At the meeting of financial analysts last week, amd was full of dry goods and announced the 5nm zen4 architecture. At the same time, it also launched a new generation of rdna2 architecture, which increased energy efficiency by 50% compared with the first generation of rDNA, which can be called the biggest change of AMD GPU in ten years.
In addition, amd has also confirmed the next generation of rdna3 architecture, but the details of rdna3 architecture are not available, and the process technology is simply defined by the fuzzy term “advanced node”.
What process will rdna3 architecture use? In response to an analyst’s question, amd CEO Su Zifeng explained the reason why it did not publish the process. Su said that it was not yet time for AMD to disclose the technology in a period of time before the release of rdna3 graphics card.
Considering that the rnda2 video card has not been officially announced, it seems a little early to make sure that rdna3 is really a little early.
Another reason may be related to the previous dispute. Amd announced rnda2 and zen3 too early. At that time, the 7Nm + process was mentioned in the roadmap, which led to a misunderstanding that it was TSMC’s second-generation 7Nm + EUV process.
Results at the meeting a few days ago, amd explained that 7Nm + does not represent the EUV process.
Of course, it is also possible that AMD planned to use the 7Nm + EUV process before, but now it is still not cost-effective, so it is not planned to use it.
In the rnda3 architecture, amd should have learned this lesson, avoid publishing too detailed parameters too early, and leave some room for itself.
However, from the perspective of progress, the GPU chip based on rdna3 architecture should also use 5nm technology, just like 5nm zen4 in 2022. After all, AMD will build 20 billion supercomputing times with future CPU and GPU. At that time, it should be the next generation RX 7000 Series graphics card.
Editor in charge: WV