TapTechNews August 27th news, IBM announced in a press release yesterday (August 26th) that for AI application scenarios, introduced the next-generation Telum II processor and Spyre AI accelerator for the latest IBM Z mainframe system.
This processor is equipped with 8 high-performance cores, with an operating clock frequency of 5.5 GHz. Each core has 36 megabytes of L2 cache, and the on-chip cache capacity is a total of 360 megabytes, which is 40% more than the previous generation.
Each processor's virtual Level-4 virtual cache is 2.88 gigabytes, which is 40% more than the previous generation.
Integrated AI accelerators can achieve low-latency and high-throughput AI inference in transactions, such as enhancing fraud detection in financial transactions. The computing power per chip is four times higher than that of the previous generation. TapTechNews attached relevant pictures as follows:
This is an enterprise-level accelerator specifically designed to provide scalable capabilities for complex AI models and generative AI use cases.
It has up to 1 terabyte of memory and can work in series on eight cards in a regular IO drawer to support AI model workloads on the mainframe, while the power consumption of each card does not exceed 75 watts.
Each chip will have 32 computing cores and support int4, int8, fp8 and fp16 data types, which can be used for low-latency and high-throughput AI applications.