CORAL Signals New Dawn for Exascale Ambitions

By Nicole Hemsoth

November 14, 2014

Just when it started to look as though the architectural course had been set for the next wave of large-scale supercomputers, today offered quite a shakeup to the standard.

And it’s not just the amount spent to turn a novel architecture into a pre-exascale reality, although to be fair, it’s rare indeed to see a lump $325 million deal from the Department of Energy to fund new systems with an extra $100 million added to support extreme scale technologies under the FastForward initiative.

Aside from the sheer investment figures, the fascinating part of what’s happening is architectural—and therefore, important in terms of what this means for how centers think about energy consumption, prioritization of extreme scale scientific and security challenges, and of perhaps to some degree, the slightly less dominant position of the U.S. in terms of its national supercomputing capability.

While many expected these first two of the new pre-exascale systems to come out of the CORAL collaboration between Oak Ridge, Lawrence Livermore, and Argonne national laboratories to follow the trends set by Titan and other accelerated Intel-powered x86 machines, those expectations were upended by IBM in today’s announcement about a new class of systems sporting GPUs via a close collaboration among OpenPower members IBM, NVIDIA and Mellanox.

Before we delve into an early overview of the systems, it’s worth noting that the very status of IBM’s role in the future of supercomputing had been called into question over the last year, making this a rather surprising announcement in its own right. From selling off their core HPC-oriented server business to Lenovo to quietly bringing the Blue Gene era to a close, it seemed that their interests were shifting toward a more general Power-based approach for all datacenters—not just HPC with its unique subsets of system choices.

To be fair though, this is still what they’re doing. The massive procurement is for systems that are not exactly distinct HPC offerings per se, but rather more advanced and forward-looking variants on the overall OpenPower push to upend Intel’s dominance. However, with the addition of key technologies from Mellanox and NVIDIA, specifically the latter’s NVLINK technology, the new generation, which we heard for the first time today is called “Power9” IBM has found a way to maintain an edge at the high end while refining the Power approach to the wider datacenter market as these technologies mature and are put to the test at scale….and massive scale, at that.

The result of all of this are two systems that will be installed in the 2017 time frame. Summit, which will be housed at Oak Ridge National Laboratory and will be dedicated to large-scale scientific endeavors ranging from climate modeling to other open science initiatives. The other, called Sierra, is set to be installed at Lawrence Livermore with emphasis on security and weapons stockpile management.

Both are GPU-accelerated systems that have fewer nodes for all the performance they’re able to pack in due to the collaboration between NVIDIA and its Volta architecture, which for those who follow these generations, is two away from where we are now with Pascal expected in 2016. The key here is the NVLink interconnect, which is set to push new limits in terms of making these the “data centric” supercomputers IBM is espousing as the next step beyond supercomputers which have traditionally been valued according only to their floating point capabilities.

We will be exploring the technology in a companion piece that will immediately follow this one and offer a deeper sense of the projected architecture from chip to interconnect. However, to kick off this series, we wanted to provide a touchstone for these first inklings at what exascale-class systems might look like in the U.S. in the years to come.

One thing is for sure, these are packing a lot of punch in a far lessened amount of space. The Summit system at Oak Ridge is expected to push the 150 to 300 peak petaflop barrier, but according to Jeff Nichols, one of the most remarkable aspects of the system is how they were able to work partners IBM, NVIDIA, and Mellanox to create an architecture that can be boiled down to a much smaller number of nodes for far higher performance and a much larger shared memory footprint.

At this stage, Summit will be 5x or more the performance of Titan at 1/5 the size—weighing in at just around 3400 nodes.

“This shared memory capability and lower node count is important to our developers going forward,” he said. “I can say as a computational chemist myself that developers love having fewer nodes to manage and more shared memory per node to work with.”

The “data-centric” approach that IBM has been wrapping around for this announcement in particular is another key feature of the Summit system said Nichols. In addition to having the 5x to 10x performance boost using accelerators, which are already in play at Oak Ridge National Lab on the Titan machine, the capabilities for managing vast amounts of complex simulation data is critical. “We can ingest more data, more varieties of data, and explore modeling and simulation data in new ways that we couldn’t do even with Titan,” he explained. “As we move toward exascale, and this is certainly an early step towards that, we do feel that we have a good path forward in terms of how we’ll develop and deploy future systems along this architectural path” with both computational and data centric needs in mind.

As NVIDIA’s Sumit Gupta told us today that each of these nodes is so powerful that four of them alone today would make the Top 500. “You probably need a couple of racks of servers to get into the Top 500 but GPU performance will advance so much that we’ll get that with just four nodes. The central reason why the largest supercomputers are using accelerators is that CPU alone is too much power. A 150 petaflop system today would be half the power of Vegas—and that isn’t going to improve much.”

Gupta added that NVLink, which will explore in depth in a follow-up technical piece, is central because the CORAL collaborators wanted a fast processor but required a data movement paradigm that would allow data to be handled quickly without extra hops. The traditional CPU and GPU connected traditionally over PCIe has been great for classical high performance computing, he noted, but with high throughput computing users at that scale need the processors to be able to move data efficiently from point to point.

These features are key for the weapons stockpile program that is central to national security where the Sierra system will offer a massive increase in performance and efficiency at Lawrence Livermore. This machine is expected to offer in excess of 100 peak petaflop performance.

As LLNL’s Mike McCoy said today, “Simulation is critical to our stockpile program—it’s critical for us to make sure we never have to return to nuclear testing. But our 3D weapons simulations codes involve 3D applications, multiple physics packages, and our major codes easily run over a million lines not to mention the databases they employ. At the end of the day, key national security decisions are made based on these calculations but the question is always how do we know these systems are going to do the work we need?

In answering his own question he explained the way value of the partnership of OpenPower members. “This is not an off the shelf approach—the partnerships are strong and we share the risk in development and deliver platforms that can rapidly come into production and serve our needs. This effort is achieved through a systems integration approach and there will be tight integration between the vendors and code development teams which is called codesign—this has been interestingly enough applied into the past and led to advances like the Blue Gene L that led to advances and performed. This partnership represents a huge opportunity to deliver these and future first gen exascale systems.

We’ve displaced an Intel-based system at ORNL and we haven’t been there for a number of years. It’s a nice achievement for us,” said IBM’s Dave Turek in a conversation today. But the real value in this news is how it could represent the first seismic shift away from the FLOPS-centric approach to large-scale systems to one that takes the problems of data to heart at the core. “We are aided here not because of anything other than what we’re seeing in terms of the evolution of the marketplace through direct measurement how necessary it is to simultaneously deal with analytics in concert with modeling and simulation. If you look at an example like seismic processing and you go back ten years, the bulk of the time would have been dedicated to the algorithm and making it faster but what’s transformed the conversation is the radical influx of data. Now when you inspect the infrastructure that’s being deployed in examples like this, there’s a tremendous amount of mundane data sorting and managing that’s taking up the compute.

Just as efforts like this have bolstered IBM’s supercomputing products overtime, this new collaboration represents a shift for the company. IBM has in fact established an entirely new HPC roadmap—all around the concept of data centric computing. With these systems, the balance of performance, data movement, memory, and overall footprint are balanced with the needs of the new generations of highly scalable codes under development now with assistance from NVIDIA and IBM.

Follow up with us during your SC travels over the weekend and on Monday for more detail about the architectural features we’ve been able to tease out of a few conversations with IBM, NVIDIA, Mellanox and others.

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industy updates delivered to you every week!

Pittsburgh Supercomputer Powers Machine Learning Analysis of Rare East Asian Stamps

September 27, 2021

Setting aside the relatively recent rise of electronic signatures, personalized stamps have been a popular form of identification for formal documents in East Asia. These identifiers – easily forged, but culturally ubi Read more…

Purdue Researchers Peer into the ‘Fog of the Machine Learning Accelerator War’

September 27, 2021

Making sense of ML performance and benchmark data is an ongoing challenge. In light of last week’s release of the most recent MLPerf (v1.1) inference results, now is perhaps a good time to review how valuable (or not) Read more…

Quantum Monte Carlo at Exascale Could Be Key to Finding New Semiconductor Materials

September 27, 2021

Researchers are urgently trying to identify possible materials to replace silicon-based semiconductors. The processing power in modern computers continues to increase even as the size of the silicon on which components a Read more…

The Case for an Edge-Driven Future for Supercomputing

September 24, 2021

“Exascale only becomes valuable when it’s creating and using data that we care about,” said Pete Beckman, co-director of the Northwestern-Argonne Institute of Science and Engineering (NAISE), at the most recent HPC Read more…

Three Universities Team for NSF-Funded ‘ACES’ Reconfigurable Supercomputer Prototype

September 23, 2021

As Moore’s law slows, HPC developers are increasingly looking for speed gains in specialized code and specialized hardware – but this specialization, in turn, can make testing and deploying code trickier than ever. Now, researchers from Texas A&M University, the University of Illinois at Urbana... Read more…

AWS Solution Channel

Introducing AWS ParallelCluster 3

Running HPC workloads, like computational fluid dynamics (CFD), molecular dynamics, or weather forecasting typically involves a lot of moving parts. You need a hundreds or thousands of compute cores, a job scheduler for keeping them fed, a shared file system that’s tuned for throughput or IOPS (or both), loads of libraries, a fast network, and a head node to make sense of all this. Read more…

Qubit Stream: Monte Carlo Advance, Infosys Joins the Fray, D-Wave Meeting Plans, and More

September 23, 2021

It seems the stream of quantum computing reports never ceases. This week – IonQ and Goldman Sachs tackle Monte Carlo on quantum hardware, Cambridge Quantum pushes chemistry calculations forward, D-Wave prepares for its Read more…

Quantum Monte Carlo at Exascale Could Be Key to Finding New Semiconductor Materials

September 27, 2021

Researchers are urgently trying to identify possible materials to replace silicon-based semiconductors. The processing power in modern computers continues to in Read more…

The Case for an Edge-Driven Future for Supercomputing

September 24, 2021

“Exascale only becomes valuable when it’s creating and using data that we care about,” said Pete Beckman, co-director of the Northwestern-Argonne Institut Read more…

Three Universities Team for NSF-Funded ‘ACES’ Reconfigurable Supercomputer Prototype

September 23, 2021

As Moore’s law slows, HPC developers are increasingly looking for speed gains in specialized code and specialized hardware – but this specialization, in turn, can make testing and deploying code trickier than ever. Now, researchers from Texas A&M University, the University of Illinois at Urbana... Read more…

Qubit Stream: Monte Carlo Advance, Infosys Joins the Fray, D-Wave Meeting Plans, and More

September 23, 2021

It seems the stream of quantum computing reports never ceases. This week – IonQ and Goldman Sachs tackle Monte Carlo on quantum hardware, Cambridge Quantum pu Read more…

Asetek Announces It Is Exiting HPC to Protect Future Profitability

September 22, 2021

Liquid cooling specialist Asetek, well-known in HPC circles for its direct-to-chip cooling technology that is inside some of the fastest supercomputers in the world, announced today that it is exiting the HPC space amid multiple supply chain issues related to the pandemic. Although pandemic supply chain... Read more…

TACC Supercomputer Delves Into Protein Interactions

September 22, 2021

Adenosine triphosphate (ATP) is a compound used to funnel energy from mitochondria to other parts of the cell, enabling energy-driven functions like muscle contractions. For ATP to flow, though, the interaction between the hexokinase-II (HKII) enzyme and the proteins found in a specific channel on the mitochondria’s outer membrane. Now, simulations conducted on supercomputers at the Texas Advanced Computing Center (TACC) have simulated... Read more…

The Latest MLPerf Inference Results: Nvidia GPUs Hold Sway but Here Come CPUs and Intel

September 22, 2021

The latest round of MLPerf inference benchmark (v 1.1) results was released today and Nvidia again dominated, sweeping the top spots in the closed (apples-to-ap Read more…

Why HPC Storage Matters More Now Than Ever: Analyst Q&A

September 17, 2021

With soaring data volumes and insatiable computing driving nearly every facet of economic, social and scientific progress, data storage is seizing the spotlight. Hyperion Research analyst and noted storage expert Mark Nossokoff looks at key storage trends in the context of the evolving HPC (and AI) landscape... Read more…

Ahead of ‘Dojo,’ Tesla Reveals Its Massive Precursor Supercomputer

June 22, 2021

In spring 2019, Tesla made cryptic reference to a project called Dojo, a “super-powerful training computer” for video data processing. Then, in summer 2020, Tesla CEO Elon Musk tweeted: “Tesla is developing a [neural network] training computer called Dojo to process truly vast amounts of video data. It’s a beast! … A truly useful exaflop at de facto FP32.” Read more…

Enter Dojo: Tesla Reveals Design for Modular Supercomputer & D1 Chip

August 20, 2021

Two months ago, Tesla revealed a massive GPU cluster that it said was “roughly the number five supercomputer in the world,” and which was just a precursor to Tesla’s real supercomputing moonshot: the long-rumored, little-detailed Dojo system. “We’ve been scaling our neural network training compute dramatically over the last few years,” said Milan Kovac, Tesla’s director of autopilot engineering. Read more…

Esperanto, Silicon in Hand, Champions the Efficiency of Its 1,092-Core RISC-V Chip

August 27, 2021

Esperanto Technologies made waves last December when it announced ET-SoC-1, a new RISC-V-based chip aimed at machine learning that packed nearly 1,100 cores onto a package small enough to fit six times over on a single PCIe card. Now, Esperanto is back, silicon in-hand and taking aim... Read more…

CentOS Replacement Rocky Linux Is Now in GA and Under Independent Control

June 21, 2021

The Rocky Enterprise Software Foundation (RESF) is announcing the general availability of Rocky Linux, release 8.4, designed as a drop-in replacement for the soon-to-be discontinued CentOS. The GA release is launching six-and-a-half months after Red Hat deprecated its support for the widely popular, free CentOS server operating system. The Rocky Linux development effort... Read more…

Intel Completes LLVM Adoption; Will End Updates to Classic C/C++ Compilers in Future

August 10, 2021

Intel reported in a blog this week that its adoption of the open source LLVM architecture for Intel’s C/C++ compiler is complete. The transition is part of In Read more…

Hot Chips: Here Come the DPUs and IPUs from Arm, Nvidia and Intel

August 25, 2021

The emergence of data processing units (DPU) and infrastructure processing units (IPU) as potentially important pieces in cloud and datacenter architectures was Read more…

AMD-Xilinx Deal Gains UK, EU Approvals — China’s Decision Still Pending

July 1, 2021

AMD’s planned acquisition of FPGA maker Xilinx is now in the hands of Chinese regulators after needed antitrust approvals for the $35 billion deal were receiv Read more…

Google Launches TPU v4 AI Chips

May 20, 2021

Google CEO Sundar Pichai spoke for only one minute and 42 seconds about the company’s latest TPU v4 Tensor Processing Units during his keynote at the Google I Read more…

Leading Solution Providers

Contributors

HPE Wins $2B GreenLake HPC-as-a-Service Deal with NSA

September 1, 2021

In the heated, oft-contentious, government IT space, HPE has won a massive $2 billion contract to provide HPC and AI services to the United States’ National Security Agency (NSA). Following on the heels of the now-canceled $10 billion JEDI contract (reissued as JWCC) and a $10 billion... Read more…

Julia Update: Adoption Keeps Climbing; Is It a Python Challenger?

January 13, 2021

The rapid adoption of Julia, the open source, high level programing language with roots at MIT, shows no sign of slowing according to data from Julialang.org. I Read more…

10nm, 7nm, 5nm…. Should the Chip Nanometer Metric Be Replaced?

June 1, 2020

The biggest cool factor in server chips is the nanometer. AMD beating Intel to a CPU built on a 7nm process node* – with 5nm and 3nm on the way – has been i Read more…

Quantum Roundup: IBM, Rigetti, Phasecraft, Oxford QC, China, and More

July 13, 2021

IBM yesterday announced a proof for a quantum ML algorithm. A week ago, it unveiled a new topology for its quantum processors. Last Friday, the Technical Univer Read more…

Intel Launches 10nm ‘Ice Lake’ Datacenter CPU with Up to 40 Cores

April 6, 2021

The wait is over. Today Intel officially launched its 10nm datacenter CPU, the third-generation Intel Xeon Scalable processor, codenamed Ice Lake. With up to 40 Read more…

Frontier to Meet 20MW Exascale Power Target Set by DARPA in 2008

July 14, 2021

After more than a decade of planning, the United States’ first exascale computer, Frontier, is set to arrive at Oak Ridge National Laboratory (ORNL) later this year. Crossing this “1,000x” horizon required overcoming four major challenges: power demand, reliability, extreme parallelism and data movement. Read more…

Intel Unveils New Node Names; Sapphire Rapids Is Now an ‘Intel 7’ CPU

July 27, 2021

What's a preeminent chip company to do when its process node technology lags the competition by (roughly) one generation, but outmoded naming conventions make it seem like it's two nodes behind? For Intel, the response was to change how it refers to its nodes with the aim of better reflecting its positioning within the leadership semiconductor manufacturing space. Intel revealed its new node nomenclature, and... Read more…

The Latest MLPerf Inference Results: Nvidia GPUs Hold Sway but Here Come CPUs and Intel

September 22, 2021

The latest round of MLPerf inference benchmark (v 1.1) results was released today and Nvidia again dominated, sweeping the top spots in the closed (apples-to-ap Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire