According to MLD‘s sources, NVIDIA’s upcoming Ada Lovelace GPUs for next-generation GeForce RTX 40 Gaming graphics cards could have an advantage over AMD’s upcoming RDNA 3 GPUs, using one of TSMC’s latest etching processes: 4N.
Nvidia RTX 40xx Ada Lovelace: made in 4N and not in 4nm
So be careful, the engraving processes have become commercial arguments, we must not confuse everything. Contrary to what one might think, TSMC 4N has nothing to do with 4nm and is in fact an evolution of their 5nm process. It seems that the TSMC 4N process is designed exclusively for NVIDIA and thus has a range of optimizations that allow for better energy efficiency, performance and a slight increase in density compared to the “classic” TSMC 5nm node. This choice of Nvidia obviously echoes the various rumors about the gargantuan consumption of the next RTX 40. On the AMD side, even if a mix of 5nm and 6nm processes will be used for its next GPUs based on the RDNA 3 graphics architecture, some of them will benefit from an MCM approach that should be very efficient. At Intel, leaks on this subject have also fueled questions with surprisingly low TDP. It should also be noted that this 4N process does not come out of nowhere since it is the one used by Green for the Hopper GPUs for the HPC market of data centers.