Rumors swirling around the upcoming NVIDIA GeForce gaming GPUs suggest that the highly anticipated RTX 5080 and 5090 might not be as groundbreaking as initially thought. According to reputable tipster kopite7kimi, the next-gen GPUs with codenames GB202, GB203, GB204, and GB206 are expected to maintain a similar Graphics Processing Cluster (GPC) configuration to their Ada Lovelace predecessors.
The AD103 GPU, which powers the RTX 4080 and 4080 Super, is equipped with 7 GPCs, each consisting of multiple Streaming Multiprocessors (SMs) housing FP32 cores. This results in a total of 10240 FP32 cores, shaping the core count for the upcoming RTX 5080 to be quite similar.
On the other hand, the flagship RTX 5090 is speculated to boast a slightly higher core count than its predecessor, the RTX 4090. The GB202 die is expected to feature a 12 GPCs x 8 TPCs configuration, potentially offering a substantial 24576 cores if all units are activated, showing a glimpse of the performance leap expected from this powerful GPU.
Additionally, sources hint at a memory upgrade for the RTX 5090, featuring a 448-bit memory bus compared to the 384-bit on the RTX 4090. This enhancement is projected to result in a significant boost in external bandwidth to 1568 GB/s, showcasing a remarkable advancement in memory speed and capacity for the next-gen lineup.
The RTX 5080, on the other hand, is anticipated to maintain a similar memory configuration to the 4080 Super but with the addition of faster GDDR7 28 Gbps memory, promising a higher peak bandwidth of 896 GB/s, elevating the GPU’s performance credentials.
