GamingGPUsNews

AMD Reportedly Prepping a 16K Core Radeon RX 6000 GPU to Fight the NVIDIA RTX 4090?

AMD is reportedly keeping a backup option ready in its RDNA 3 arsenal in case the Radeon RX 7900 XT fails to keep up with NVIDIA’s RTX 4090. According to Greymon55, there’s a 16,384 core (8 shader engines) part based on the RDNA 3 architecture but it’s unclear whether it’s meant for gaming or professional users. Keep in mind that this is not the Navi 31 die, and represents a completely different graphics core. It could be an MCM design with two GCDs but a monolithic GCD is the more probable scenario. In case it does exist, it should feature the same memory bus and Infinity Cache (L3) as the RX 7900 XT, and perhaps 32GB of GDDR6 memory?

The Radeon RX 7900 XT (Navi 31) will feature a single 5nm GCD consisting of 12,288 shaders across 96 CUs and 48 WGPs. A total of 384MB of L3 “Infinity” Cache will be distributed across the six MCDs and paired with the memory controllers. The memory bus is said to be an impressive 384-bit wide, a first for the RDNA family. As expected from the N5 process, the boost clocks should easily hit 3GHz. And finally, the TDP should also increase, though not as much as Lovelace. A 350W TGP for the RX 7900 XT is my guess.

It’s unlikely that AMD would design an entirely new GPU just for one high-end SKU but it’s possible that this will be shared with the Radeon Pro lineup as well. The professional rendering GPUs can never have too much compute performance, and 16K cores should be just the thing content creators need. Still, considering that we’ve heard very little about this part, I’m still skeptical about it actually existing outside of blueprints.

Areej

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.