Home Z Telum IBM Telum processor family
Integrated AI accelerators for near real-time analytics and enhanced performance on the IBM Z platform
See what's new Register for the webinar
Overhead shot of person working with components featuring IBM Telum Processor

The Telum® processor is an advanced processor that powers many of the features and benefits of the IBM z16®. It is designed to help customers achieve business insights at scale across banking, finance, trading, insurance applications and customer interactions.

The IBM Telum processor represents a significant advancement in microprocessor technology for IBM z16. Designed to enhance enterprise workloads, the Telum processor integrates deep learning inference capabilities directly into the hardware, enabling real-time fraud detection and improved decision-making processes. This next-generation chip showcases innovations, setting a new standard for enterprise computing by delivering unprecedented speed and efficiency for complex data environments through:

  • High processing power
  • Accelerated AI integration
  • Advanced security
  • Energy efficiency
  • High availability
Download the Operationalizing fraud prevention Celent report
Introducing IBM Telum II Processor and Spyre Accelerator

Welcome IBM Telum II processor and Spyre Accelerator™1, built for high-speed inferencing. Get faster insights, solutions and time-to-value across your business with AI acceleration built to run where it is needed most—at the source of your data. 

Built to deliver deep learning and foundation model capabilities by using a multimodel ensemble method that focuses on performance and accuracy, this next-generation AI acceleration allows clients to use increasingly larger, more complex models for in-transaction workloads as well as certain generative AI use cases. 

Meet the processors

First generation Telum The IBM Telum processor is a 7nm microprocessor designed to deliver AI-driven insights without sacrificing response time in high-volume transactional workloads. It features 8 cores, each clocked over 5GHz and supported by a 32MB private L2 cache, which combines to form a 256MB virtual L3 and 2GB L4 cache. With 1.5x more cache per core than its predecessor, Telum boosts per-thread performance and overall capacity, helping ensure rapid response times for complex transactions, especially when enhanced by real-time AI inference. Read the blog

Second generation Telum II The IBM Telum II Processor offers a built-in low-latency data processing unit (DPU) for accelerated IO. The main processor contains 4 interconnected core clusters, each with 8 programmable micro-controller cores. This helps ensure there is cache coherency among the 32 cores. Also, with 10, 36 MB Level-2 caches on the chip, one per core, one for the DPU and the 10th as the overall chip cache capacity(2). Also, there is a larger pool of virtual L3 and virtual L4 cache that are also turned into meaningful performance improvements. Each has grown by 40% in Telum II, to 360 MB and 2.8 GB(3). Read the blog

Processor features Accelerated AI integration

The Telum processor includes on-chip AI acceleration, allowing for near real-time analytics and decision-making directly on the processor without the need for another hardware. On-chip AI acceleration enables real-time data analysis and decision-making, allowing businesses to gain immediate insights and respond proactively to emerging trends and issues.

Telum II enhances AI acceleration and inferencing with 4x the available compute power of the Telum processor and coherently connected processing clusters4.

High availability

The Telum processor's reliability and high availability help ensure continuous operation with minimal downtime, which is crucial for maintaining business continuity in mission-critical applications.

The Telum II offers redundant connections to processors allowing for interruption-free service to maintain operations through the attached IO via an alternative path while a processor is going through a repair action.

Low latency

The Telum processor’s architecture is optimized for low latency, which is crucial for high-speed transaction processing and real-time analytics.

Telum II increases overall available cache by 40%, in both the number of L2 (10) and the size of L2, L3 and L4 cache. All demonstrating improved off-chip bandwidth or latency5.

High processing power

The Telum processor is designed for high-performance computing, enabling the IBM z16 to handle large-scale transactions and analytics efficiently. The high processing power and low latency provided by the Telum processor result in faster transaction processing and more efficient handling of large-scale data analytics, enhancing overall productivity. 

Telum II is projected to clock in at 5.5 GHz, offering improved speed while still delivering efficient management of large-scale transactions6.

Advanced security

The Telum processor supports advanced encryption techniques, including quantum-safe encryption, helping ensure robust data protection. The support for advanced encryption techniques, including quantum-safe encryption, helps ensure that sensitive data is protected against both current and future cyberthreats, providing peace of mind and compliance with regulatory standards.

Offering improved security on A and Z bus links, Telum II is designed to be one more element in protecting your business7.

Energy efficiency

The design of the Telum processor optimizes power consumption, contributing to the energy efficiency of the IBM z16. The energy-efficient design of the Telum processor helps reduce operational costs associated with power consumption, contributing to overall cost savings.

Energy efficiency is further improved with Telum II with up to 15% core power reduction based on centralized IO and DPU design8.

Spyre Accelerator
PCiE chip Spyre Accelerator Delivered via a 75W gen 5x PCIe-attached AI accelerator with 128 GB of LPDDR5 memory. IBM Spyre Accelerator is geared to handle larger, more complex AI use cases, including Generative AI and multimodal LLMs, with 32 (+2) cores, 2 MB scratchpad on each and >55% core utilization. Scalable by card and drawer, clients are able to take advantage of this efficient solution for next-generation AI challenges. Read the blog
Take the next step

Discover IBM z16—built with first-in-industry on-chip acceleration Telum processor, designed for real-time AI inferencing to help identify fraud.

Explore IBM z16
Footnotes

The IBM Spyre Accelerator is current in tech preview.

2, 3, 4, 5, 6, 7, 8 All projections for Telum II are based on pre-release hardware measurements and configurations.