NVIDIA’s next-generation flagship graphics card, the GeForce RTX 5090, is generating immense buzz, and recent leaks paint a picture of an absolute behemoth. Information shared on X by the leaker “MEGAsizeGPU” reveals a card that dwarfs even the current king, the RTX 4090.
The heart of the RTX 5090 is the GB202 “Blackwell” GPU, a chip boasting a reported die size of 744mm². This represents a significant 22% increase compared to the AD102 “Ada Lovelace” GPU found in the RTX 4090 (609mm²), signifying a substantial leap in processing power. The package size is equally impressive, jumping 35% to approximately 3528mm² (63x56mm) compared to the RTX 4090’s 2601mm² (51x51mm). While this might seem enormous, it’s important to remember the TU102 “Turing” GPU, at 754mm², paved the way for Tensor and RT Cores in GeForce gaming GPUs, with subsequent generations showcasing refined optimization and reduced die sizes.
But the sheer size isn’t the only impressive aspect. Rumors suggest the GB202 packs a staggering 21,760 CUDA cores—a considerable upgrade from the RTX 4090’s 16,384. This raw processing power is further amplified by the upgraded Blackwell architecture, offering substantial performance improvements over the Ada Lovelace architecture. Memory is another area of significant enhancement. NVIDIA is set to utilize the cutting-edge GDDR7 standard, equipping the RTX 5090 with a massive 32GB of GDDR7 memory. This is a major leap forward from the RTX 4090’s 24GB of GDDR6X, promising smoother, higher-resolution gaming experiences.
The rumored 512-bit memory bus translates to a breathtaking potential memory bandwidth of up to 2.0TB/s (assuming 32Gbps GDDR7 modules). This is double the 1.0TB/s bandwidth of the RTX 4090, enabling seamless handling of massive textures and data-intensive applications. Power consumption is also significantly higher, with a projected 600W TDP—a 150W increase over the RTX 4090’s 450W. While some high-end, overclocked RTX 4090 models already approach 600W, the RTX 5090’s power draw suggests a new level of performance capability, likely exceeding 600W in custom AIB models.
Performance expectations are sky-high. Early predictions suggest a 50-70% performance boost across the board compared to the RTX 4090, particularly noticeable at 4K resolution. 4K 120FPS gaming will become even more commonplace, and the RTX 5090 is poised to be the go-to card for future 4K 240FPS gaming. Ray tracing performance is anticipated to see a massive improvement – 2-3x gains or even more are projected in certain scenarios, powered by the new Blackwell architecture.
The addition of DLSS 4, an AI-powered upscaling technology exclusive to the RTX 50 series, promises even higher image quality and performance boosts compared to DLSS 3.x. Finally, despite the increased power draw, NVIDIA aims for significantly improved performance-per-watt compared to the current generation, making the RTX 5090 a powerful yet potentially more efficient choice. The upcoming GeForce RTX 5090 promises to redefine high-end gaming, and the wait to see its capabilities unleashed is certainly a thrilling one.
