IBM’s z17 mainframe has the processing power to deliver 50% more AI inference operations daily.
IBM has unveiled the IBM z17 mainframe, a leap forward in its enterprise technology portfolio, engineered with AI at its core. Built around the IBM Telum II processor, the z17 extends beyond transactional AI, introducing robust new workloads that empower businesses to innovate and achieve more.
The IBM z17 can process 50% more AI inference operations daily than its predecessor, the z16. The z17 offers tools for mitigating loan risk, enhancing chatbot interactions, advancing medical image analysis, and preventing retail theft, among many others. Developed in collaboration with over 100 clients and drawing on insights from IBM Research and Software teams, the z17 embodies five years of research and development. Its design features more than 300 patented innovations that emphasize enhanced performance, security, and system usability.
IBM Telum II
At the heart of the z17’s AI capabilities is the on-chip accelerator integrated into the IBM Telum II processor. With increased frequency, enhanced compute capacity, and a 40% growth in cache, the processor supports over 450 billion inference operations per day. The addition of the IBM Spyre Accelerator, expected to be available in late 2025, will further enhance AI compute capabilities.
The on-chip and PCIe-attached AI acceleration is delivered via the IBM Telum II processor and the Spyre Accelerator card. This combination improves high-speed, low-latency inferencing and extends generative AI and LLMs to enterprise speed and scale. For small language models (SLMs), the Telum II allows up to eight billion parameters.
Additionally, the Telum II features a new data processing unit (DPU) designed to accelerate complex input/output (I/O) protocols for networking and storage in the mainframe. The DPU simplifies system operations and enhances the performance of key components.
Beyond hardware, IBM z17 incorporates advanced software features to improve the efficiency of developers and IT operations. AI assistants and agents, such as IBM watsonx Code Assistant for Z and IBM watsonx Assistant for Z, will assist in streamlining incident detection and resolution. Integrating these tools with IBM Z Operations Unite provides a new level of AI-driven operational insight. With z/OS 3.2, slated for release in 2025, the operating system will introduce hardware-accelerated AI capabilities, support for NoSQL databases, and expanded hybrid cloud data processing. This enables more accessible enterprise data, facilitating predictive analytics and business insights.
Security remains a cornerstone of the IBM z17’s design. The system includes AI-powered tools for discovering and classifying sensitive data, ensuring critical assets are safeguarded against evolving cyber threats. IBM Vault, a comprehensive secrets management solution, standardizes security across hybrid environments, while new AI-driven solutions provide anomaly detection and threat identification capabilities, further strengthening the platform’s defense posture.
| IBM z17 Specifications | |
|---|---|
| Maximum number of engines | 208 |
| Maximum number of drawers | 4 |
| Maximum number of IO drawers | 12 |
| Number of frames | 4 |
| Colocate with storage/switch | No |
| Frequency | 5.5 GHz |
| Telum chip | Yes |
| Maximum memory | 64 TB |
| Sizes | 43, 90, 136, 183, 208 |
IBM z17 also enhances the support experience through IBM Technology Lifecycle Services. These services optimize performance, address risks, and streamline incident remediation, leveraging IBM watsonx technology for improved resolution times. In addition, the platform integrates seamlessly with the IBM Storage DS8000 series, offering secure, agile, and high-performance storage solutions that align with the demands of mission-critical workloads.
Available on June 18, the IBM z17 represents a transformative shift in mainframe technology, combining hardware, software, and AI to deliver unprecedented business value. For more information, visit IBM.com/z17.
Engage with StorageReview
Newsletter | YouTube | Podcast iTunes/Spotify | Instagram | Twitter | TikTok | RSS Feed
