New Processor is Introduced by Graphcore For AI Hardware Supremacy

The UK-based AI chip creator company Graphcore has uncovered a novel set of hardware and software innovations to expand the walls of research and development in the core domain of Artificial Intelligence. The essence of the second generation of the flagship Intelligence Processing Unit (IPU) chip was launched by the company which is GC200 or the Colossus MK2.

 

In accordance with the Graphcore, GC200 is the most compact processor that is made till date, the IPU chip is situated at the gist of each IPU-Machine M2000, and the plug-and-play Machine Intelligence to calculate edge that has been developed for smooth deployment and to promote systems and operations that can expand over high-scale. 

 

The Senior Analyst at Moor Insights, Karl Freund affirmed that these sorts of developments place Graphcore ahead in the lineup to NVIDIA challenges concerning datacenter AI.   

 

GC200 combines 1,472 isolated IPU-cores and proficient enough to execute 8,832 individual parallel estimated threads of connection. Every single IPU processor core obtains a performance hike out of an array of floating-points technologies designed through Graphcore, known as AI-Float.

 

The chip creator also professed to produce around 250 Trillion Operations/second (TOPS) over 1,472 cr and 900MB of In-Processor Memory interlinked crossed a 2.8Tb/s less-latency stuff. Each chip comprises of almost 59.4 billion transistors over a single 823 sqm die that is designed with the TSMC’s recent 7nm processing technology.

 

The M2000, IPU-machine is accessible or might be acquired from data center pods that enable the connection of 64k IPU chips for controlling the most difficult tasks.