7 Comments
User's avatar
Sharad Mehrotra's avatar

This is a nice overview, thanks. Your drawing where you show a quad of GPUs connected by point-to-point NVLinks replaced by switches is incorrect.

That said, the point you make that point-to-point links statically divide node bandwidth in proportion to the number links, and that using switches instead allows use of the full bandwidth of the link bandwidth to be used for any connection, is correct. It’s just the drawing that needs fixing.

Dr. Ashish Bamania's avatar

Thanks, Sharad. I've fixed it!

Naveen Venugopal's avatar

Thanks man basic and simple

Dr. Ashish Bamania's avatar

Thank you! Glad you found it helpful :)

Ramanth Kumar Reddy Polu's avatar

Great way of simplifying things. Enjoyed reading