Tuesday, 16 May 2017

Deep Learning using 2 Power8 CPUs (16 cores) @4Ghz and 2 Nvidia P100 GPUs (7186 cores)

Today we are breaking the time to complete a Kitti benchmark using 100 epochs with an IBM Minsky server.
Reference machine: Supermicro Intel based server with 2xP100 Nvidia cards. Test time: 300 minutes.
Our server: IBM Minsky Power8 based server with 2xP100 Nvidia cards. Predicted Test time: 131 minutes. (currently running) Final time 2h20mins (140 mins)

More details to come in next days...




No comments:

Flying in the AVE