The introduction of the Telum-based system is planned for the first half of 2022.
IBM unveiled at the annual Hot Chips Conference the details of the upcoming new Telum processor, which aims to link in-depth learning capabilities to the corporate workload to help prevent fraud in real time. prevention. Telum is IBM’s first processor to include in-chip acceleration to draw artificial intelligence in-transaction conclusions. The breakthrough new chip-based hardware acceleration, which has been under development for three years, is designed to help customers scale business insights into banking, finance, commerce, insurance applications and customer interactions. The introduction of the Telum-based system is planned for the first half of 2022.
Today, businesses typically use post-fraud detection techniques, which is a time-consuming and computationally intensive process due to the limitations of today’s technology, especially when fraud analysis and detection takes place away from critical transactions and data. Due to delay requirements, complex fraud detection is often not feasible in real time, which means that a criminal could have already successfully purchased goods with a stolen credit card before the retailer could become aware of the fraud.
According to the U.S. Federal Trade Commission’s 2020 Consumer Sentinel Network Databook, consumers lost more than $ 3.3 billion in fraud in 2020, compared to $ 1.8 billion in 2019. Telum can help its customers shift their mindset from fraud detection to fraud prevention, from exploring many fraud cases today to potentially entering a new era of fraud prevention without affecting service level agreements (SLAs) before the transaction is completed.
The innovative centralized design of the new chip allows customers to leverage the full power of the processor for AI-specific workloads, making it ideal for financial services such as fraud detection, credit processing, trade clearing and settlement, money laundering and risk analysis. With these innovations, customers can apply rule-based fraud detection or machine learning that can speed up credit authorization processes, improve customer service and profitability, identify which transactions or transactions may fail, and propose solutions to create more efficient clearing processes.
Telum involves the joint creation and integration of hardware and software in accordance with IBM’s traditional design approach, covering silicon, the system, firmware, operating systems, and leading software frameworks.
The chip contains 8 processor cores with a deep, superscalar, out-of-order instruction path that operates at a clock speed of more than 5 GHz and is optimized for the needs of heterogeneous enterprise-class workloads. The completely redesigned cache and chip connectivity infrastructure provides 32 MB of cache per core and is scalable up to 32 Telum chips. The two-chip module design includes 22 billion transistors and 19 miles of wire on 17 metal layers.
Telum is the first IBM chip to be built with technology developed by the IBM Research AI Hardware Center. In addition, Samsung, as IBM’s technology development partner, has developed a 7nm EUV technology node for Telum.