IBM’s new Telum Processor is the company’s first with an on-chip AI accelerator

The microprocessor comprises 8 processor cores and is designed to course of delicate knowledge in hybrid cloud deployments, in line with the corporate.


Picture: GettyImages/gremlin

The brand new Telum Processor would be the central processor chip for the subsequent technology IBM Z and LinuxONE methods, the corporate introduced Monday. The processor has a devoted on-chip accelerator for AI inference and the design improves efficiency, safety and availability. The corporate introduced the information on the Scorching Chip convention.

Extra about synthetic intelligence

The microprocessor comprises 8 processor cores, clocked at greater than 5GHz, with every core supported by a redesigned 32MB personal level-2 cache. The extent-2 caches collectively kind a 256MB digital Stage-3 and 2GB Stage-4 cache, in line with the corporate. 

The 1.5x progress of cache per core over the z15 technology is designed to allow a major improve in each per-thread efficiency and complete capability, in line with the corporate. The corporate stated in a press launch that Telum’s efficiency enhancements will enable fast response instances in advanced transaction methods, particularly when augmented with real-time AI inference.

SEE: World chip scarcity: How producers can cope over the long run (TechRepublic)

Telum additionally options safety enhancements together with clear encryption of essential reminiscence. Telum’s Safe Execution enhancements are designed to supply elevated efficiency and value for Hyper Protected Digital Servers and trusted execution environments, equivalent to processing delicate knowledge in hybrid cloud deployments.

The predecessor IBM z15 chip was designed to allow seven nines availability (solely 3.16 seconds of downtime per yr) for IBM Z and LinuxONE methods. IBM stated Telum’s redesigned 8-channel reminiscence improves on that chip’s availability on account of an interface able to tolerating full channel or DIMM failures.

The chip is the primary one created by the IBM Analysis AI {Hardware} Heart. The corporate expects monetary companies to make use of the chip to stop fraud as a substitute of merely detecting it.

NVIDIA’s AI chips

NVIDIA introduced an “AI-on-5G” machine in June to increase assist for Arm-based CPUs within the NVIDIA Aerial A100 platform. The computing platform makes use of the NVIDIA Aerial software program growth package and can incorporate 16 Arm Cortex-A78 processors into the Nvidia BlueField-3 A100. 

SEE: World chip scarcity is hitting near dwelling (TechRepublic)

In keeping with the corporate, this can be a self-contained, converged card that delivers enterprise edge AI functions over cloud-native 5G vRAN with improved efficiency per watt and quicker time to deployment. BlueField-3 is a next-generation knowledge processing unit constructed for AI and accelerated computing and optimized for 5G connectivity. The machine contains an NVIDIA A100 GPU, a BlueField 2 knowledge processing chip and a processor, both ARM or x86.

The corporate introduced its A100 AI chip in June 2020. It has 54 billion transistors and may execute 54 petaflops of efficiency. That chip is the third technology of the corporate’s’ DGX platform.

Additionally see

Recent Articles


Related Stories

Leave A Reply

Please enter your comment!
Please enter your name here

Stay on op - Ge the daily news in your inbox