IBM has unveiled its new 7nm Telum processor at the Hot Chips 33 conference. The processor is designed to bring deep learning inference to enterprise workloads, which can help address frauds in real-time. After three years of development, IBM has finally managed to create this breakthrough technology. The first system to feature the chip is expected to come out in the first half of 2022.
The chip is designed to help customers achieve business insights at scale across banking, finance, trading, insurance, and customer interactions. With the new processor, applications can run efficiently where the data resides. IBM says that this approach is far better than traditional enterprise AI approaches that tend to require significant memory and data movement capabilities to handle inferencing.
Since the new processor is now in close proximity to mission-critical data and applications, it will be easier for enterprises to conduct high volumes of inferencing for real-time sensitive transactions, without having to invoke platform AI solutions that might impact performance in the long run. IBM stated:
"The chip contains 8 processor cores with a deep super-scalar out-of-order instruction pipeline, running with more than 5GHz clock frequency, optimised for the demands of heterogenous enterprise-class workloads."
IBM further added that it has completely redesigned the cache and chip-interconnection infrastructure that provides 32MB cache per core, and can scale to 32 Telum chips, and that the new dual-chip module design contains 22 billion transistors and 19 miles of wire on 17 metal layers.
While IBM has developed the processor, it was actually built on the 7nm extreme ultraviolet technology by Samsung and is the first chip made by the IBM Research AI Hardware Center. The research center has also recently announced scaling to 2nm mode which is a benchmark in silicon and semiconductor innovation.