Close Menu
  • Home
  • AI Models
    • DeepSeek
    • xAI
    • OpenAI
    • Meta AI Llama
    • Google DeepMind
    • Amazon AWS AI
    • Microsoft AI
    • Anthropic (Claude)
    • NVIDIA AI
    • IBM WatsonX Granite 3.1
    • Adobe Sensi
    • Hugging Face
    • Alibaba Cloud (Qwen)
    • Baidu (ERNIE)
    • C3 AI
    • DataRobot
    • Mistral AI
    • Moonshot AI (Kimi)
    • Google Gemma
    • xAI
    • Stability AI
    • H20.ai
  • AI Research
    • Allen Institue for AI
    • arXiv AI
    • Berkeley AI Research
    • CMU AI
    • Google Research
    • Microsoft Research
    • Meta AI Research
    • OpenAI Research
    • Stanford HAI
    • MIT CSAIL
    • Harvard AI
  • AI Funding & Startups
    • AI Funding Database
    • CBInsights AI
    • Crunchbase AI
    • Data Robot Blog
    • TechCrunch AI
    • VentureBeat AI
    • The Information AI
    • Sifted AI
    • WIRED AI
    • Fortune AI
    • PitchBook
    • TechRepublic
    • SiliconANGLE – Big Data
    • MIT News
    • Data Robot Blog
  • Expert Insights & Videos
    • Google DeepMind
    • Lex Fridman
    • Matt Wolfe AI
    • Yannic Kilcher
    • Two Minute Papers
    • AI Explained
    • TheAIEdge
    • Matt Wolfe AI
    • The TechLead
    • Andrew Ng
    • OpenAI
  • Expert Blogs
    • François Chollet
    • Gary Marcus
    • IBM
    • Jack Clark
    • Jeremy Howard
    • Melanie Mitchell
    • Andrew Ng
    • Andrej Karpathy
    • Sebastian Ruder
    • Rachel Thomas
    • IBM
  • AI Policy & Ethics
    • ACLU AI
    • AI Now Institute
    • Center for AI Safety
    • EFF AI
    • European Commission AI
    • Partnership on AI
    • Stanford HAI Policy
    • Mozilla Foundation AI
    • Future of Life Institute
    • Center for AI Safety
    • World Economic Forum AI
  • AI Tools & Product Releases
    • AI Assistants
    • AI for Recruitment
    • AI Search
    • Coding Assistants
    • Customer Service AI
    • Image Generation
    • Video Generation
    • Writing Tools
    • AI for Recruitment
    • Voice/Audio Generation
  • Industry Applications
    • Finance AI
    • Healthcare AI
    • Legal AI
    • Manufacturing AI
    • Media & Entertainment
    • Transportation AI
    • Education AI
    • Retail AI
    • Agriculture AI
    • Energy AI
  • AI Art & Entertainment
    • AI Art News Blog
    • Artvy Blog » AI Art Blog
    • Weird Wonderful AI Art Blog
    • The Chainsaw » AI Art
    • Artvy Blog » AI Art Blog
What's Hot

Google’s Big Day In Court Wasn’t That Big

Sparkvia AI — Inside the First AI-Powered Writing

Perplexity AI makes its play for government use

Facebook X (Twitter) Instagram
Advanced AI News
  • Home
  • AI Models
    • OpenAI (GPT-4 / GPT-4o)
    • Anthropic (Claude 3)
    • Google DeepMind (Gemini)
    • Meta (LLaMA)
    • Cohere (Command R)
    • Amazon (Titan)
    • IBM (Watsonx)
    • Inflection AI (Pi)
  • AI Research
    • Allen Institue for AI
    • arXiv AI
    • Berkeley AI Research
    • CMU AI
    • Google Research
    • Meta AI Research
    • Microsoft Research
    • OpenAI Research
    • Stanford HAI
    • MIT CSAIL
    • Harvard AI
  • AI Funding
    • AI Funding Database
    • CBInsights AI
    • Crunchbase AI
    • Data Robot Blog
    • TechCrunch AI
    • VentureBeat AI
    • The Information AI
    • Sifted AI
    • WIRED AI
    • Fortune AI
    • PitchBook
    • TechRepublic
    • SiliconANGLE – Big Data
    • MIT News
    • Data Robot Blog
  • AI Experts
    • Google DeepMind
    • Lex Fridman
    • Meta AI Llama
    • Yannic Kilcher
    • Two Minute Papers
    • AI Explained
    • TheAIEdge
    • The TechLead
    • Matt Wolfe AI
    • Andrew Ng
    • OpenAI
    • Expert Blogs
      • François Chollet
      • Gary Marcus
      • IBM
      • Jack Clark
      • Jeremy Howard
      • Melanie Mitchell
      • Andrew Ng
      • Andrej Karpathy
      • Sebastian Ruder
      • Rachel Thomas
      • IBM
  • AI Tools
    • AI Assistants
    • AI for Recruitment
    • AI Search
    • Coding Assistants
    • Customer Service AI
  • AI Policy
    • ACLU AI
    • AI Now Institute
    • Center for AI Safety
  • Business AI
    • Advanced AI News Features
    • Finance AI
    • Healthcare AI
    • Education AI
    • Energy AI
    • Legal AI
LinkedIn Instagram YouTube Threads X (Twitter)
Advanced AI News
IBM

IBM Power Processor, This One Goes to 11

By Advanced AI EditorSeptember 8, 2025No Comments5 Mins Read
Share Facebook Twitter Pinterest Copy Link Telegram LinkedIn Tumblr Email
Share
Facebook Twitter LinkedIn Pinterest Email


Hot Chips 25 was held August 24-26 on the Stanford University campus again this year, with many exciting and interesting presentations. I’ve noticed an overall trend with more focus being placed on overall systems rather than the socket. As the conference name suggests, there’s a history of showcasing chips, but with the increased emphasis on AI and related large-scale computing, efficiency of the total system is also of primary importance.

William Starke from IBM presented, “IBM’s Power11 Processor”. IBM announced Power11 on July 8, 2025, so this was more than a chip presentation, as systems are already being shipped. Their Power Processor roadmap is shown below. While IBM builds 2-4 socket systems, they also go up to 16 sockets. Although the IBM Power chips don’t typically have as many cores per socket as others, their emphasis is on higher per core performance.

Fig. 1: IBM Power Processor roadmap.

IBM sells to customers from small businesses to the world’s largest enterprises, so robust scaling is important to enable flexible deployments for on-prem, hybrid and public cloud scenarios. With modern workloads incorporating more AI, Starke said that he believes that system design and performance are going to be more important going forward and that a system-focused, instead of socket-focused approach is needed.

With 99.9999% of uptime, roughly equivalent to ~31.5 seconds of downtime per year, Power11 is designed to be the most resilient server in the history of the IBM Power platform. The article “Reducing Voltage Guard Band,” published this past March, looked at an IBM power management approach for exceptional reliability.

IBM claims customers for larger systems aren’t looking so much for more capacity but want faster thread performance, so IBM decided to go with Samsung’s Enhanced 7nm process rather than going to 5nm. IBM also worked with Samsung for their packaging of a 2.5D ISC Architecture.

Going from Power10 to Power11 (2-socket systems), the Power11 has 50% more cores at a higher clock speed. For the largest system, that translates into a boost in clock speed from 4.0 GHz to 4.3 GHz (with more cores).

Fig. 2: IBM Power10 to Power11 performance comparison.

Typically, industry implementations use direct attach memory, which isn’t very “beachfront” efficient and burns up quite a bit of the I/O space to communicate with say 8 DDR ports. Instead, IBM uses a hierarchical buffered memory architecture and goes through 16 intermediate buffers that connect to 32 DDR ports (2 DDR ports per buffer) per processor socket. This gives the IBM system effectively a 4x Bandwidth and 4x Capacity (per processor) advantage.

The communication between the processor and the buffers uses high-speed SerDes links running at 38.4 Gb that have very good resilience characteristics with excellent bandwidth and signal integrity.

Figure 3 below shows the specialized OMI DIMM form factor with the buffer chip sitting underneath the copper heatsink.

Fig. 3: IBM Power Processor OMI D-DIMM.

Since the buffer is attached to the D-DIMM, upgrading to newer versions of DDR is straightforward, as the communication from the processor to the memory is through the SerDes links.

Figure 4 below shows the newer Odyssey DDR5 2-port D-DIMM with a 50% increase in bandwidth on the DDR ports going from 25.6 GB/s to 38.4 GB/s and a 2x increase in ports going to 2-ports from 1-port in the older Explorer DDR4 D-DIMM. This gives the Power11 processor a 3x increase in DRAM bandwidth per socket and up to 8 TB of total DRAM capacity.

Fig. 4: IBM Power Processor roadmap.

The inclusion of in-core matrix math accelerators (MMA) in Power10 was a fortunate early feature that has paid dividends in terms of performance for accelerated AI large models and tuning. In addition, Power11 systems will also be able to utilize IBM’s external Spyre accelerators. IBM’s Chris Berry presented IBM Telum II Processor and IBM Spyre Accelerator Chip for AI at last year’s Hot Chips 2024. The IBM Spyre Accelerator is expected to be available in Q4 2025. So, in all, there are really 3 “Hot Chips” in the new system: the Power11, the Odyssey memory buffer, and the Spyre accelerator.

Figure 5 below provides some insight into the direction of the next-generation Power processor, referred to as “Power Future.” It will move to a chiplet-based architecture with an I/O hub in the center.

Fig. 5: IBM Power Processor roadmap – beyond Power11.

Using memory chips with SerDes clearly decreases beachfront and increases bandwidth. As features get smaller, the difference between on and off chip connectivity lines is 10x+. Chiplets also leverage packaging (2.5D interposer) technology.

The systems are buffered memory optimized and IBM has been working on this for decades. The SerDes run synchronously with the DDR on the buffer so that they can have extremely long predictable pipelines with several bypasses. The whole Power11 OMI memory is implemented in such a way that there’s only an extra 6-8ns of latency for going to memory. Starke stated that if you’re starting out with about 90-100ns of latency, the additional 6-8ns of latency isn’t really that noticeable and HBM doesn’t offer a high enough capacity now to fit their profile. For their target applications, capacity is really important, and IBM’s architecture enables this capacity in a very efficient manner.



Source link

Follow on Google News Follow on Flipboard
Share. Facebook Twitter Pinterest LinkedIn Tumblr Email Copy Link
Previous ArticleMIT Takes The Crown In 2025-2026 BlueSky Composite Ranking
Next Article AI in recruitment: OpenAI’s new jobs platform targets workers and small businesses
Advanced AI Editor
  • Website

Related Posts

IBM, AMD team up on quantum AI supercomputers

September 8, 2025

Shadow AI enters workforce, employees embrace AI adoption: IBM

September 6, 2025

IBM’s Strategic Focus in the Chinese Market Has Shifted, Multinational Tech Giants Favor AI Manufacturing_Dekkers_Yicai

September 6, 2025

Comments are closed.

Latest Posts

Tony Shafrazi and the Art of the Comeback

Basquiats Linked to 1MDB Scandal Auctioned by US Government

US Ambassador to UK Fills Residence with Impressionist Masters

New Code of Ethics Implores UK Museums to End Fossil Fuel Sponsorships

Latest Posts

Google’s Big Day In Court Wasn’t That Big

September 8, 2025

Sparkvia AI — Inside the First AI-Powered Writing

September 8, 2025

Perplexity AI makes its play for government use

September 8, 2025

Subscribe to News

Subscribe to our newsletter and never miss our latest news

Subscribe my Newsletter for New Posts & tips Let's stay updated!

Recent Posts

  • Google’s Big Day In Court Wasn’t That Big
  • Sparkvia AI — Inside the First AI-Powered Writing
  • Perplexity AI makes its play for government use
  • The Impact of AI Voice Assistants on Customer Interaction Trends » World Business Outlook
  • ASML takes €1.3B stake in Mistral AI

Recent Comments

  1. JoshuaAstEp on 1-800-CHAT-GPT—12 Days of OpenAI: Day 10
  2. секс-кукла on 1-800-CHAT-GPT—12 Days of OpenAI: Day 10
  3. slot 88 on 1-800-CHAT-GPT—12 Days of OpenAI: Day 10
  4. Daviddit on 1-800-CHAT-GPT—12 Days of OpenAI: Day 10
  5. ChesterSponi on Foundation AI: Cisco launches AI model for integration in security applications

Welcome to Advanced AI News—your ultimate destination for the latest advancements, insights, and breakthroughs in artificial intelligence.

At Advanced AI News, we are passionate about keeping you informed on the cutting edge of AI technology, from groundbreaking research to emerging startups, expert insights, and real-world applications. Our mission is to deliver high-quality, up-to-date, and insightful content that empowers AI enthusiasts, professionals, and businesses to stay ahead in this fast-evolving field.

Subscribe to Updates

Subscribe to our newsletter and never miss our latest news

Subscribe my Newsletter for New Posts & tips Let's stay updated!

LinkedIn Instagram YouTube Threads X (Twitter)
  • Home
  • About Us
  • Advertise With Us
  • Contact Us
  • DMCA
  • Privacy Policy
  • Terms & Conditions
© 2025 advancedainews. Designed by advancedainews.

Type above and press Enter to search. Press Esc to cancel.