According to multiple sources (shown below), AMD’s Radeon RX 7900 XT and 7800 XT, powered by the Navi 31 die will feature dual chiplets, in addition to an I/O die, resulting in an overall core count of 10,240 cores (5,120 per die). This will reportedly make the Navi 31 based RDNA 3 GPUs 2.5-3x faster than the RX 6900 XT. Furthermore, the ray-tracing performance is expected to be a notch higher than RDNA 2, being more or less on par with NVIDIA.
Navi 31+32 = 2 chiplets for compute, 80 CU each, 1 IO die.
Navi 33 = 80 CU, monolithic.
Navi 31 is about 2.5x faster than Navi 21. Similar RT perf to nvidia (though a tiny bit slower like for like). ML still handled as lower precision ops
However, before you make any conclusions, I’d like to draw your attention to previous Big GPU launches from AMD. Most of them fell far short of what the rumors had claimed, and it’s likely that the same will happen with this particular rumor as well. The Navi 31 GPU is likely going to be 1.5x to 2.5x faster than Navi 21 which is considerable for an inter-generational gain, and much more reasonable than the 3x gain quoted by multiple outlets.
Regardless of how this turns out, it’ll be interesting to see how AMD implements a chiplet design with its GPUs considering that they’re very sensitive to latency. SLI and XFX have failed for this very reason as it’s hard for games to work with a multi-GPU configuration due to the inter-frame dependency brought about by temporal upscaling and AA technologies.
Computer hardware enthusiast, PC gamer, and almost an engineer. Former co-founder of Techquila (2017-2019), a fairly successful tech outlet. Been working on Hardware Times since 2019, an outlet dedicated to computer hardware and its applications.
