Both Nvidia and AMD will release a new generation of GPUs this year. The former has attracted everyone’s attention as the NVIDIA GTC Conference is approaching, as well as many revelations and even physical pictures of GeForce RTX 40 series graphics cards. Although the latter’s Radeon RX 7000 series graphics cards will come later, and there is not much recent news.
AMD Radeon RX GPU with RDNA3 demo, Source: AMD
Sasa Marinkovic, director of gaming marketing at AMD, recently
tweeted on his Twitter account, said that
“bigger doesn’t necessarily mean better.” Although the content does not mention specific products, it is speculated that it should not be related to the Ryzen 7000 series desktop processors, but may be related to the Radeon RX 7000 series GPUs based on the RDNA 3 architecture.
Some analysts pointed out that Sasa Marinkovic’s remarks may point to a large number of recent leaks involving GeForce RTX 4090 graphics cards. The so-called larger may point to the fact that this graphics card has a larger-scale cooling system, resulting in an oversized volume with a thickness of four slots. AMD’s cooling solution seems to be more friendly. It is rumored that the new graphics card is a 2.5-slot thickness, still a three-fan configuration, and only equipped with two 8Pin external power supply interfaces.
Sam Naffziger, AMD senior vice president, corporate researcher, and product technology architect, said in an interview with the media that
the total power consumption of the next-generation GPU will increase. However, combined with Sasa Marinkovic’s statement, it seems that the efficiency may be higher than the competition. According to AMD’s official statement in the past, the performance per watt of the RDNA 3 architecture will increase by 50%.
The Radeon RX 7000 series is notable for its Navi 3x series GPU, which is the first time that a small chip design has been introduced on a consumer graphics card. Among them, Navi 31 and Navi 32 will be packaged in MCM multi-chip, equipped with one GCD (Graphics Computing Chip) and four or six MCDs. At the same time, two different processes will be used, the former is TSMC’s 5nm, and the latter is 6nm.
