Companies in the technology sector tend to plan and develop for the long term, so AMD is already working on RDNA 4the future architecture that will presumably be used in the generation of Radeon RX 8000 graphics. According to the first information available on RDNA 4, which would be better to take with a grain of salt considering the dates we are on, they suggest that it would provide an improvement in the 50 to 60 percent higher performance per watt than its predecessor, RDNA 3, which is implemented in the RX 7000.
According to the YouTube channel RedGamingTech, the source of the information that is circulating around RDNA 4, this will introduce some changes such as the presence of a single Graphics Computing Dice (GCD) that would be divided into one or more GCX. Navi 41, future flagship model and top of the range (at least initially), could be made up of 64 or 128 Logical Arithmetic Units per Computing Unit (CU), in addition to supposedly having 48 CUs and 3 shader engines ( SE) for each GCX.
Chips based on the Memory Cache Dice (MCD) aim to stay, while the Memory type would be GDDR7 (at least for top-of-the-range models), PCIe 5 interface and would use TSMC’s 3 and 4 nanometer manufacturing processes. The size of the cache and the Infinite Cache are two fronts that will presumably see their capacities increase, but the filter has not provided data in this regard. On the other hand, the second version of the Wave Matrix Multiple Accumulation (WMMA) Instructions would aim to speed up the performance per CU for certain instructions up to double it.
As we have already said, one of the alleged outstanding improvements is the 50 to 60 percent performance per watt improvement compared to RDNA 3 and RDNA 4 it would be able to exceed 3.5GHz. Navi 41, which would become the main graphic to use the future architecture, would incorporate a total of 48 CUs per GCX and there would be a total of three GCXs per GCD, resulting in a total of 144 CUs. versus the 96 CUs of the RX 7900 XTX. For its part, the shader count in Navi 41 would stand at 18,432.
As for the VRAM, AMD would be considering two configurations for Navi 41, one with four MCDs with 32GB of GDDR7 memory and another with six MCDs and variants with 24 and 48 gigabytes of GDDR7. Navi 42 would aim to be a more “modest” graphic with two GCXs to presumably add 96 CUs, 12,228 shaders and a total of 4 MCD. Finally, Navi 43 would have a single GCX for a total of 48 CUs and two MCDs with GDDR7 memory.
It is still too early to confirm anything, but seeing the bittersweet premiere of the RX 7900 XT and RX 7900 XTX graphics, surely many will perceive all this information more as smoke than anything else. Let’s hope that AMD and Intel get their act together, since NVIDIA begins to give the feeling of winning only by inertia.
