IBM and NVIDIA are establishing a new in-person accelerator lab to help usher in the next wave for accelerated computing. Work directly with IBM and NVIDIA experts and the new OpenPOWER HPC Server: the first server combining NVIDIA Tesla P100 and POWER8 with NVLink Technology.
Early access to NVLink technology
- This platform directly connects the POWER8 CPU to the NVIDIA Tesla P100 GPU with the new NVLink bus — delivering up to 2.5X the bandwidth of PCI-E x16 3.0.
- Running your code on systems with POWER8 with NVLink Technology and Tesla P100 can rapidly result in a substantial application speedup.
Team up with NVIDIA and IBM on advanced acceleration
- Many accelerated applications mask data transfers --- presenting a ceiling on performance. The broad NVLink connection from POWER8 to Tesla P100 results in far less complexity for transfers and simpler data management.
- Take advantage of the wider bus and realize explosive new performance potential.
Unlock what was previously impossible
- Some applications were deemed too difficult to bring to GPUs: too much data movement, too large a memory footprint. The Page Migration Engine in NVIDIA’s UVM Architecture dramatically reduces the effort to implement GPU computing.
- Coupled with POWER8 with NVlink, UVM lowers the barriers to GPU computing. The application aperture has broadened.