According to a patent that was recently published by AMD, the Company designs multi-block and multi-core graphics cards. This generation of GPUs will put an end to monolithic graphics cards and would put AMD ahead of competitors such as Nvidia.
MCM project for graphics cards
AMD appears to be on the verge of changing the design of its graphics cards. According to the recently published patent, AMD is on a project allowing different graphics GPUs to work together. It is therefore an MCM (multi-chip module) project like it did for its processors. It consists of producing models that can work from components previously produced separately. This is the case, for example, for its consumer Ryzen processor models and EPYC servers. AMD thus produces a chip composed of different blocks that can be separated from each other (memory controllers, cores, etc.). The number of cores can be increased as desired via a fairly advanced design and production technique. The different blocks are interconnected via very fast speed buses. These types of products are breaking into the market but they are also economical and efficient. The production of MCM processors has allowed AMD to get ahead of its competitor Intel in different areas. It will therefore try to reproduce its feat to beat its direct competitors (Nvidia, etc.)
End of monolithic chips?
This may be the end of the reign of monolithic chips. The two graphics card giants (AMD and Nvidia) have had the design of MCM graphics cards in their objectives for quite some time. We will therefore see which of the two will manage to produce one first. In any case, this recent patent from AMD confirms its good progress in the field. This patent is in fact titled “GPU Chiplets using high-speed cross-links”. The Company led by Lisa Su provides in the documentation the reasons why AMD had not yet started to realize this project. Some of these reasons were the existence of a communication latency between the different blocks and the difficulty of implementation in terms of parallelism. These problems were able to be circumvented by AMD engineers. To do this, they set up an internal high-speed communication network called: “high bandwidth passive crosslink”. This network allows each block to communicate with the others, but also with the CPU. In addition, each GPU would have its own caches and all the elements necessary for it to operate autonomously. Each block would be manageable directly at the operating system level.
The design of these new graphics cards differs a little from that of processors. In fact, the cores of a processor are put into a block with a single input/output device. The graphics card GPUs will be small in size, can connect to each other and work together. This new solution will become reality according to rumors after the RDNA3 generation which will be produced this year and next year. We can therefore have the first cards with new architecture around the year 2023.
In short, the patent which was recently posted online suggests that AMD is committed to this project of producing graphics cards with different interconnectable blocks. The cores will be able to work together, communicate with each other and with the CPU. These blocks will be directly manageable at the operating system level. This project seems to be underway and will be effective after the RDNA3 generation currently in production.
