo video of this system - https://www.youtube.com/watch?v=tKnnrSqHJYE

We have many other GPU accelerated inspection application for those interested in reading - https://bluewrist.com/casestudy/

Today is International Quality Day so lets cheers to all the Quality Engineers out there!


_39A9605_DxO.jpg

_39A9613_DxO.jpg

_39A9615.jpg


Scan.jpg


Speaker.jpg
_39A9602_DxO.jpg
 

NTMBK

Lifer
Nov 14, 2011
10,208
4,940
136
Coolest thing I ever worked on with a GPU was using Tesla GPUs to visualise medical scan data during live surgery. 3D rendering of a volume of data, aligned with live fluoroscopy data from the X ray camera, guiding a probe being fed into the patient's brain.
 

Borealis7

Platinum Member
Oct 19, 2006
2,914
205
106
I work for a company that developed and ASIC unit (we call it an APU - Associative Processing Unit) for running complex vector calculations and algorithms in-memory instead of in a "core" (be it CPU or CUDA). We test and benchmark massive parallel calculations on our ASIC against nVidia cards since those are the ones that are usually used by people running these sort of calculations. we do Hashing, Brute force attacks, Hamming Distance, Cosine Similarity, anything that is massively parallel basically.
Of course, our units perform better and draw less power per TFLop :) and no, we do not mine crypto...
 
Last edited:
  • Like
Reactions: lightmanek

CP5670

Diamond Member
Jun 24, 2004
5,508
586
126
I worked at a tech startup some years ago where we were using deep learning to do face recognition at long range. The main product was a laser imaging camera that picked up a 3D mesh of a face alongside a 2D image, and could recognize people walking around at off-angles at long range. Back then a lot of the modern deep learning libraries did not exist, and you had to write everything from scratch in CUDA.
 
  • Like
Reactions: NTMBK and traderjay