Question How do I get my two Tesla M40s to work together on my Asus B450-F Gaming II Board?

[jonathynblythe]

Hello,

I have an AMD Ryzen 9 (no X for the video, I'm using an NVIDIA video card for that) and an Asus ROG Strix B450-F Gaming II board. I'm using two NVIDIA Tesla M40 GPUs, but every time I try to use an AI (like Faceswap), it only uses the one GPU (PCIEX16_1 slot) and does nothing with the GPU in the PCIEX16_2 slot. Why is this? My understanding is that this Asus board supports SLI and Crossfire, but how do activate it? They're both the same model of GPU, so can't I get them to work simultaneously for a total of 6000+ Cuda cores (they each have about 3000+) for deep learning and machine learning?

Thanks in advance. Jon.

 

[egda23]

Training a model in a data-distributed fashion requires use of advanced algorithms like allreduce or parameter-server algorithms.
You do not need SLI or Crossfire.

You can find some explanation here:
View: https://www.reddit.com/r/MachineLearning/comments/4tx9mj/using_two_different_gpus_in_dl_workstation/

 

[jonathynblythe]

Thank you for the quick response! I'm not exactly clear as to what all that means. Is there anything for the layman as far as that goes? Thanks again!