Imagine this, every time you swipe your credit card for a purchase, the system is already checking to see if it’s fraudulent. Imagine how much more convenient life would be with these instant transactions.
IBM has announced their new IBM Telum Processor, a CPU chip that can now facilitate deep learning inference at an unprecedented scale. This is due to on-chip acceleration for AI inferencing which could lead to breakthroughs in combating fraud, credit approval, claims, and settlements or financial trading with systems quick enough so as not to interfere with transaction speed.
To take advantage of AI, it is essential that you have a heterogeneous system with both CPU and AI cores tightly integrated on the same chip. This will allow your application workflows to stay low-latency for very rapid inference, so they don’t lose any efficiency in their calculations or data processing.
Telum builds AI cores with the power to run deep learning workloads while still splitting time between general-purpose software applications. The tight coupling of these two types of cores allows for fast data exchange, which is a fantastic feature in big business endeavors where speed and efficiency are key factors.
Telum’s full-stack approach to system design is an innovative way of solving the most complex problems. This means Telum can process tens of thousands of transactions infused with AI every second and do it for clients who need this performance level in their critical workloads.
This new chip contains eight processor cores, running at more than 5GHz clock frequency. The wholly redesigned cache and chip-interconnection infrastructure provide 32MB of cache per core for enterprise-class workloads, with a total of 22 billion transistors and 19 miles of wire on 17 metal layers.
Using Telum, banks will be able to prevent fraudulent activity and catch instances of fraud while the transaction is still in progress. The dedicated AI core in Telum makes future IBM systems ideal for deploying artificial intelligence across multiple industries that deal with payment processing, clearing trades, detecting money laundering activities, and assessing risk analysis situations. Banking isn’t the only industry that could benefit from this revolutionary technology. There is a possibility that its use cases will extend to other fields, such as natural language processing, computer vision, and speech recognition programs.