Commodity Memory for a High Performance World: The View from Samsung

By Nicole Hemsoth

February 21, 2012

As CPUs get ever more powerful, their hunger for data is outstripping the ability of memory systems to feed them. This is especially true in high performance computing, where memory capacity is often constrained by cost, and more recently, energy consumption. With larger supercomputers now incorporating upwards of hundreds of terabytes of main memory, and often with multiple terabytes of flash storage, this challenge will continue to grow.

As one of the world leaders in memory solutions, Samsung Semiconductor has been a key supplier of DRAM and NAND components that end up in high performance computing systems. Dr. Byungse So, who heads the Memory Product Planning & Application Engineering team at Samsung, shares his thoughts about the memory technologies needed by performance-minded users today and what might come next.

HPCwire: The memory wall is perhaps one of the biggest impediments in high performance computing today. What is the main reason that DRAM performance has failed to keep pace with processor performance?

Byungse So: In current system architectures, the CPU always needs high bandwidth and high density memory with low cost to keep feeding data to the computing engine. Therefore system memory has typically been a module-based commodity solution using multiple standard DRAM components. Due to its cost-driven design, system performance is compromised. It is time to shift gears by switching from cost-optimized to performance-optimized designs via application specific memory.

HPCwire: So what types of innovations will be needed? Are there DRAM technologies on the horizon that promise to do this?

So: Samsung has been exploring ways of narrowing the gap. Short term, Samsung will continue to provide the highest speed system and graphic memory. In the long run, Samsung will be investigating new interface technologies to meet high bandwidth requirements aimed at double-digit performance improvements over DDR3. To make this happen, close collaboration with CPU producers, platform builders as well as system users, is required.

HPCwire: Memory capacity can be an additional impediment to performance. Given the need for terabytes of memory for some HPC applications and the expense that entails, is there a way to build reasonably-priced systems that can support big memory applications?

So: Samsung has been leading the market with the highest capacity memory solutions and is already providing solutions which allow building terascale system memory for a single dual-processor system. Combined with the most advanced production process, Samsung will keep investigating how new stacking technologies including 3DS — 3D-Stacking –can allow adequate memory support of even most capacity-hungry systems.

HPCwire: The commodity DRAM business is well known for being rather unforgiving to suppliers, with some firms like OCZ exiting the market and others like Qimonda eventually going bankrupt. Why is the DRAM business so precarious and how has Samsung managed to maintain its viability in this segment?

So: DRAM as commodity memory has been under continuous pressure from its primary users, that is, the PC and notebook market, to continuously lower its cost. This enormous cost pressure has kept a number of suppliers in the red for a long time, ultimately pushing some suppliers out of the market, while also leading to designs compromising between cost and performance. Samsung’s way out of this dilemma is a combined focus on industry-best economies of scale, industry leading process technology and optimized, application-oriented product solutions for both mobile and server applications.

HPCwire: How do you think NAND memory technology can best be applied in high performance computing systems?

So: Solid state drives are the next-generation storage devices that have overcome the shortcomings of rotating media in high performance computing systems. The SSD adoption rate in volume servers is supposed to grow dramatically. Its interface will slowly be changing from SATA to SAS or PCIe for easy migration.

The most important parts in this emerging product line are the NAND flash components as well as the controllers which manage the input and output of data. Other major components are DRAM, which help buffer during data input and output, and the firmware that drives an SSD.

Samsung is able to combine its extensive know-how, stemming from the design and the production of billions of NAND flash and DRAM components, with its long history of designing countless controllers and firmware, as a company not being “only” vertically integrated.

HPCwire: Are there other solid state memory technologies that you believe will prove their worth commercially? If so, how will they change the memory-storage hierarchy?

So: Many studies and researches on fast non-volatile memory have been conducted in the past years, such as PRAM and MRAM as potential future memory technologies. However, more time is needed to examine their commercial value, as currently no existing hardware can fully utilize the features of these NVM solutions in their different roles.

Based on the maturity of the chosen technology, the usage model might also affect the memory hierarchy. It seems quite obvious that DRAM speed like fast NVM can change the existing memory hierarchy. OS experts may have better answers to this question.

HPCwire: Will these newer technologies eventually displace NAND or even DRAM?

So: It will take more than just better technology to replace existing memories, as they possibly impact system architecture and the value chain. Hence we expect that these new technologies will slowly penetrate into different parts of systems and industries and start to gradually build a new ecosystem. Eventually though, a large portion of NAND and DRAM will be replaced.

Dr. So will be delivering a keynote on advanced memory technology at this year’s International Supercomputing Conference (ISC’12) in Hamburg Germany. His presentation will take place on Tuesday, June 19.

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!

HOKUSAI’s BigWaterfall Cluster Extends RIKEN’s Supercomputing Performance

February 21, 2018

RIKEN, Japan’s largest comprehensive research institution, recently expanded the capacity and capabilities of its HOKUSAI supercomputer, a key resource managed by the institution’s Advanced Center for Computing and C Read more…

By Ken Strandberg

Neural Networking Shows Promise in Earthquake Monitoring

February 21, 2018

A team of Harvard University and MIT researchers report their new neural networking method for monitoring earthquakes is more accurate and orders of magnitude faster than traditional approaches. Read more…

By John Russell

HPE Wins $57 Million DoD Supercomputing Contract

February 20, 2018

Hewlett Packard Enterprise (HPE) today revealed details of its massive $57 million HPC contract with the U.S. Department of Defense (DoD). The deal calls for HPE to provide the DoD High Performance Computing Modernizatio Read more…

By Tiffany Trader

HPE Extreme Performance Solutions

Experience Memory & Storage Solutions that will Transform Your Data Performance

High performance computing (HPC) has revolutionized the way we harness insight, leading to a dramatic increase in both the size and complexity of HPC systems. Read more…

Topological Quantum Superconductor Progress Reported

February 20, 2018

Overcoming sensitivity to decoherence is a persistent stumbling block in efforts to build effective quantum computers. Now, a group of researchers from Chalmers University of Technology (Sweden) report progress in devisi Read more…

By John Russell

HOKUSAI’s BigWaterfall Cluster Extends RIKEN’s Supercomputing Performance

February 21, 2018

RIKEN, Japan’s largest comprehensive research institution, recently expanded the capacity and capabilities of its HOKUSAI supercomputer, a key resource manage Read more…

By Ken Strandberg

Neural Networking Shows Promise in Earthquake Monitoring

February 21, 2018

A team of Harvard University and MIT researchers report their new neural networking method for monitoring earthquakes is more accurate and orders of magnitude faster than traditional approaches. Read more…

By John Russell

HPE Wins $57 Million DoD Supercomputing Contract

February 20, 2018

Hewlett Packard Enterprise (HPE) today revealed details of its massive $57 million HPC contract with the U.S. Department of Defense (DoD). The deal calls for HP Read more…

By Tiffany Trader

Fluid HPC: How Extreme-Scale Computing Should Respond to Meltdown and Spectre

February 15, 2018

The Meltdown and Spectre vulnerabilities are proving difficult to fix, and initial experiments suggest security patches will cause significant performance penal Read more…

By Pete Beckman

Brookhaven Ramps Up Computing for National Security Effort

February 14, 2018

Last week, Dan Coats, the director of Director of National Intelligence for the U.S., warned the Senate Intelligence Committee that Russia was likely to meddle in the 2018 mid-term U.S. elections, much as it stands accused of doing in the 2016 Presidential election. Read more…

By John Russell

AI Cloud Competition Heats Up: Google’s TPUs, Amazon Building AI Chip

February 12, 2018

Competition in the white hot AI (and public cloud) market pits Google against Amazon this week, with Google offering AI hardware on its cloud platform intended Read more…

By Doug Black

Russian Nuclear Engineers Caught Cryptomining on Lab Supercomputer

February 12, 2018

Nuclear scientists working at the All-Russian Research Institute of Experimental Physics (RFNC-VNIIEF) have been arrested for using lab supercomputing resources to mine crypto-currency, according to a report in Russia’s Interfax News Agency. Read more…

By Tiffany Trader

The Food Industry’s Next Journey — from Mars to Exascale

February 12, 2018

Global food producer and one of the world's leading chocolate companies Mars Inc. has a unique perspective on the impact that exascale computing will have on the food industry. Read more…

By Scott Gibson, Oak Ridge National Laboratory

Inventor Claims to Have Solved Floating Point Error Problem

January 17, 2018

"The decades-old floating point error problem has been solved," proclaims a press release from inventor Alan Jorgensen. The computer scientist has filed for and Read more…

By Tiffany Trader

Japan Unveils Quantum Neural Network

November 22, 2017

The U.S. and China are leading the race toward productive quantum computing, but it's early enough that ultimate leadership is still something of an open questi Read more…

By Tiffany Trader

AMD Showcases Growing Portfolio of EPYC and Radeon-based Systems at SC17

November 13, 2017

AMD’s charge back into HPC and the datacenter is on full display at SC17. Having launched the EPYC processor line in June along with its MI25 GPU the focus he Read more…

By John Russell

Researchers Measure Impact of ‘Meltdown’ and ‘Spectre’ Patches on HPC Workloads

January 17, 2018

Computer scientists from the Center for Computational Research, State University of New York (SUNY), University at Buffalo have examined the effect of Meltdown Read more…

By Tiffany Trader

IBM Begins Power9 Rollout with Backing from DOE, Google

December 6, 2017

After over a year of buildup, IBM is unveiling its first Power9 system based on the same architecture as the Department of Energy CORAL supercomputers, Summit a Read more…

By Tiffany Trader

Nvidia Responds to Google TPU Benchmarking

April 10, 2017

Nvidia highlights strengths of its newest GPU silicon in response to Google's report on the performance and energy advantages of its custom tensor processor. Read more…

By Tiffany Trader

Fast Forward: Five HPC Predictions for 2018

December 21, 2017

What’s on your list of high (and low) lights for 2017? Volta 100’s arrival on the heels of the P100? Appearance, albeit late in the year, of IBM’s Power9? Read more…

By John Russell

Russian Nuclear Engineers Caught Cryptomining on Lab Supercomputer

February 12, 2018

Nuclear scientists working at the All-Russian Research Institute of Experimental Physics (RFNC-VNIIEF) have been arrested for using lab supercomputing resources to mine crypto-currency, according to a report in Russia’s Interfax News Agency. Read more…

By Tiffany Trader

Leading Solution Providers

Chip Flaws ‘Meltdown’ and ‘Spectre’ Loom Large

January 4, 2018

The HPC and wider tech community have been abuzz this week over the discovery of critical design flaws that impact virtually all contemporary microprocessors. T Read more…

By Tiffany Trader

Perspective: What Really Happened at SC17?

November 22, 2017

SC is over. Now comes the myriad of follow-ups. Inboxes are filled with templated emails from vendors and other exhibitors hoping to win a place in the post-SC thinking of booth visitors. Attendees of tutorials, workshops and other technical sessions will be inundated with requests for feedback. Read more…

By Andrew Jones

How Meltdown and Spectre Patches Will Affect HPC Workloads

January 10, 2018

There have been claims that the fixes for the Meltdown and Spectre security vulnerabilities, named the KPTI (aka KAISER) patches, are going to affect applicatio Read more…

By Rosemary Francis

GlobalFoundries, Ayar Labs Team Up to Commercialize Optical I/O

December 4, 2017

GlobalFoundries (GF) and Ayar Labs, a startup focused on using light, instead of electricity, to transfer data between chips, today announced they've entered in Read more…

By Tiffany Trader

Tensors Come of Age: Why the AI Revolution Will Help HPC

November 13, 2017

Thirty years ago, parallel computing was coming of age. A bitter battle began between stalwart vector computing supporters and advocates of various approaches to parallel computing. IBM skeptic Alan Karp, reacting to announcements of nCUBE’s 1024-microprocessor system and Thinking Machines’ 65,536-element array, made a public $100 wager that no one could get a parallel speedup of over 200 on real HPC workloads. Read more…

By John Gustafson & Lenore Mullin

Flipping the Flops and Reading the Top500 Tea Leaves

November 13, 2017

The 50th edition of the Top500 list, the biannual publication of the world’s fastest supercomputers based on public Linpack benchmarking results, was released Read more…

By Tiffany Trader

V100 Good but not Great on Select Deep Learning Aps, Says Xcelerit

November 27, 2017

Wringing optimum performance from hardware to accelerate deep learning applications is a challenge that often depends on the specific application in use. A benc Read more…

By John Russell

SC17: Singularity Preps Version 3.0, Nears 1M Containers Served Daily

November 1, 2017

Just a few months ago about half a million jobs were being run daily using Singularity containers, the LBNL-founded container platform intended for HPC. That wa Read more…

By John Russell

  • arrow
  • Click Here for More Headlines
  • arrow
Share This