NVIDIA is stepping up its game with the upcoming Blackwell GB202 GPU, poised to power the new GeForce RTX 5090. This flagship GPU is set to be significantly larger than its predecessor, the Ada-based RTX 4090, promising impressive performance gains while utilizing TSMC’s advanced 5nm process.
The GeForce RTX 5090 will be the crown jewel of NVIDIA’s RTX 50 "Blackwell" lineup, anticipated to deliver substantial improvements over the RTX 4090. Interestingly, this advancement comes with a notable increase in die size. As MEGAsizeGPU highlighted on Twitter, the GB202 GPU in the RTX 5090 will have a 744mm² die size, measuring approximately 24x31mm. In comparison, the RTX 4090’s AD102 had a die size of 609mm². The packaging area has also grown to 3528mm² from 2601mm² in the previous model, marking a 22% enlargement in die size and a 35% increase in the overall packaging area.
While a bigger die doesn’t necessarily mean higher power consumption or heat production, it’s worth noting that the die size of the RTX 5090 is still smaller than the 754mm² Turing TU102 chip. This is noteworthy because it marked the debut of Tensor and RT core technology on NVIDIA’s GeForce "Gaming" GPUs. Since then, advancements have enabled a more compact die size with each new process node.
Focusing on the specifications, the RTX 5090 stands out with a PG144/145-SKU30 PCB design incorporating the GB202-300-A1 GPU core. It will enable 170 of the 192 Streaming Multiprocessors (SMs), translating to 21,760 cores—an 11.4% reduction from its full die, slightly higher than the RTX 4090’s 11.1% reduction.
When it comes to memory, the NVIDIA GeForce RTX 5090 is no slouch. It will boast 32 GB of GDDR7 VRAM on a 512-bit interface, capable of running at 28 Gbps for a considerable 1792 GB/s of bandwidth. Boosted by a larger L3 cache and enhanced memory compression, the GPU will achieve even better bandwidth delivery. This card is designed with a 600W Total Board Power (TBP) rating. However, it’s important to keep in mind that this figure doesn’t directly translate to actual power draw during real-world usage, where it’s typically lower. Speculations suggest the Founders Edition will feature a two-slot cooler.
Set to make its debut at CES 2025, the NVIDIA GeForce RTX 5090 will likely reveal more intriguing details in the months to come. For now, gamers and tech enthusiasts alike are eagerly anticipating NVIDIA’s next big leap in graphics technology.
NVIDIA GeForce RTX 50 Series Preliminary Specifications:
| Graphics Card | 5090 | 5080 | 5070 | 5060 |
|---|---|---|---|---|
| GPU Name | Blackwell GB202-300 | Blackwell GB203-400 | Blackwell GB205 | Blackwell GB206? |
| SMs | 170 (192 Full) | 84 (84 Full) | TBD | TBD |
| Cores | 21,760 (+33%) | 10,752 (+11%) | TBD | TBD |
| Clock Speeds | TBD | TBD | TBD | TBD |
| L2 Cache | TBD | TBD | TBD | TBD |
| Memory Capacity | 32 GB GDDR7 (+33%) | 16 GB GDDR7 (0%) | 12 GB GDDR7 | TBD |
| Memory Bus | 512-bit (+33%) | 256-bit (0%) | 192-bit (0%) | 128-bit? (0%) |
| Memory Speed | 28 Gbps | 32 Gbps | 28 Gbps | 28 Gbps? |
| Bandwidth | 1792 GB/s | 1024 GB/s | 672 GB/s | 448 GB/s |
| TBP | 600W (+33%) | 400W (+25%) | 250W (+14%) | TBD |
| Power Interface | 1 12V-2×6 (16-Pin) | 1 12V-2×6 (16-Pin) | 1 12VHPWR (16-Pin) | 1 12VHPWR (16-Pin) |
As NVIDIA fans anticipate the launch, the big question remains: Which GeForce RTX 50 GPU will grab your attention the most? Keep the conversation going and share your thoughts on social media!
