AMD’s Radeon RX 7000 GPUs are going to be some of the most energy-efficient parts we’ve ever since in the industry. Featuring a chiplet design with up to seven dies, including a 5nm GCD (Graphics Compute Die) and six 6nm MCDs (Memory Complex Dies), the RDNA 3 graphics architecture promises a 50% performance/watt uplift over RDNA 2.
In addition to new process nodes, the next-gen Radeon GPUs will also feature 3D packaging (MCDs and Infinity Cache), wider Compute Unit (128 ALUs), and WGPs (256 ALUs), and a fine-grained graphics pipeline.
According to Moore’s Law is Dead, the Radeon RX 7900 XT is going to draw well below the 450W limit (allegedly) set for the NVIDIA RTX 4090, the next-gen GeForce flagship. Tom’s sources have confirmed that although some board partner models will have an upper power limit of 400W, the reference model should stay under 400W, possibly between 350W to 375W.
Furthermore, it looks like AMD has given up on making the absolute fastest GPU. Instead, Team Red is focusing on power efficiency and upgradability/accessibility.
Meanwhile, NVIDIA’s RTX 40 series SKUs will reportedly draw up to 800W of power, and at least 660W for certain. The latest rumors from various sources claim that the RTX 4090 will have two TGP modes: 450W and 660W. The former will be used on the Founders Edition variant and the latter on factory overclocked, possibly liquid-cooled models. Some high-end SKUs will come with a BIOS switch to flip between the two modes, much like existing RTX 3090 flagships.