Developed by TSMC, a system that will allow two GPUs, either from NVIDIA or AMD, to be introduced in the same package using a special stacking technology.
The scalability of the manufacturing processes or lithographs of the processors and graphics cards is reaching a quite complex point, where making the leap costs much more than it had been anticipated, we just have to see how Intel is constantly delaying 10nm due to the problems of mass producing these chips. Taiwan Semiconductor Manufacturing Company, also known as TSMC, has developed a technology that will increase the power of NVIDIA and AMD graphics, without the need to physically make them larger.
This exciting new technology has been dubbed wafer-on-wafer and is based on, or mimics, 3D NAND technology, which is used in current solid-state memory. Basically it is stacking several layers vertically, instead of distributing all the components horizontally on the printed circuit board, something that would involve taking up more space.
Thanks to this design, the number of transistors in a GPU could be increased, but logically, this would cause an increase in consumption and therefore heat generation, so you have to play with these aspects. We may see an increase in the number of transistors and an increase in performance, but not as big as one might expect, since we must try to find a balance between performance and consumption.
Currently, NVIDIA and AMD GPUs are based on a single silicon, but TSMC has figured out how to stack two silicons in a single package. The lower silicon does not undergo changes, but that second silicon, is flipped or what is the same, they are put 'face to face' and then joined. The upper silicon would have the input and output connections, unchanged, therefore a dual system is created using flip-chip technology.
Source: DT
