• Home
  • News
  • Tutorials
  • Analysis
  • About
  • Contact

TechEnablement

Education, Planning, Analysis, Code

  • CUDA
    • News
    • Tutorials
    • CUDA Study Guide
  • OpenACC
    • News
    • Tutorials
    • OpenACC Study Guide
  • Xeon Phi
    • News
    • Tutorials
    • Intel Xeon Phi Study Guide
  • OpenCL
    • News
    • Tutorials
    • OpenCL Study Guide
  • Web/Cloud
    • News
    • Tutorials
You are here: Home / News / Inside the IBM POWER Systems 2x Price-Performance over x86 Claims

Inside the IBM POWER Systems 2x Price-Performance over x86 Claims

September 8, 2016 by Rob Farber Leave a Comment

Today (Sept. 8, 2016) IBM POWER Systems unveiled a new family of servers that IBM claims will transform the economics of the Data Center: purpose-built for AI and machine learning and that can deliver a 2X Price-Performance Advantage over x86-Based Servers. Designed in collaboration with the OpenPOWER Foundation (including some of the biggest and innovative tech companies in the world), IBM claims this hardware is setting new performance benchmarks for the industry. Some IBM provided benchmark results are provided below. Systems will be available starting September 26.

The New POWER Systems

The following graphic shows the technical details of the three new systems. The S822LC for HPC system incorporates the new POWER8 with NVLink on the Processor. The other two systems utilize the PCIe bus for GPU communications.

09-08-16_ibm_servers-fs8

The first NVLink-enabled CPUs

The S822LC for HPC finally gives users the ability to see NVLink, Pascal GPUs, and POWER8 processors working together – without the limitations of the PCIe bus – to make offload mode programming an optimization rather than a requirement.

Arthur S. (Buddy) Bland, Oak Ridge Leadership Computing Facility project director at ORNL notes,  “As a long-time user of GPUs, we believe that this will improve the performance of our applications and make it easier for the users to deliver great science.

Buddy Bland, project director at the OLCF

As we have seen, NVIDIA’s Pascal NVLink GPUs are now in the wild.

NVIDIA_Pascal_delivery-fs8

Click to see more

TechEnablement spoke with Ian Buck  (VP Accelerated Computing, NVIDIA) today (Sept. 8. 2016) about the availability of the NVLink-enabled NVIDIA P100 Pascal GPUs. He states NVIDIA is in full production, which means there should not be a problem with customers getting Pascal GPUs for the S822LC for HPC systems. We note that all the new servers are CAPI enabled, which means that  other accelerators can be used with this new IBM server family.

Ian Buck

Ian Buck (VP Accelerated Computing, NVIDIA)

Try your own benchmarks!

To help developers best take advantage of new hardware and see the potential for performance gains for themselves, IBM established the Accelerated Discovery Lab. Contact them at acceldis@us.ibm.com to get access.

We are happy to note that IBM Fellow Brad McCredie (Vice President and IBM Fellow) is apparently making good on his promise for 3rd parties to get access to an OpenPower system with NVlink during the 3rd quarter of this year. Read more about what McCredie said in, Power 9 info plus IBM to make NVlink demo system available.

It is likely that access to these systems will be popular given the significant interest in GPU vs. CPU benchmarks! Here is one example, NVIDIA – “[Intel] Should Get Their Facts Straight” on Machine Learning Benchmarks. Of course, POWER8 is now thrown into the mix as well as x86 and GPUs. Let us know your benchmark results when they come in!

Cloud and Database Price/Performance

IBM shared their benchmark and Price/Performance data with TechEnablement. Customers who want a water-cooled option for the CPUs, GPUs, and memory can pay IBM an extra $1,500 for water blocks and piping. Water cooling will offer about a 15% performance boost over air cooling as the Power8 chips can run in Turbo mode and the Teslas to run in GPUBoost mode in a sustained fashion. To our knowledge, the following benchmarks were performed on an air-cooled system. Alas, no GPU-based, NVLink benchmarks were provided.

09-08-16_ibm_lc_benchmarks-1

09-08-16_ibm_lc_benchmarks-2

09-08-16_ibm_lc_benchmarks-3

09-08-16_ibm_lc_benchmarks-4

 

 

Share this:

  • Twitter

Filed Under: Analysis, Featured article, Featured news, News Tagged With: HPC, IBM

Leave a Reply Cancel reply

Your email address will not be published. Required fields are marked *

Tell us you were here

Recent Posts

Farewell to a Familiar HPC Friend

May 27, 2020 By Rob Farber Leave a Comment

TechEnablement Blog Sunset or Sunrise?

February 12, 2020 By admin Leave a Comment

The cornerstone is laid – NVIDIA acquires ARM

September 13, 2020 By Rob Farber Leave a Comment

Third-Party Use Cases Illustrate the Success of CPU-based Visualization

April 14, 2018 By admin Leave a Comment

More Tutorials

Learn how to program IBM’s ‘Deep-Learning’ SyNAPSE chip

February 5, 2016 By Rob Farber Leave a Comment

Free Intermediate-Level Deep-Learning Course by Google

January 27, 2016 By Rob Farber Leave a Comment

Intel tutorial shows how to view OpenCL assembly code

January 25, 2016 By Rob Farber Leave a Comment

More Posts from this Category

Top Posts & Pages

  • DARPA Goals, Requirements, and History of the SyNAPSE Project
  • Recovering Speech from a Potato-chip Bag Viewed Through Soundproof Glass - Even With Commodity Cameras!
  • Paper Compares AMD, NVIDIA, Intel Xeon Phi CFD Turbulent Flow Mesh Performance Using OpenMP and OpenCL
  • Lustre Delivers 10x the Bandwidth of NFS on Intel Xeon Phi
  • South Africa Team Wins Their Second Student Supercomputing Competition At ISC14

Archives

© 2025 · techenablement.com