Home / News EN / AMD OpenAI: GPT-optimized chips deliver 14% PPA improvement

AMD OpenAI: GPT-optimized chips deliver 14% PPA improvement

AMD OpenAI. AMD has announced a collaboration with OpenAI that led to a 14% improvement in its chip performance. The optimization, achieved through the use of AI models, concerns the PPA (Power, Performance, Area) of future graphics processors and, to a lesser extent, CPUs as well.

AMD OpenAI

The partnership between the two companies involved using GPT models to directly analyze the code used in integrated circuit design. This approach allows for identifying potential optimizations in power consumption, performance, and chip footprint, aspects crucial for manufacturers constantly seeking to improve product efficiency.

AMD OpenAI: why it matters

PPA is a fundamental parameter in hardware design. An improvement in this area does not automatically translate into a proportional increase in video game performance, but it contributes to creating more efficient and high-performing chips, offering manufacturers greater flexibility in design.

AMD has extended the use of these models to its processor development as well. Although the gains obtained are lower than for GPUs, the company highlights how this technology can significantly accelerate product development cycles. OpenAI estimates that the work completed in less than five months would have taken approximately one year using traditional methodologies.

The company highlighted the importance of this innovative approach, emphasizing how artificial intelligence could revolutionize hardware design. Automating part of the traditionally complex and time-consuming design process could lead to reduced development times and the market release of more advanced components.

AMD expects other industry players, such as Intel and Nvidia, may adopt similar strategies to improve their product performance. However, no official information has been released by these companies regarding this matter at present.

What is PPA in chip design?

PPA (Power, Performance, Area) is a set of three metrics that measure power consumption, performance, and the physical space occupied by a chip, respectively. Manufacturers constantly seek to optimize these three aspects, often having to face trade-offs between them.

How can AI like GPT help design physical chips?

AI models can analyze and directly modify the code used for circuit design, identifying patterns and optimizations that would be difficult to spot manually. This allows part of the hardware design process, traditionally highly specialized and slow, to be automated.

Does a 14% PPA improvement translate to more FPS in video games?

Not necessarily. PPA measures chip design efficiency, not direct gaming performance. A PPA improvement can lead to chips that are more energy-efficient or compact, but the final impact on FPS depends on many other hardware and software factors.

* Content created with the assistance of artificial intelligence systems.