Hello - I understand there are 8 gpus connecting to the backend network, and there are 4 ethernet ports connected to the frontend network. My question is when these gpus are requesting training data how and what controls what ethernet ports to send this traffic over? There must be some kind of load balancing, what controls/maps these flows.
Related topics
Topic | Replies | Views | Activity | |
---|---|---|---|---|
bandwidth of GPUs | 6 | 2057 | June 23, 2010 | |
GPU cluster | 16 | 13528 | January 9, 2008 | |
four 9800GX2 cards: will it work? | 33 | 23388 | May 28, 2008 | |
S870 Configurations... | 2 | 5200 | December 6, 2007 | |
Using more than 1 CUDA card at a time. Physics simulations flat out flying on GPU | 12 | 12596 | March 12, 2010 | |
The fastest platform of GPU computing | 38 | 40412 | January 19, 2010 | |
very large data set (big matrix) | 10 | 3067 | October 17, 2009 | |
multiGPU poor performance up to 10x lowest performance in multiGPU | 14 | 10813 | January 18, 2008 | |
4 GPUs | 10 | 8798 | November 10, 2007 | |
Best solution for maximizing bandwidth? More then 5.7G H->D bandwidth except Tesla | 24 | 11141 | December 26, 2008 |