News for hackers. Stuff that matters.
Re:U.S. To Build Two Flagship Supercomputers for National Labs by Jtsummers on Friday November 14, 2014 @03:32PM (#8609321)
From the video in one of the links it seems that it's still nodes containing both CPU (POWER) and GPU (Tesla) (some number of each). But instead of PCI-e, it uses a new high-speed bus (NVLink) which allows for direct GPU-GPU communication and for the GPU and CPU to share memory at the same speeds (doesn't rely on PCI-e to connect GPU to CPU).
Re:U.S. To Build Two Flagship Supercomputers for National Labs by pmalynin on Friday November 14, 2014 @04:32PM (#8609621)
There is now also an RDMA protocol that understands GPU RAM