In possibly just mere weeks or even days before the announcement, NVIDIA’s upcoming flagship gaming GPU, the GeForce RTX 5090, has apparently been leaked – the images surfaced online thus far shows the board design and the silicon itself, designated as GB202-300.
Meet RTX 5090’s Massive GB202 Silicon
The GB202-300 silicon that is expected to power the next-gen flagship is demonstrably huge: according to Videocardz, the silicon measures around 24mm x 21mm, which accounts for 744mm² of die area. That figure is substantially larger than RTX 3090 and RTX 4090’s silicon (GA102 and AD102, respectively), which measures just north of 600mm² – though it’s just shy of the massive silicon that powers the RTX 2080 Ti, the Turing-based TU102 chip that measures at 754mm².
Based on the leaks, the GB202-300 silicon is expected to pack 21,760 CUDA cores – more than double of the RTX 5080 GPU also reportedly launching soon. It’ll be coupled with 32GB of GDDR7 VRAM manufactured by Samsung, and here we can see there are sixteen modules each packing 2GB and provides 28Gbps of bandwidth. In total, the overall memory bandwidth is expected to reach nearly 1.8TB/s.
The RTX 50 series (codenamed Blackwell) is expected to be announced in CES 2025, which kicks off in January 6, 2025. As a side note – rumors suggest that RTX 5080 will be launched first, with RTX 5090 set to follow shortly after, though no specific date has been confirmed at this point.
Pokdepinion: Perhaps the biggest question for most people at this point isn’t how fast it’ll be, but rather how expensive it’ll be.
