During AMD Financial Analyst Day, the company announced new details, such as its next-generation graphics cards AMD Radeon RX7000architecture-based AMD RDNA3will experience an improvement in performance 50% per watt consumed compared to the current generation of AMD RDNA2-based GPUs.
One of the main improvements to make this possible is to launch a new crafting process, the 5nm from TSMC, while the current generation is made by the same foundry but at 7nm. In addition, the company indicates that it used a new packaging system because its high-end graphics feature a chiplet design, the compute units have been redesigned, the graphics pipeline has been optimized, and a new generation of its memory is used AMD Infinite Cache, which will now have more capacity and would also have increased its bandwidth. That is to say, the AMD Radeon RX 7000s arrive to give war to Nvidia.
So far all the information provided by the company in graphics, in addition to announcing that next year, in 2023, the AMD Radeon RX 8000 architecture-based AMD RDNA4, which will achieve a more advanced manufacturing process. Unfortunately, this is the only information provided on its next generation of GPUs.
AMD CDNA3 GPUs will also use a chip design
On the other hand, the company’s line of professional GPUs based on the architecture CDNA3 will also achieve a manufacturing process of 5nmexcept that an internal design is already used here 3d chiplet where it will enable the company to create the first exascale APU.
Otherwise, with regard to this architecture for data centers, it will also use the new generation of AMD Infinity Cache memory, the 4th generation of the Infinity architecture, and the use of unified memory integrated into silicon to create these powerful APUs. . The company has not advanced any details on the predictable architecture of CDNA4.