IBM has unveiled the Telum II processor at the Hot Chips 2024 conference in Palo Alto, which is set to significantly enhance the performance of IBM Z systems expected in early 2025.

The Telum II processor features a built-in AI accelerator, designed to handle mission-critical tasks and demanding AI workloads, thereby supporting enterprise solutions.

Accompanying the Telum II is the IBM Spyre Accelerator, which offers up to 1TB of memory and operates with low power consumption, supporting AI workloads across multiple cards.

This new processor delivers an impressive 24 trillion operations per second (TOPS) with INT8 precision, providing four times the computational power of its predecessor.

The architecture of the Telum II has been optimized for high throughput and low-latency inferencing, supporting enhanced cache sizes that have increased by 40%.

IBM claims that approximately 70% of the world's transactions by value are processed through its mainframes, and the new technology aims to integrate generative AI into these critical operations.

The AI accelerator within the Telum II is engineered for real-time AI workloads, ensuring low latency and faster transaction processing.

The Telum II is equipped with eight high-performance cores running at 5.5GHz and features a total of 360MB cache, along with a 40% increase in L2 cache.

Additionally, the processor includes a virtual level-4 cache of 2.88GB per drawer and an integrated I/O Acceleration Unit DPU, which enhances data handling density by 50%.

These innovations are expected to enhance applications such as fraud detection and anti-money laundering, with availability projected for 2025.

The DPU is designed with four processing clusters, each containing eight microcontrollers, to streamline data transfers and improve power efficiency.

IBM's advancements in AI technology are focused on transforming workloads in data centers, allowing for secure management and deployment of AI models on-premises.