According to wccftech and very well-identified components leaker @KittyYYuko, AMD’s upcoming-technology Radeon graphics playing cards should be a large upgrade—as considerably as 3 instances faster than AMD’s present flagship graphics cards.
2.5x is also tiny.
Consider a guess :p
— Yuko Yoshida (@KittyYYuko) May perhaps 2, 2021
In the previously mentioned reply to RedGamingTech’s Paul Eccleston, @KittyYYuko promises a 3x efficiency boost when upgrading from AMD’s flagship Navi 21 (as discovered in latest-gen RX 6800XT and RX 6900 XT cards) to Navi 31.
RDNA3—the architecture beneath the new GPUs—will shift approach-node size from 7 nm down to 5 nm, which tends to end result in big performance gains and even bigger general performance-per-watt gains. The downshift in course of action node isn’t adequate to account for the claimed efficiency uplift by itself, though—that arrives from earning the GPU by itself appreciably larger and meaner.
The current flagship, Navi 21, offers 80 compute units and 5,120 shading units. Navi 31 will reportedly double these numbers, providing a whopping 160 CUs and above 10,000 shading models, delivered in two 80 CU chiplets on a solitary die. This is probably only practical because of to the downshift in approach size—cooling and power provide are significant worries in flagship GPUs, which have a tendency to call for as significantly of the two as OEMs and personal builders can easily cram into a circumstance.
RedGamingTech indicates that Navi 31’s ray-tracing effectiveness need to be roughly on par with Nvidia’s Ampere GPUs, but device-finding out tasks will even now be handled by using reduced-precision operations on the major GPU by itself alternatively than on focused hardware like Nvidia’s Tensor Cores.
We hope RDNA 3 playing cards to become obtainable sometime in late 2022 or early 2023.