As Generative Artificial Intelligence projects that use large language models (LLMs) transition from the experimental stage to everyday use, the urgent need for power-efficient, secured, and scalable technologies has emerged. To support such foundational models and AI workloads, IBM has unveiled a processor and accelerator to enhance enterprise-scale AI.
In August 2021, IBM announced Telum microprocessors, specifically designed for IBM’s Z mainframe computers. The Telum was IBM’s first processor to integrate AI capabilities directly into the hardware, improving processing speed and fraud detection. Now, IBM has revealed the specifications for its upcoming Telum II Processor and Spyre Accelerator, designed for better processing capacity across next-generation IBM Z systems that leverage AI in traditional models and LLMs.
Samsung Foundry will manufacture the Telum II processor and the IBM Spyre Accelerator using its 5nm node technology. They can potentially improve performance “up to 70% across key system components.” These chips are powered by eight high-performance cores that run at 5.5GHz, increasing frequency and memory capacity with a 40 percent growth in cache. The hardware also supports higher caching levels, L3 and L4, expanding to 360MB and 2.88GB, respectively.
Fifth Generation Snap Spectacles: Augmented Reality Redefined: IBM Announces Telum II Processors, Upscaling LLM and Generative AIIBM’s Telum II has Accelerator Units Integrated
Compared to its predecessor, Telum II’s key integrations are accelerators like the IO acceleration unit and IBM Spyre Accelerator. The new Data Processing Unit (DPU) on the Telum II processor chip simplifies system operations and improves component performance. Moreover, the Spyre Accelerator chip assists Telum II in calculating several possibilities and displays more accurate results than its companions.
With each accelerator chip attached via a 75-watt PCIe adapter and the combined compute powers, accelerators are expected to reach 24 trillion operations per second (TOPS).
“Our robust, multi-generation roadmap positions us to remain ahead of the curve on technology trends, including escalating demands of AI,” said Tina Tarquinio, Product Management, IBM Z. “The Telum II Processor and Spyre Accelerator are designed to deliver high-performance, secured, and more power-efficient enterprise computing solutions. After years in development, these innovations will be introduced in our next-generation IBM Z platform so clients can leverage LLMs and generative AI at scale.”
The Telum II processor and Spyre Accelerator utilize various AI models to better apply generative AI use cases. For instance, insurance claim fraud detection, anti-money laundering, or AI assistance in day-to-day activities like knowledge transfer or code explanation.
IBM has revealed that Telum II will be the core processor powering IBM Z platforms. Telum II and Spyre Accelerator are expected to be available in 2025.
