The Hot Chips 2024 conference was held in Palo Alto, where IBM unveiled a new processor and AI accelerator. The innovative products were designed mainly for enterprise applications and performance improvements for AI tasks.
The new IBM Telum II processor has 8 cores at 5.5 GHz and operates on 5-nanometer technology. The processor is equipped with the latest AI accelerator and data processing unit (DPU).
The main purpose of creating the latest processor was to improve performance, if compared to previous models. Telum II will have 4 times higher computing power, which will be - 24 TOPS. The accelerator architecture is fully optimized to work with a variety of AI models and large language models.
The INT8 data type will also be supported, which will significantly impact the computational capabilities of the processor. The overall performance of Telum II will be able to reach 192 TOPS, such performance became possible due to more efficient load balancing between 8 AI accelerators in the module.
The new Spyre accelerator will have 32 cores, similar to those in the Telum II. It will be possible to connect multiple powerful accelerators to the I/O subsystem to accelerate AI tasks.
Both of these latest products are mainly focused on different scenarios of using artificial intelligence, and they are scheduled for release in 2025.
