National Energy Research Scientific Computing Center (NERSC) General NERSC Info, 24 Feb 2016

Previous Meeting

Next Meeting

⏯

youtube image

►

From YouTube: NERSC Pushes The Limits Of Discovery | Intel Business

Description

Katie Antypas, Department Head, Scientific Computing and Data Services, National Energy Research Scientific Computing Center (NERSC) talks about how the NERSC Cori supercomputer, Cray*, and the Intel® Xeon® Phi coprocessor (codenamed Knights Landing) are contributing to a broad range of science pursuits such as astrophysics, chemistry, biology, and geo science. Learn more at https://software.intel.com/moderncode

Subscribe now to Intel Business on YouTube: http://intel.ly/intelitcenteryt

About Intel Business:
Get all the IT info you need, right here. From data center to devices, the Intel® Business Center has the resources, guidance, and expert insights you need to get your IT projects done right.

Connect with Intel Business:
Visit Intel Business's WEBSITE: http://intel.ly/itcenter
Follow Intel Business on TWITTER: https://twitter.com/IntelBusiness
Follow Intel Business on LINKEDIN: https://www.linkedin.com/showcase/intel-business/
Follow Intel Business on FACEBOOK: https://www.facebook.com/IntelBusiness

NERSC Pushes The Limits Of Discovery | Intel Business
https://www.youtube.com/intelitcenter

A

Newark is the national energy research scientific computing center? We are the mission high-performance computing center for the Department of Energy Office of Science. As such, we support an incredibly broad range of science from astrophysics chemistry, biology, Geoscience, really anything that the Department of Energy supports and those users that need high performance computing. They come to nurse our latest supercomputer is corey, is a cray xe6 system with intel processors.

A

Applications are going to have to adapt, there's new, o technologies and Cory there's new technologies in the Knights landing processor, and these applications may not be prepared to take advantage of this architecture. So what we've decided to do is really focus on about 20 different applications, and these applications are in a program that we call nice app. The nurse exascale science application.

A

These are teams that want to focus on their science. They want to focus on their applications and right now we're at this this turning point where its applications teams don't spend some time. Looking at application performance, they really could be left behind, and so the nice app program is providing that focal point and allowing users kind of the time and the resources to be able to really focus on a code. Modernization preparing their codes for these next-generation architectures.

A

One of the reasons we're moving to the quarry architecture is due to energy efficiency. The quarry architecture with the Knights landing processor is one of the steps that we need to take in order to move this very diverse workload at nurse to more energy efficient, architectures nice landing processor will have more than sixty four cores on a chip with up to four Hardware threads each.

A

So this is an incredible increase in the amount of parallelism our users are used to, and so what they need to do is make sure their applications really scale, whether that's using open and P, and P I or newer programming models and they'll need to make sure that threading capabilities are enabled in their applications. But now we're seeing you know, vector units of 512 bits and that's something users can no longer ignore and so we're working with our users with Intel to get really good vectorization on the Knights landing processor.

A

One of the things I'm really excited about is the high bandwidth on package memory on the Knights landing processor. So one of the challenges for our applications will be how exactly they can take advantage of that high bandwidth memory. Are there certain arrays that they can put into the high bandwidth memory? Should they use it as cash? This high bandwidth memory is really an opportunity for our application teams to get much higher memory bandwidth much higher application performance.

A

Preparing these applications for new pre exascale architecture requires a lot of different skill sets. You need people who understand the architecture. You need to people that understand the application performance. Well, you can't do these in a vacuum. You have to have this the science and the application team really close by so that you know the improvements. Are you making in the coder are scientifically viable? What we think the nice a program really does is bring together those experts from Intel from nurse and the application teams, and you see these people all working together.

A

You know closely and then they come out after a deep dive session with provements of maybe two or three times.

A

There are about 20 application teams that are part of the nice app program and the I think 11 says this most exciting is to see the improvements in these climate models as part of the nice at program.

A

You know these climate models are incredibly important for understanding our environment and the warming that's happening on the earth, and so what we want to do is make sure that, as we move to more advanced architectures in the next five and ten years that these climate codes really can take advantage of all the features on these new architectures, we want to make sure we're making changes and encouraging changes that will last throughout the next five ten hour. 15 years.

A

One of the things I've always enjoyed about computer science and computing is that the computers they don't know who I? Am they don't know if I'm a man I'm a woman, my race, my Creed, they just know if I am computing right or wrong. You need everyone in high-performance, computing and scientific computing that we can get, because we are not going to be able to solve the problems of this next century, with only half of the workforce participating.

A