NVIDIA's next-gen GeForce RTX 5090 features 'physically monolithic' GB202 GPU die

NVIDIA's next-generation flagship GeForce RTX 5090 graphics card will sport the "physically monolithic" GB202 "Blackwell" GPU die, according to the latest leaks.

We should expect the GB202 GPU to power the GeForce RTX 5090 with up to 192 SMs that will see a monster 24,567 CUDA cores (if the GB202 chip retains its 128 cores per SM design from the AD102 "Ada Lovelace" GPU inside of the RTX 4090). In a new reply on X, leaker "kopite7kimi" said that the GB202 GPU is a "physically monolithic" chip.

The reason this is important, is that NVIDIA used a chiplet design for its HPC and AI GPUs including the Blackwell B100 and B200 AI GPUs, but looks to use a physically monolithic GPU design for its gaming-focused GeForce RTX 50 series GPUs. We should also see NVIDIA rolling out next-generation GDDR7 memory on its RTX 50 series GPUs, with the RTX 5090 rumored to feature a hefty 32GB of ultra-fast GDDR7 memory.

• Read more: NVIDIA GeForce RTX 5090 rumor: 16 next-gen GDDR7 memory modules, 32GB GDDR7
• Read more: NVIDIA GeForce RTX 5090 Founders Edition PCB reportedly split into three parts

3

GB203 details: NVIDIA's new GB203 "Blackwell" GPU is expected to feature a similar CUDA core count to the AD103 "Ada Lovelace" GPU inside of the RTX 4080, so we should have 96 SMs (Streaming Multiprocessors) and 12288 CUDA cores. We should have performance upgrades with the Blackwell GPU architecture, so expect 40-50% more performance over the AD103-based RTX 4080 and RTX 4080 SUPER.

GDDR7 details: Samsung has its next-gen GDDR7 memory ready for NVIDIA, with GDDR7 memory speeds of 28Gbps and 32Gbps. We could see the faster 32Gbps GDDR7 on the RTX 5090 and 28Gbps GDDR7 on the RTX 5080, and probably on the rest of the RTX 50 series GPU fleet.

• Read more: NVIDIA GeForce RTX 5090 and RTX 5080 will be available within a few weeks of each other
• Read more: NVIDIA testing 250W to 600W coolers for next-gen GeForce RTX 50 series GPUs
• Read more: NVIDIA's next-gen GeForce RTX 5080 rumored to now launch BEFORE the RTX 5090
• Read more: NVIDIA's next-gen RTX 5090, RTX 5080 Laptop GPUs rumored with 16GB VRAM
• Read more: NVIDIA rumored to ONLY launch GeForce RTX 5090 this year
• Read more: NVIDIA's next-gen GeForce RTX 5090, RTX 5080 rumored for Computex 2024 reveal
• Read more: NVIDIA to unveil top-end GeForce RTX 5090, and high-end RTX 5080 at first later this year

4K performance: NVIDIA's next-generation ultimate gaming GPU in the GeForce RTX 5090 should have 50-70% more performance across the board compared to the RTX 4090, especially at the higher-end 4K resolution. 4K 120FPS gaming should be an even easier achievement for the RTX 5090 than it is for the RTX 4090. We've got 4K 240FPS coming this year and in 2025, so the RTX 5090 will be the GPU of choice for 4K 240FPS gaming in the future.

3

RT performance: This is where the biggest performance improvements of the next-gen Blackwell-based GeForce RTX 50 series GPUs will come: ray tracing. Expect some rather large 2-3x performance gains using RT, probably 4x or more in some cases with the RTX 5090 and RTX 5080 against their RTX 40 series counterparts. Throw new DLSS 4 on top, and you've got some wowzers RT performance.

DLSS 4: This is probably my personal favorite part of the excitement of next-gen GPUs from NVIDIA, AI-powered upscaling with a next-gen DLSS 4 that works only on the new Blackwell-based RTX 50 series GPUs. We should expect even higher image quality than offered by DLSS 3.x and new levels of performance with DLSS 4 enabled on a new RTX 5090 or RTX 5080 graphics card.

Power efficiency: NVIDIA's current-gen GeForce RTX 4090 can use anywhere between 450W and 600W of power depending on the model and overclocking, but the new GeForce RTX 50 series GPUs will offer far more performance-per-watt of the leading RTX 4090 and RTX 4080 SUPER graphics cards. Another exciting part to see unravel in the near future.