It is normal that we see a large number of advances created by the most important companies in the technology sector, and they are the ones who end up deciding what the key aspects will be, for example, the new components that we can find for a computer. In this case since Intel They want to offer a new type of architecture for GPU that would allow us to achieve a higher level of performance, as well as energy efficiency, and that is because a patent that would reveal what the company’s plans are related to these components.
Being one of the main technology companies on the market, Intel presents a large number of new solutions every year that offer really interesting changes to the technologies that we can currently find, although it does not imply that all of them will reach the market, as we can well imagine. Even so, there are many patents that we can find both from this company and from others that also share the same technological challenges, seeking to offer new capabilities to improve the performance and efficiency of each and every one of the components that we can find in a computer.
From Intel they know where the future of GPUs is
One of the considerations that Intel would be considering to use in your GPU It is a disaggregated architecture, and the company has filed a patent that would offer the ability to change the typical monolithic structure they have to adopt one based on chiplets specialized. In the event that the brand chose to adopt a design of this type, we could find a series of really big improvements in the GPU designs, and this implementation would greatly improve all the aspects that these components can offer. .
The approach to this design involves, as we have mentioned, using small chiplets that are then interconnected with each other, using the company’s technology to divide the GPU design into said components, allowing the manufacturer to adjust each of these chiplets for a specific use. (graphics, AI…). Among other things, this type of architecture stands out for having the capacity to be much more efficient at an energy level, and the fact that it is divided into individual “fragments” allows them to be activated depending on the use, that is, they can turned off to save energy, something that allows us to see that this type of designs are a reference for the future.
Although one of the most curious things that we can find about this type of patent that focuses on the use of chiplets and how individual units are managed in different modes is quite similar to the one that AMD presented with the same objective, seeking to offer a structure of a multichiplet module. This would imply that both companies have recognized this architectural design as one of the most important segments for GPUs, and we could be facing a race between the two companies to see who of the two manages to develop it first.
