Brown GPU nodes available for use

March 20, 2018 – April 9, 2018
BrownGPU, HalsteadGPU

The Brown cluster's GPU nodes have entered friendly-user mode, and are available for purchase. To make the cost of the GPUs more affordable, ITaP is pleased to continue with a cloud-like, annual subscription-based model that was first used on Halstead GPU.

Your feedback from HalsteadGPU made clear that a key requirement was "larger", to keep up with demand - therefore, the new system consists of 15 compute nodes, each with 2 Intel Xeon Gold 6126 processors (24 cores total), 192GB of RAM, EDR Infiniband, and 3 Nvidia P100 GPU accelerators - a total of 45 GPUs with a total 211.5 TF of double-precision performance (418.5 TF single-precision).

The GPU nodes have access to the same high-speed parallel scratch (3.4PB) as the main Brown cluster. These P100 “Pascal” GPUs are well suited to accelerate computations in molecular dynamics, machine learning, quantum chemistry, and many other domains.

This annual charge gets you and your group access to a shared queue, only available to researchers who have purchased access. If you purchase access to the system, you can add your students to the shared queue exactly like you do on other community clusters.

Alternatively, if you require a dedicated queue like you may have on other community clusters, that is available as well.

You can purchase access to the Brown GPU nodes here: https://www.rcac.purdue.edu/purchase/.

We look forward to your feedback as we continue to invest in providing the right resources for you via our community cluster program. Our goal is to provide GPU systems that are as cost-effective, efficiently utilized, and capable as our CPU-based community clusters, so that you can focus on your research rather than operating computer systems.

If you have any questions, please don’t hesitate to contact us at rcac-help@purdue.edu.

Originally posted: March 20, 2018  10:49am