AMD CEO Lisa Su expressed her thoughts on the future development of the semiconductor industry and the prospects for creating new processors for high-complexity computing tasks. She believes the current model of using graphics processing units (GPUs) to perform tasks has exhausted its potential.
Although GPUs were initially good at large-scale computing and parallel tasks, they have limitations in their programming capabilities. So, they can no longer be the best solution for the future. Lisa Su believes that over the next five years, there will be an active transition to the use of specialized chips for artificial intelligence (AI) and computing tasks that will replace GPUs.
The main reason for this transition is the need for a stable and specialized solution. In the case of GPUs, each new generation is becoming more complex in terms of heat dissipation and power consumption, which is associated with an increase in the size and density of the crystal. This approach is only acceptable in the short term, but with artificial intelligence and big computing becoming an integral part of the future, there is a need for more optimal solutions.
Many tech companies, such as Google and Amazon, have already started developing their specialized chips for internal use. Other manufacturers, such as Broadcom, as well as ARM companies, are also striving to create their specialized solutions.
Today, GPUs are a versatile tool for a variety of tasks, from 3D graphics to artificial intelligence computing. But this very versatility prevents them from being the most efficient. That's why the market is moving towards the specialization of computing accelerators.
It is also important to reduce dependence on a single leader, Nvidia, which currently dictates the pace of the industry and controls the GPU market.
Lisa Su noted that to further develop the field of artificial intelligence, it is necessary to create alternatives in order not to slow down progress and avoid overdependence on one supplier.
