DailyPost 2750
ARE ORDINARY CHIPS OUT?
The onset of AI has been literally transforming everything, but what has been impacting the growth of AI at the backend. A hardware which is destined to define the growth trajectory of AI, its compute capabilities, the pricing of its products and services and varieties of its offshoots is the GPU. The question that needs to be answered is why ordinary chips are not good enough for the AI machines. The ordinary chip we are referring to, is our very old CPU (Central Processing Unit), which has ruled the roost, since computers became our mainstay. The computer brain as well it, performed basic arithmetic, logic, input/output operations specified by program instructions.
CPUs sequentially process instructions. This and for a variety of other technical reasons ranging from machine learning compute to gaming and video rendering GPU (graphics processing unit) has gained prominence. It has its roots from 3D graphics in video games. Real time 3D graphics rendering was a giant technological leap. From there to the ultimate AI enabler has been the journey of Nvidia, which has for long become synonymous with this chip. The importance of this chip grew as AI advanced. What is at the core of AI?
It is the training of AI systems requiring massive amounts of data that need to be processed instantly. The magic computing power which the GPU enabled in conjunction with the emergence of machine learning and deep learning techniques were the direct tech facilitators that helped create powerful AI solutions. Primarily serving the gaming industry Nvidia has been a dominant player since the 1990s. CEO Jensen Huang’s bet that GPUs could be used for AI workloads paid off. It seems Nvidia was waiting for this day. Their research and development exploited every AI compute opportunity thrown to the world. In 2019, Nvidia’s market-cap was at $111.35 billion. Nvidia’s market cap has shot almost 20 times to $2.24 trillion, making it the third largest company globally.
What sets the AI chips apart? “AI chips are specialised hardware components designed to efficiently process AI tasks.” It scores over the traditional CPUs in terms of parallel processing; critical to handling the massive computations, energy efficiency; being more power efficient than CPUs makes them ideally suited for AI workloads, high memory bandwidth; fast access to memory and customised architecture; purpose built for AI tasks allowing better performances. The research, growth, availability and pricing would have a direct impact on the AI revolution.
THE POWER OF THE GPU IS DIRECTLY PROPORTIONAL TO THE POWER OF THE AI REVOLUTION.
Sanjay Sahay
Have a nice evening.