Nvidia’s RTX 5090 to Reportedly Feature Monolithic Blackwell Die

Nvidia‘s upcoming GeForce RTX 5090 GPU will use a large monolithic GB202 “Blackwell” die, according to recent reports. This significant development follows Nvidia’s decision to utilize a chiplet design for its HPC and AI GPUs but retain a monolithic design for its gaming GPUs.

Key Features and Expectations

The RTX 5090 is expected to feature next-gen GDDR7 memory with up to 32GB capacity and a 512-bit interface. This upgrade aims to boost performance and solidify Nvidia’s dominance in the gaming market. Rumors also suggest that the GPU will retain the 128 cores per SM design from the RTX 4090’s Ada Lovelace GPU, potentially offering up to 24,567 CUDA cores across 192 SMs.

Nvidia might also introduce new technologies with this generation, possibly including DLSS 4. Enhancements in efficiency and ray tracing performance are anticipated, along with solutions to address issues related to the 12VHPWR connector.

Comparison with Competitors

The monolithic GB202 Blackwell GPU design contrasts with Nvidia’s chiplet-based HPC and AI GPUs. This approach may offer better inter-die communication and scalability. The GB202 GPU is expected to utilize the TSMC 4NP process node, which should improve density by 30%, further enhancing performance.

Nvidia’s decision to maintain a monolithic design for the RTX 5090 might give it a significant edge over competitors. With AMD rumored to be stepping back from the ultra-high-end graphics performance segment, Nvidia could further extend its lead in the gaming GPU market.

Launch Timeline

The GeForce RTX 5090 is expected to launch a few weeks after the RTX 5080, which will be the first Blackwell gaming GPU to hit the market. This strategic release plan indicates Nvidia’s commitment to maintaining its position as a leader in the gaming industry.

Nvidia’s GeForce RTX 5090 is shaping up to be a powerhouse in the gaming world, with its advanced features and high performance expected to set new standards in the industry.