In this podcast, the Radio Free HPC team reviews the highlights of the GPU Technology Conference.
The rapid growth in deep learning workloads has driven the need for a faster and more scalable interconnect, as PCIe bandwidth increasingly becomes the bottleneck at the multi-GPU system level. NVLink is a great advance to enable eight GPUs in a single server, and accelerate performance beyond PCIe. But taking deep learning performance to the next level will require a GPU fabric that enables more GPUs in a single server, and full-bandwidth connectivity between them.
NVIDIA NVSwitch is the first on-node switch architecture to support 16 fully-connected GPUs in a single server node and drive simultaneous communication between all eight GPU pairs at an incredible 300 GB/s each. These 16 GPUs can be used as a single large-scale accelerator with 0.5 Terabytes of unified memory space and 2 petaFLOPS of deep learning compute power.”
For more details on DGX-2, check out our insideHPC interview with NVIDIA’s Marc Hamilton.
After that, we do our Catch of the Week:
- Henry is very proud of his new sport coat.
- Shahin notes that the data privacy the key story these days. Facebook is on the ropes, and their CEO Mark Zuckerberg will appear before Congress this week to answer questions about the Cambridge Analytics scandal.
- Dan is not optimistic about Apple switching to their own ARM processor for laptops. Rich notes that they have done it successfully before when they switched from POWER to Intel 12 or so years ago.
- Rich is hitting the road once again. Stay tuned for videos from Swiss HPC Conference in Lugano and the HPC User Forum in Tucson.