insideHPC Performance Guru Looks at Nvidia’s New NVLink

Bill D'Amico

“For NVLink to have its highest value it must function properly with unified memory. That means that the Memory Management Units in the CPUs have to be aware of NVLink DMA operations and update appropriate VM structures. The operating system needs to know when memory pages have been altered via NVLink DMA – and this can’t be solely the responsibility of the drivers. Tool developers also need to know details so that MPI or other communications protocols can make use of the new interconnect.”

Managing the GPUs of Your Cluster in a Flexible Way with rCUDA

rcuda

In this talk, we introduce the rCUDA remote GPU virtualization framework, which has been shown to be the only one that supports the most recent CUDA versions, in addition to leverage the InfiniBand interconnect for the sake of performance. Furthermore, we also present the last developments within this framework, related with the use of low-power processors, enhanced job schedulers, and virtual machine environments.”

High Performance Computing at CSCS

Schulthess

“With around 3.2 billion computer operations (3.2 gigaflops) per watt, the combination of GPUs CPUs makes “Piz Daint” one of the world’s most energy-­efficient supercomputers in the petaflop performance class.”

Monitoring and Management Interfaces for GPU Devices in a Cluster Environment

1529987_10153950357685284_1755270361_o

“This presentation will provide an overview of the Nvidia Tesla Deployment Kit (TDK) from a user and a system administrator point of view. TDL contains Nvidia Management Library (NVML) and nvidia-healthmon–a tool for detecting and troubleshooting known GPU issues in a cluster environment. Usage models within a cluster environment will be presented along with a discussion on how existing resource management tools can be extended to improve allocation and accounting of GPU resources.”

GTC 2014 Keynote Time Lapse from insideHPC

timelapse

I produced this time lapse video featuring clips from the opening keynote at GTC 2014. “Nvidia really knows how to wow the audience with a blend of showmanship, state-of-the-art graphics, and dazzling technology.”

Managing GPUs by SLURM

1960911_10153950357690284_914246519_o

“SLURM is an open-source workload manager designed for Linux clusters of all sizes. It provides workload management on many of the most powerful computers in the world and its design is very modular with dozens of optional plugins. This talk will present an overview of SLURM and an analysis of the Consumable Resource Allocation Plugin and its utilization in connection with GPUs.”

Video: Introduction to GPUs

dimitry

“This talk will provide a general introduction to the GPU Development ecosystem by example of typical data processing tasks, compared to other solutions. It will give an overview of key architecture aspects that make amazingly high GPU performance possible.”

Nvidia’s Steve Oberlin on his New Role as CTO for Accelerated Computing

oberlin

In this video from GTC 2014, Steve Oberlin from Nvidia describes his new role as Chief Technical Officer for Accelerated Computing. Along the way, he discusses: the HPC lessons learned from the CRAY T3E and other systems, Nvidia’s plans to tackle the challenges of the HPC Memory Wall, the current status on Project Denver, and how Nvidia plans to couple to the POWER architecture in future systems.

Radio Free HPC Looks at GTC 2014 Technology Announcements

RFHPC

With their upcoming Pascal generation of GPUs, Nvidia plans to bring NVLink and 3D stacked memory to market in their ongoing battle against the HPC memory wall. “This is our 50th episode of Radio Free HPC, so I’d like to offer a Tip of the Hat to my co-hosts, Henry Newman and Dan Olds!”

Video: Rich and Dan Test Drive a Tesla at GTC 2014

tesla

In this video, Dan Olds and Rich Brueckner from Radio Free HPC take a Tesla Motors electric sedan for a test drive. Dan has a lead foot, so hang on and come along for the ride!